Online First
Articles in press have been peer-reviewed and accepted, which are not yet assigned to volumes/issues, but are citable by Digital Object Identifier (DOI).
Display Method:
, Available online
Abstract:
In order to meet the application requirements of high-definition fish eye lens in the information age, this paper combines CODE V and ZEMAX optical software to design a large aperture glass plastic hybrid fish eye lens optical system matching 1 INCH CCD. The system can clearly image in the range of 486 nm-656 nm and 850 nm. The field angle is 210 °, the f-number is 2.0, the focal length is 4.1 mm, the f-theta distortion is less than 7%, and the edge illumination is more than 68%. At normal temperature, the MTF of 0.707 field of view is greater than 0.5 and the MTF of the full field of view is greater than 0.35 at the Nyquist frequency of 91 lp/mm. At 40 ℃ ~ + 75 ℃, the MTF of 0.707 field of view is greater than 0.3, and the MTF of the full field of view is greater than 0.2, which can meet the requirements of high and low temperature environment. The design adopts a 7-piece glass plastic hybrid structure, which has the characteristics of large field angle, large target surface and large aperture. It can be widely used in high-definition camera, security monitoring, industrial application and other fields.
In order to meet the application requirements of high-definition fish eye lens in the information age, this paper combines CODE V and ZEMAX optical software to design a large aperture glass plastic hybrid fish eye lens optical system matching 1 INCH CCD. The system can clearly image in the range of 486 nm-656 nm and 850 nm. The field angle is 210 °, the f-number is 2.0, the focal length is 4.1 mm, the f-theta distortion is less than 7%, and the edge illumination is more than 68%. At normal temperature, the MTF of 0.707 field of view is greater than 0.5 and the MTF of the full field of view is greater than 0.35 at the Nyquist frequency of 91 lp/mm. At 40 ℃ ~ + 75 ℃, the MTF of 0.707 field of view is greater than 0.3, and the MTF of the full field of view is greater than 0.2, which can meet the requirements of high and low temperature environment. The design adopts a 7-piece glass plastic hybrid structure, which has the characteristics of large field angle, large target surface and large aperture. It can be widely used in high-definition camera, security monitoring, industrial application and other fields.
, Available online
Abstract:
The characteristics of traditional direct and oblique laser displacement sensor are analyzed. Combined with their advantages, a compact laser displacement sensor based on auto collimation is designed. In the structure, the linear structure is used to replace the light path layout of the large triangle, the auto collimating optical path structure is formed. A beam splitter with a beam splitting ratio of 50% is introduced and placed at an angle of 45° with the optical axis to combine the focusing lens and imaging lens into one, and make the condenser, beam splitter and photo detector coaxial, simplify the system structure, reduce the volume of the instrument and compact structure. At the same time, the Scheimpflug condition conforming to the principle of auto collimation. The optical design software Zemax is used to simulate the optical system. The system adopts the structure of single lens, the front surface is Forbes aspheric, and the entrance pupil diameter\begin{document}$ D=4\;\text{mm} $\end{document} ![]()
![]()
\begin{document}$ \text{2}\omega =\text{4}^{\circ} $\end{document} ![]()
![]()
\begin{document}$L = 20\;{\text{mm}}$\end{document} ![]()
![]()
The characteristics of traditional direct and oblique laser displacement sensor are analyzed. Combined with their advantages, a compact laser displacement sensor based on auto collimation is designed. In the structure, the linear structure is used to replace the light path layout of the large triangle, the auto collimating optical path structure is formed. A beam splitter with a beam splitting ratio of 50% is introduced and placed at an angle of 45° with the optical axis to combine the focusing lens and imaging lens into one, and make the condenser, beam splitter and photo detector coaxial, simplify the system structure, reduce the volume of the instrument and compact structure. At the same time, the Scheimpflug condition conforming to the principle of auto collimation. The optical design software Zemax is used to simulate the optical system. The system adopts the structure of single lens, the front surface is Forbes aspheric, and the entrance pupil diameter
, Available online
Abstract:
In order to solve the problem of color inconsistency between the projection pictures of various projectors in a multi-channel projection display system, a multi projection color correction method based on free deformation technology is designed. Firstly, the free-form deformation model is established by Bernstein basis function, the color conversion relationship between the original image of each projector and the image taken by the camera is established, and the parameters of the free-form deformation model are determined. Then, collect the original image set and the projection display picture set taken by the camera to determine the parameters of the free-form deformation technology model. Then, through MATLAB analysis, it is verified that each color channel of projection image is affected by each other. Finally, the original image is color distorted, and the color correction method in this paper is compared with the traditional method for histogram similarity evaluation.The experimental results show that, compared with the generalized color correction method and the color correction method based on B-spline curve, the method in this paper reduces the average difference of color intensity of the projected image by 2.50 in the B channel, 2.34 in the G channel, and 3.57 in the R channel. The correlation of histogram is increased by 8.9%, and the Pap distance is reduced by 9.7%.The multi projection color correction method based on free-form deformation makes the projection images smoothly connected, giving users a better immersive feeling.
In order to solve the problem of color inconsistency between the projection pictures of various projectors in a multi-channel projection display system, a multi projection color correction method based on free deformation technology is designed. Firstly, the free-form deformation model is established by Bernstein basis function, the color conversion relationship between the original image of each projector and the image taken by the camera is established, and the parameters of the free-form deformation model are determined. Then, collect the original image set and the projection display picture set taken by the camera to determine the parameters of the free-form deformation technology model. Then, through MATLAB analysis, it is verified that each color channel of projection image is affected by each other. Finally, the original image is color distorted, and the color correction method in this paper is compared with the traditional method for histogram similarity evaluation.The experimental results show that, compared with the generalized color correction method and the color correction method based on B-spline curve, the method in this paper reduces the average difference of color intensity of the projected image by 2.50 in the B channel, 2.34 in the G channel, and 3.57 in the R channel. The correlation of histogram is increased by 8.9%, and the Pap distance is reduced by 9.7%.The multi projection color correction method based on free-form deformation makes the projection images smoothly connected, giving users a better immersive feeling.
, Available online
Abstract:
The whispering gallery mode resonator have a wide variety of potential application in microwave photonics、nonlinear optics and quantum optics, due to their ultra-high Q factor and small model volume. The loss of whispering gallery mode resonator is discussed. The key technology affecting whispering gallery mode resonator Q factor was the fabrication process. The main limiting factors were material classes and surface roughness. The structure of whispering gallery mode resonator was designed with DUV MgF2 crystal assuming the roughness of whispering gallery mode resonator less than 0.7 nm. The limit loss of whispering gallery mode resonator is 4.09×10−10 and the maximum Q factor of whispering gallery mode resonator is 2.44×109 according to the theoretical calculation. The fabrication process of MgF2 crystal whispering gallery mode resonator was rough machining, ultra precision turning and precision polishing. The manufacture of ultra-high Q whispering gallery mode resonators are realized. The test results show that the Q value of whispering gallery mode resonator is 2.054×109@1 550 nm by Q value measurement system, the surface roughness of whispering gallery mode resonator is 0.669 nm(Ra value) by whit light interferometer, and the shape error of whispering gallery mode resonator is 6.767 nm(PV value) by whit light interferometer.
The whispering gallery mode resonator have a wide variety of potential application in microwave photonics、nonlinear optics and quantum optics, due to their ultra-high Q factor and small model volume. The loss of whispering gallery mode resonator is discussed. The key technology affecting whispering gallery mode resonator Q factor was the fabrication process. The main limiting factors were material classes and surface roughness. The structure of whispering gallery mode resonator was designed with DUV MgF2 crystal assuming the roughness of whispering gallery mode resonator less than 0.7 nm. The limit loss of whispering gallery mode resonator is 4.09×10−10 and the maximum Q factor of whispering gallery mode resonator is 2.44×109 according to the theoretical calculation. The fabrication process of MgF2 crystal whispering gallery mode resonator was rough machining, ultra precision turning and precision polishing. The manufacture of ultra-high Q whispering gallery mode resonators are realized. The test results show that the Q value of whispering gallery mode resonator is 2.054×109@1 550 nm by Q value measurement system, the surface roughness of whispering gallery mode resonator is 0.669 nm(Ra value) by whit light interferometer, and the shape error of whispering gallery mode resonator is 6.767 nm(PV value) by whit light interferometer.
, Available online
Abstract:
Apparent distance is an important parameter to evaluate the performance of low-level-light (LLL) night vision imaging system. With the development of LLL night vision detection technology, the simulation results of the classical apparent distance model show some deviations from the actual measurement data, especially the simulation results are not ideal under the low illumination of 10-3 lx, which causes great obstacles to the practical application of the LLL night vision system. Aiming at this problem, the classical apparent distance detection model was modified from three aspects: the first is considering the influence of atmospheric transmittance on the apparent distance of LLL night vision system and modifying the atmospheric transmittance factors in the classical apparent distance model, the second is optimizing the apparent distance model based on noise factors of image intensifier, the third is considering the influence of human visual transmission characteristics on the apparent distance of LLL night vision system, and the simplified human visual system was added into the transfer function model of the system. The improved apparent distance model was derived, and its effectiveness as well as practicability were verified by the field test data, which had certain guiding significance for the design, evaluation and application of LLL night vision system.
Apparent distance is an important parameter to evaluate the performance of low-level-light (LLL) night vision imaging system. With the development of LLL night vision detection technology, the simulation results of the classical apparent distance model show some deviations from the actual measurement data, especially the simulation results are not ideal under the low illumination of 10-3 lx, which causes great obstacles to the practical application of the LLL night vision system. Aiming at this problem, the classical apparent distance detection model was modified from three aspects: the first is considering the influence of atmospheric transmittance on the apparent distance of LLL night vision system and modifying the atmospheric transmittance factors in the classical apparent distance model, the second is optimizing the apparent distance model based on noise factors of image intensifier, the third is considering the influence of human visual transmission characteristics on the apparent distance of LLL night vision system, and the simplified human visual system was added into the transfer function model of the system. The improved apparent distance model was derived, and its effectiveness as well as practicability were verified by the field test data, which had certain guiding significance for the design, evaluation and application of LLL night vision system.
, Available online
Abstract:
Aiming at the problem that transmittance parameters of infrared optical system cannot be accurately measured, the measurement and calibration technology for spectral transmittance of infrared optical system were studied. Based on the analysis of measurement methods such as integrating sphere method and large-area uniform source method, a calibration device for spectral transmittance of infrared optical system based on method of reflection was studied and constructed, and the complete quantity traceability and quantity transmission chain of spectral transmittance were established. The device was used to measure the spectral transmittance in the wavelength range of 2 µm~14 µm, and the uncertainty analysis of the measurement results showed that the result was 2%. Compared with the previous measurement device, the measurement results of the proposed device have higher accuracy and reliability.
Aiming at the problem that transmittance parameters of infrared optical system cannot be accurately measured, the measurement and calibration technology for spectral transmittance of infrared optical system were studied. Based on the analysis of measurement methods such as integrating sphere method and large-area uniform source method, a calibration device for spectral transmittance of infrared optical system based on method of reflection was studied and constructed, and the complete quantity traceability and quantity transmission chain of spectral transmittance were established. The device was used to measure the spectral transmittance in the wavelength range of 2 µm~14 µm, and the uncertainty analysis of the measurement results showed that the result was 2%. Compared with the previous measurement device, the measurement results of the proposed device have higher accuracy and reliability.
, Available online
Abstract:
DETR is a target detection algorithm based on transformer, which has the advantages of fast detection speed and good detection effect. A measurement system based on DETR and binocular vision principle for people, cars, bicycles, signal lights and other targets in road environment is introduced. The principles of binocular ranging, camera calibration, target detection and target matching are analyzed, and the measurement system is constructed on this basis. The target detection algorithm is used to detect the target in the field of vision, and the principle of binocular vision is used to measure the distance of the detected target. The source of measurement error in the measurement system is analyzed and the influence on the results is calculated. The algorithm is tested in Kitti data set and real environment. The system baseline is 45 cm, the detection rate of 15-80 m specified target is higher than 90.6%, and the ranging error is less than 5.8%. The algorithm in this paper runs on RTX 2080ti platform and can run in real time.
DETR is a target detection algorithm based on transformer, which has the advantages of fast detection speed and good detection effect. A measurement system based on DETR and binocular vision principle for people, cars, bicycles, signal lights and other targets in road environment is introduced. The principles of binocular ranging, camera calibration, target detection and target matching are analyzed, and the measurement system is constructed on this basis. The target detection algorithm is used to detect the target in the field of vision, and the principle of binocular vision is used to measure the distance of the detected target. The source of measurement error in the measurement system is analyzed and the influence on the results is calculated. The algorithm is tested in Kitti data set and real environment. The system baseline is 45 cm, the detection rate of 15-80 m specified target is higher than 90.6%, and the ranging error is less than 5.8%. The algorithm in this paper runs on RTX 2080ti platform and can run in real time.
, Available online
Abstract:
The detection of concealed targets has always been a key military research field. With the development of hyperspectral imaging technology, a new solution is provided for this field. The high spectral resolution of hyperspectral data can be used to distinguish the hidden target from the background in some wavebands. The hyperspectral covert target detection technology based on ACE algorithm proposed in this paper makes full use of the different spectral characteristics of background and covert target in different wavebands, and introduces spectral rearrangement technology and first-order differential technology to make the background spectrum oscillate, so as to extract the target. Compared with other algorithms, this algorithm has higher extraction accuracy, less interference and better detection effect of hidden targets.
The detection of concealed targets has always been a key military research field. With the development of hyperspectral imaging technology, a new solution is provided for this field. The high spectral resolution of hyperspectral data can be used to distinguish the hidden target from the background in some wavebands. The hyperspectral covert target detection technology based on ACE algorithm proposed in this paper makes full use of the different spectral characteristics of background and covert target in different wavebands, and introduces spectral rearrangement technology and first-order differential technology to make the background spectrum oscillate, so as to extract the target. Compared with other algorithms, this algorithm has higher extraction accuracy, less interference and better detection effect of hidden targets.
, Available online
Abstract:
In this paper, the modulation phase is solved based on the geometric steady-phase method, and the Fourier transform is carried out through the Fourier lens, and the corresponding flat-top beam is focused at the rear focal plane, so that the circular Gaussian beam is shaped into a square flat-top beam. The optical path parameters will have a great influence on the shaping quality. In the experiment, it is found that the square flat-top beam is affected by the zero-order light and has poor uniformity. Therefore, the causes of the zero-order light are analyzed in detail and effective solutions are proposed. The zero-order light in the experimental results is directly removed by stacking the shining grating. The experimental results show that the energy utilization rate of square flat-topped beam is 72.3%, and the beam uniformity is up to 97.2%, which is better than the traditional shaping method. Furthermore, the quality of beam shaping by stationary phase method are further analyzed for three factors: the offset of hologram, defocus and waist radius of incident beam. The research provides convenience for the application of beam shaping by geometric shaping method.
In this paper, the modulation phase is solved based on the geometric steady-phase method, and the Fourier transform is carried out through the Fourier lens, and the corresponding flat-top beam is focused at the rear focal plane, so that the circular Gaussian beam is shaped into a square flat-top beam. The optical path parameters will have a great influence on the shaping quality. In the experiment, it is found that the square flat-top beam is affected by the zero-order light and has poor uniformity. Therefore, the causes of the zero-order light are analyzed in detail and effective solutions are proposed. The zero-order light in the experimental results is directly removed by stacking the shining grating. The experimental results show that the energy utilization rate of square flat-topped beam is 72.3%, and the beam uniformity is up to 97.2%, which is better than the traditional shaping method. Furthermore, the quality of beam shaping by stationary phase method are further analyzed for three factors: the offset of hologram, defocus and waist radius of incident beam. The research provides convenience for the application of beam shaping by geometric shaping method.
, Available online
Abstract:
Vertical-scanning white light interferometry (VSWLI) is a non-contact three-dimensional surface profile measurement method. However, batwings make it challenging to obtain a satisfactory result when the step height of the sample under measurement is smaller than the coherence length of the light source. The phase-shifting interferometry does not suffer from it but has phase-ambiguity problems. In this paper, a white-light interferometric demodulation algorithm that combines Carré equal-step phase-shifting algorithm with fast Fourier transform (FFT) coherence-peak-sensing technique was proposed to overcome the above problem. This algorithm is based on successive variational mode decomposition (SVMD) combined with Hausdorff distance (HD) denoising. A micro-component sample with 500 nm and 1 200 nm steps and a standard step with 10 µm height were used as test specimens for experimental verification. The proposed algorithm can effectively reduce the batwings effect at the step edges, removing the phase-ambiguity problems and improving the measurement precision of VSWLI.
Vertical-scanning white light interferometry (VSWLI) is a non-contact three-dimensional surface profile measurement method. However, batwings make it challenging to obtain a satisfactory result when the step height of the sample under measurement is smaller than the coherence length of the light source. The phase-shifting interferometry does not suffer from it but has phase-ambiguity problems. In this paper, a white-light interferometric demodulation algorithm that combines Carré equal-step phase-shifting algorithm with fast Fourier transform (FFT) coherence-peak-sensing technique was proposed to overcome the above problem. This algorithm is based on successive variational mode decomposition (SVMD) combined with Hausdorff distance (HD) denoising. A micro-component sample with 500 nm and 1 200 nm steps and a standard step with 10 µm height were used as test specimens for experimental verification. The proposed algorithm can effectively reduce the batwings effect at the step edges, removing the phase-ambiguity problems and improving the measurement precision of VSWLI.
, Available online
Abstract:
Aiming at the problems of unsatisfactory processing effect and long processing time in single frame infrared image with complex background, a hierarchical convolution filter detection algorithm is proposed. It is mainly divided into two parts: first, according to the characteristics of small infrared targets, a hierarchical convolution filter operator is designed to filter the image, so as to achieve the effect of increasing the efficiency of small targets in the image and suppressing the background; Secondly, the adaptive threshold method based on the maximum value is used to binarize the image to filter the background clutter, and finally extract the target to be detected. Experiments in a large number of infrared images with different backgrounds show that the performance quantization results of the algorithm are better than the performance results of the existing five typical infrared dim and small target detection algorithms, and the average processing time is only 30.42% of Laplacian of Gaussian (LoG) filter algorithm. Finally, through the experimental comparison, Hierarchical convolution filtering method can effectively solve the problem of small target detection in infrared images in different complex backgrounds.
Aiming at the problems of unsatisfactory processing effect and long processing time in single frame infrared image with complex background, a hierarchical convolution filter detection algorithm is proposed. It is mainly divided into two parts: first, according to the characteristics of small infrared targets, a hierarchical convolution filter operator is designed to filter the image, so as to achieve the effect of increasing the efficiency of small targets in the image and suppressing the background; Secondly, the adaptive threshold method based on the maximum value is used to binarize the image to filter the background clutter, and finally extract the target to be detected. Experiments in a large number of infrared images with different backgrounds show that the performance quantization results of the algorithm are better than the performance results of the existing five typical infrared dim and small target detection algorithms, and the average processing time is only 30.42% of Laplacian of Gaussian (LoG) filter algorithm. Finally, through the experimental comparison, Hierarchical convolution filtering method can effectively solve the problem of small target detection in infrared images in different complex backgrounds.
, Available online
Abstract:
To achieve high precision 3D measurement of object surface profile, a telecentric measurement system with large FOV (Field of View), WD (Working Distance) and measuring range was proposed. The composition of the system and the principle of optical triangulation were introduced, the construction of the light path with the Angle of 45 ° between the projection and imaging light paths and the normal line of the measured surface can reduce the requirement of DOF (Depth of Field) of the imaging objective lens. The theoretical model of system depth calibration was established, and it was obtained that the object-space depth corresponding to each CCD pixel is 0.5544 mm. The magnification, NA and other optical parameters were calculated and the optical path was constructed. The resolution, measuring range, minimum width of light strip and edge quality of the system were analyzed by experiments. The experimental results show that the length of the projection strip is 1.5 mm, the minimum line width is 8.96 μm, the strip edge is clear and smooth. The imaging distortion of the system is small, and the optical resolution of 1.62 µm and the depth pixel resolution of 0.54 µm can be achieved. The WD is 65 mm, the depth measurement range is no less than 140 µm. The system has good measurement performance.
To achieve high precision 3D measurement of object surface profile, a telecentric measurement system with large FOV (Field of View), WD (Working Distance) and measuring range was proposed. The composition of the system and the principle of optical triangulation were introduced, the construction of the light path with the Angle of 45 ° between the projection and imaging light paths and the normal line of the measured surface can reduce the requirement of DOF (Depth of Field) of the imaging objective lens. The theoretical model of system depth calibration was established, and it was obtained that the object-space depth corresponding to each CCD pixel is 0.5544 mm. The magnification, NA and other optical parameters were calculated and the optical path was constructed. The resolution, measuring range, minimum width of light strip and edge quality of the system were analyzed by experiments. The experimental results show that the length of the projection strip is 1.5 mm, the minimum line width is 8.96 μm, the strip edge is clear and smooth. The imaging distortion of the system is small, and the optical resolution of 1.62 µm and the depth pixel resolution of 0.54 µm can be achieved. The WD is 65 mm, the depth measurement range is no less than 140 µm. The system has good measurement performance.
, Available online
Abstract:
The optical system with large collecting aperture and high resolution is the future development directions of optical system in the sky. Once in the sky, it needs to correct the misalignment errors in optical system introduced by the effect of temperature change and vibration shock. For the purpose of decreasing the complex degree of alignment in the sky and increasing the reliability, the nonlinear function mathematical model between mirror misalignments and Zernike coefficients was established, and the strong correlative alignment variables had eliminatded by means of Bhattacharyya coefficient. Then a certain simulated co-axial three-mirror-anastigmat system was corrected, the result indicated that the wavefront RMS reduced to 0.025 λ, the difference between the corrected value and designed value less than 0.014 λ. It satisfied the demands of alignment in the sky.
The optical system with large collecting aperture and high resolution is the future development directions of optical system in the sky. Once in the sky, it needs to correct the misalignment errors in optical system introduced by the effect of temperature change and vibration shock. For the purpose of decreasing the complex degree of alignment in the sky and increasing the reliability, the nonlinear function mathematical model between mirror misalignments and Zernike coefficients was established, and the strong correlative alignment variables had eliminatded by means of Bhattacharyya coefficient. Then a certain simulated co-axial three-mirror-anastigmat system was corrected, the result indicated that the wavefront RMS reduced to 0.025 λ, the difference between the corrected value and designed value less than 0.014 λ. It satisfied the demands of alignment in the sky.
, Available online
Abstract:
The application of nanofluids in spectral beam splitting (SBS) hybrid PV/T system can improve its efficiency, in which nanoparticles (NPs) with appropriate size can effectively filter sunlight "outside the response band of photovoltaic cells". The optical properties of Au, Ag, Cu, Fe3O4, ZnO and TiO2 NPs were simulated by finite difference time domain (FDTD) method. Moreover, taking spectral response of monocrystal silicon solar cells as example, the optical absorption properties of six NPs with diameters ranging from 20 nm to 200 nm were studied, and degree of appropriateness (DOA) was used as the evaluation index to optimize the particle size. The results show that the optical properties of NPs are very sensitive to their particle size. By changing the particle size of NPs, the positions of scattering, absorption and extinction peaks can be adjusted in a wide range, and the peaks increase with the increase of particle size. The absorption capacity of metal NPs to solar radiation is better than that of non-metal NPs. The maximum absorption power per unit volume of six kinds of NPs are 21.88, 17.95, 20.16, 2.54, 1.02, 0.27 GW/m3, respectively. The DOA index analysis showed that the optimal particle sizes of six NPs suitable for SBS hybrid PV/T system are 20, 50, 20, 170, 110 and 20 nm, respectively.
The application of nanofluids in spectral beam splitting (SBS) hybrid PV/T system can improve its efficiency, in which nanoparticles (NPs) with appropriate size can effectively filter sunlight "outside the response band of photovoltaic cells". The optical properties of Au, Ag, Cu, Fe3O4, ZnO and TiO2 NPs were simulated by finite difference time domain (FDTD) method. Moreover, taking spectral response of monocrystal silicon solar cells as example, the optical absorption properties of six NPs with diameters ranging from 20 nm to 200 nm were studied, and degree of appropriateness (DOA) was used as the evaluation index to optimize the particle size. The results show that the optical properties of NPs are very sensitive to their particle size. By changing the particle size of NPs, the positions of scattering, absorption and extinction peaks can be adjusted in a wide range, and the peaks increase with the increase of particle size. The absorption capacity of metal NPs to solar radiation is better than that of non-metal NPs. The maximum absorption power per unit volume of six kinds of NPs are 21.88, 17.95, 20.16, 2.54, 1.02, 0.27 GW/m3, respectively. The DOA index analysis showed that the optimal particle sizes of six NPs suitable for SBS hybrid PV/T system are 20, 50, 20, 170, 110 and 20 nm, respectively.
, Available online
Abstract:
In order to build a re-detection module suitable for long-term tracking, inspired by the GlobalTrack method which improves two-stage detection network, this paper proposes an efficient deep network for end-to-end re-detection of template targets. First, in order to fuse the template features with large-scale search features more efficiently, we improve the depth-wise correlation method by constructing a cross-information enhancement module which encodes the information of search and template features with cross channel-attention information; In addition, we replace RPN and RCNN structure of traditional two-stage detection network with a dynamic instance interaction module, guiding the classification-and-regression stage of the detection network with template information as well as building an end-to-end re-detection structure. Comparing results on LaSOT and OxUva long-term tracking datasets, the performance of this method is improved by 3%, and the real-time frame rate is improved by 173% compared with the original method. The experimental results show that our method can re-detect template targets more accurately and quickly in the whole image range.
In order to build a re-detection module suitable for long-term tracking, inspired by the GlobalTrack method which improves two-stage detection network, this paper proposes an efficient deep network for end-to-end re-detection of template targets. First, in order to fuse the template features with large-scale search features more efficiently, we improve the depth-wise correlation method by constructing a cross-information enhancement module which encodes the information of search and template features with cross channel-attention information; In addition, we replace RPN and RCNN structure of traditional two-stage detection network with a dynamic instance interaction module, guiding the classification-and-regression stage of the detection network with template information as well as building an end-to-end re-detection structure. Comparing results on LaSOT and OxUva long-term tracking datasets, the performance of this method is improved by 3%, and the real-time frame rate is improved by 173% compared with the original method. The experimental results show that our method can re-detect template targets more accurately and quickly in the whole image range.
, Available online
Abstract:
In view of the low accuracy of the current bridge disease detection algorithm based on convolutional neural network, an improved YOLOX algorithm is proposed to improve the detection accuracy. By using the feature information of the shallow layer of the backbone network, the feature extraction enhancement network is improved, and the feature information of the same layer is added for fusion; an improved coordinate attention mechanism is introduced to combine the position information and the channel information to enhance the network's ability to detect bridge diseases. Recognition; at the same time, the localization loss function has been improved. The experimental results show that the accuracy of the improved YOLOX network structure for bridge disease detection reaches 92.11%, which is 4.40% higher than the original network.
In view of the low accuracy of the current bridge disease detection algorithm based on convolutional neural network, an improved YOLOX algorithm is proposed to improve the detection accuracy. By using the feature information of the shallow layer of the backbone network, the feature extraction enhancement network is improved, and the feature information of the same layer is added for fusion; an improved coordinate attention mechanism is introduced to combine the position information and the channel information to enhance the network's ability to detect bridge diseases. Recognition; at the same time, the localization loss function has been improved. The experimental results show that the accuracy of the improved YOLOX network structure for bridge disease detection reaches 92.11%, which is 4.40% higher than the original network.
, Available online
Abstract:
In order to reduce the volume and energy consumption of the laser communication terminal, and to ensure the aperture and magnification, an improved Dall-Kirkham receiving telescope is designed according to the Gaussian optics and Seidel aberration theory of catadioptric telescopes. The lens consists of two mirrors and five lenses, and the wavelengths of the received signal light are 974 nm and 1550 nm, the aperture diameter is 60 mm, the relative aperture is 1/3.38, the exit pupil distance is 50 mm, the field of view angle is 6 mrad, and the total length is 111.04 mm. The working temperature is 20 ℃±10 ℃. The optimized lens’ wave aberration is better than l/40 λ in the 0.5 mrad field of view, and is better than l/20 λ in the 6 mrad field of view. The energy is concentrated. In order to realize the athermalization design, the selection and design of structural materials are given, and the wave aberration of optical system under ±10 ℃ is analyzed. The results show that the performance of the lens is good and meets the application requirements.
In order to reduce the volume and energy consumption of the laser communication terminal, and to ensure the aperture and magnification, an improved Dall-Kirkham receiving telescope is designed according to the Gaussian optics and Seidel aberration theory of catadioptric telescopes. The lens consists of two mirrors and five lenses, and the wavelengths of the received signal light are 974 nm and 1550 nm, the aperture diameter is 60 mm, the relative aperture is 1/3.38, the exit pupil distance is 50 mm, the field of view angle is 6 mrad, and the total length is 111.04 mm. The working temperature is 20 ℃±10 ℃. The optimized lens’ wave aberration is better than l/40 λ in the 0.5 mrad field of view, and is better than l/20 λ in the 6 mrad field of view. The energy is concentrated. In order to realize the athermalization design, the selection and design of structural materials are given, and the wave aberration of optical system under ±10 ℃ is analyzed. The results show that the performance of the lens is good and meets the application requirements.
, Available online
Abstract:
The three-channel 355 nm optical discriminator is widely used in the frequency discrimination of the backscattered signal of the space borne wind lidar. It is the core component to discriminate the wind speed Doppler frequency shift in the double-edge method. Its parameters and reliabilities determine the detection accuracy of the system. In this paper, a 355 nm three-channel air-gap etalon module based on PZT crystal tuning is developed. The effective diameter of the etalon module is 35 mm, the peak transmittance is 75%, the free spectral range is 12.5 GHz, and the full width at half maximum of the transmission peak is 1.7 GHz. By building up a 355 nm three-channel test system, the parameters including free spectral range, full width at half maximum, peak transmittance, and tuning coefficient are tested. The test results show that when the external driving voltage is 75 V, the peak transmittances of the three channels are 0.859, 0.878, and 0.735, respectively. The full width at half maximum is 1.843 GHz, 1.882 GHz, and 1.611 GHz, respectively. The tuning coefficients are 1.96 GHz/V, 1.93 GHz/V, 1.88 GHz/V. In view of the inconsistent tuning coefficients of the three channels of the PZT crystal of the FP frequency discrimination module, the influence range of the analysis on the wind speed error is ±0.1 m/s. Through the joint test of the closed-loop feedback control system and the Fabry-Perot (FP) frequency discrimination module, the system can realize the real-time locking of the 355 nm laser emission frequency, solve the problem caused by the inconsistent initial position of the FP etalon in each working state, and improve the wind speed. The frequency discrimination accuracy can achieve stable locking time of more than 30 minutes, which meets the application test requirements of space-borne Lidar. Simulation studies show that when the FP etalon interval changes by 0.08 nm, the resulting wind speed error is 1 m/s.
The three-channel 355 nm optical discriminator is widely used in the frequency discrimination of the backscattered signal of the space borne wind lidar. It is the core component to discriminate the wind speed Doppler frequency shift in the double-edge method. Its parameters and reliabilities determine the detection accuracy of the system. In this paper, a 355 nm three-channel air-gap etalon module based on PZT crystal tuning is developed. The effective diameter of the etalon module is 35 mm, the peak transmittance is 75%, the free spectral range is 12.5 GHz, and the full width at half maximum of the transmission peak is 1.7 GHz. By building up a 355 nm three-channel test system, the parameters including free spectral range, full width at half maximum, peak transmittance, and tuning coefficient are tested. The test results show that when the external driving voltage is 75 V, the peak transmittances of the three channels are 0.859, 0.878, and 0.735, respectively. The full width at half maximum is 1.843 GHz, 1.882 GHz, and 1.611 GHz, respectively. The tuning coefficients are 1.96 GHz/V, 1.93 GHz/V, 1.88 GHz/V. In view of the inconsistent tuning coefficients of the three channels of the PZT crystal of the FP frequency discrimination module, the influence range of the analysis on the wind speed error is ±0.1 m/s. Through the joint test of the closed-loop feedback control system and the Fabry-Perot (FP) frequency discrimination module, the system can realize the real-time locking of the 355 nm laser emission frequency, solve the problem caused by the inconsistent initial position of the FP etalon in each working state, and improve the wind speed. The frequency discrimination accuracy can achieve stable locking time of more than 30 minutes, which meets the application test requirements of space-borne Lidar. Simulation studies show that when the FP etalon interval changes by 0.08 nm, the resulting wind speed error is 1 m/s.
, Available online
Abstract:
Structure design is critical for a common telescope. The materials for optical mirror and metering structure of an airborne common telescope operated over ±60 ℃ isotherma difference was selected in consider of matching thermal expansion coefficient. An athermal flexure mount with high stiffness was designed and the mirrors' surface error and the system performance under 60 ℃ isothermal difference、10 ℃ axial thermal gradient and radial thermal gradient was analyzed using a custom integrated opto-mechanical code. The surface errors are less than 1/100λ after power removed for telescope made of Invar and ULE, and the RMS spot radius of the optical system is less than Airy Disk radius , the MTF@63 lp/mm is greater than 0.45. The first mode frequency is 263 Hz. Telescope made of Ti and K9 perform well under 60 ℃ isothermal difference but have a poor performance under 10 ℃ thermal gradient. The test result of the telescope made of Invar and ULE under −40 ℃ show the validity of the design and analysis.
Structure design is critical for a common telescope. The materials for optical mirror and metering structure of an airborne common telescope operated over ±60 ℃ isotherma difference was selected in consider of matching thermal expansion coefficient. An athermal flexure mount with high stiffness was designed and the mirrors' surface error and the system performance under 60 ℃ isothermal difference、10 ℃ axial thermal gradient and radial thermal gradient was analyzed using a custom integrated opto-mechanical code. The surface errors are less than 1/100λ after power removed for telescope made of Invar and ULE, and the RMS spot radius of the optical system is less than Airy Disk radius , the MTF@63 lp/mm is greater than 0.45. The first mode frequency is 263 Hz. Telescope made of Ti and K9 perform well under 60 ℃ isothermal difference but have a poor performance under 10 ℃ thermal gradient. The test result of the telescope made of Invar and ULE under −40 ℃ show the validity of the design and analysis.
, Available online
Abstract:
Using C32 molecule as the skeleton of multipolar moments, two types of substitutional co-doped fullerene derivatives C28B2N2 and C28B2P2 were designed with 16 isomers. The electronic properties, linear polarizability α and first hyperpolarizability β were studied by CAM-B3LYP method of density functional theory (DFT). The results show that the HOMO-LUMO gap of the doped molecules become smaller, and the values of α and β of C28B2P2 are larger than those of C28B2N2. Additionally, dipole molecules in these molecules have large β, while the octapolar molecules have smaller β. The structure with excellent second-order nonlinear optical (NLO) response characteristics is selected. The calculation results of time-dependent density functional theory (TD-DFT) method show that, compared with C32, the response range of absorption spectra of all doped structures becomes wider, the maximum intensity becomes weaker, and the position of the maximum absorption wave may be red-shifted or blue-shifted. Based on the sum-over-states (SOS) method, the two-level or three-level formulas are used to explain the origin of large β in two co-doped structures, and the related electronic excitation type is π → π* excitation.
Using C32 molecule as the skeleton of multipolar moments, two types of substitutional co-doped fullerene derivatives C28B2N2 and C28B2P2 were designed with 16 isomers. The electronic properties, linear polarizability α and first hyperpolarizability β were studied by CAM-B3LYP method of density functional theory (DFT). The results show that the HOMO-LUMO gap of the doped molecules become smaller, and the values of α and β of C28B2P2 are larger than those of C28B2N2. Additionally, dipole molecules in these molecules have large β, while the octapolar molecules have smaller β. The structure with excellent second-order nonlinear optical (NLO) response characteristics is selected. The calculation results of time-dependent density functional theory (TD-DFT) method show that, compared with C32, the response range of absorption spectra of all doped structures becomes wider, the maximum intensity becomes weaker, and the position of the maximum absorption wave may be red-shifted or blue-shifted. Based on the sum-over-states (SOS) method, the two-level or three-level formulas are used to explain the origin of large β in two co-doped structures, and the related electronic excitation type is π → π* excitation.
, Available online
Abstract:
In order to overcome the molding technology of high-steep spherical optical parts, a sub-aperture grinding method was proposed, taking hemispherical and hyper-hemispherical infrared optical domes as the research object. The traditional generating method grinding forming theory is expanded, the spherical surface is discretized into a series of sub-aperture rings, the grinding wheel "steps" along the ring, and a complete spherical surface is obtained by splicing and forming. In this paper, the transformation relationship between the forming spherical surface and the position coordinates of the three-axis machine tool is analyzed, the machining motion trajectory is simulated, and the radius error compensation verification experiment and the variable feed parameter optimization experiment are carried out. A method of variable radius grinding is proposed to solve the problem of over-cutting of hyper-hemisphere machining materials. The forming process test was carried out on the hot-pressed zinc sulfide and magnesium-aluminum spinel spherical dome with aspect ratios of 0.5 (hemisphere) and 0.55 (hyper-hemisphere) respectively. The sag height difference of each point on the processing surface less than 4 μm, and the surface roughness Ra<1.5 μm. The results show that the method is feasible and provides an effective solution for deep high gradient spherical processing.
In order to overcome the molding technology of high-steep spherical optical parts, a sub-aperture grinding method was proposed, taking hemispherical and hyper-hemispherical infrared optical domes as the research object. The traditional generating method grinding forming theory is expanded, the spherical surface is discretized into a series of sub-aperture rings, the grinding wheel "steps" along the ring, and a complete spherical surface is obtained by splicing and forming. In this paper, the transformation relationship between the forming spherical surface and the position coordinates of the three-axis machine tool is analyzed, the machining motion trajectory is simulated, and the radius error compensation verification experiment and the variable feed parameter optimization experiment are carried out. A method of variable radius grinding is proposed to solve the problem of over-cutting of hyper-hemisphere machining materials. The forming process test was carried out on the hot-pressed zinc sulfide and magnesium-aluminum spinel spherical dome with aspect ratios of 0.5 (hemisphere) and 0.55 (hyper-hemisphere) respectively. The sag height difference of each point on the processing surface less than 4 μm, and the surface roughness Ra<1.5 μm. The results show that the method is feasible and provides an effective solution for deep high gradient spherical processing.
, Available online
Abstract:
In order to meet the needs of digital in-situ inspection of hole drilling quality for aerospace large components, a multi-view cloud registration method based on feature locating plate is proposed, which realizes the acquisition and inspection of 3D morphology of integral hole wall. Firstly, the requirements of multi-view inspection of hole drilling are analyzed, and a method of multi-view pointcloud registration assisted by feature location plate is proposed. Then, introduce the design of feature plate and the corresponding pointcloud segmentation and recognition algorithm.Also, the method of multi-view pointcloud registration and parameter extraction based on feature self-localization was described. Finally, through experiments, titanium, aluminum and composite materials commonly used in aerospace were tested, and the average error of each is 0.011 mm, 0.034 mm and 0.041 mm respectively, which verifies the reliability of the system. The robustness of this method is demonstrated by comparing the detection results of traditional single-view method and our method.
In order to meet the needs of digital in-situ inspection of hole drilling quality for aerospace large components, a multi-view cloud registration method based on feature locating plate is proposed, which realizes the acquisition and inspection of 3D morphology of integral hole wall. Firstly, the requirements of multi-view inspection of hole drilling are analyzed, and a method of multi-view pointcloud registration assisted by feature location plate is proposed. Then, introduce the design of feature plate and the corresponding pointcloud segmentation and recognition algorithm.Also, the method of multi-view pointcloud registration and parameter extraction based on feature self-localization was described. Finally, through experiments, titanium, aluminum and composite materials commonly used in aerospace were tested, and the average error of each is 0.011 mm, 0.034 mm and 0.041 mm respectively, which verifies the reliability of the system. The robustness of this method is demonstrated by comparing the detection results of traditional single-view method and our method.
, Available online
Abstract:
Terahertz time-of-flight (THz-TOF) method is a new development direction in the field of thickness measurement, which has the advantages of fast, non-destructive and high precision. However, water vapor in atmospheric environment has strong absorption of terahertz wave, which limits the application of THz-TOF method in industrial field environment. In this paper, the propagation mechanism of terahertz wave in THz-TOF method in atmospheric environment is firstly analyzed, and then a method of water vapor elimination based on absorption model is proposed. Finally, two kinds of plastic plate samples are used as the object of application research. By comparing with the measurement results in dry environment, the results show that the method can effectively eliminate the interference of water vapor and realize the accurate measurement of material thickness and refractive index in atmospheric environment.
Terahertz time-of-flight (THz-TOF) method is a new development direction in the field of thickness measurement, which has the advantages of fast, non-destructive and high precision. However, water vapor in atmospheric environment has strong absorption of terahertz wave, which limits the application of THz-TOF method in industrial field environment. In this paper, the propagation mechanism of terahertz wave in THz-TOF method in atmospheric environment is firstly analyzed, and then a method of water vapor elimination based on absorption model is proposed. Finally, two kinds of plastic plate samples are used as the object of application research. By comparing with the measurement results in dry environment, the results show that the method can effectively eliminate the interference of water vapor and realize the accurate measurement of material thickness and refractive index in atmospheric environment.
, Available online
Abstract:
The traditional interferometry method of large optical element relies on artificially changing the beam expanding lens and optical path structure according to different test samples. This method inevitably introduces some systematic errors. Therefore, a set of corresponding dual-line control scheme is proposed according to the functional requirements of the dual-optical path interferometer and the simulation experiment. Through the cooperation of Bluetooth communication, serial communication and mechanical structure, the optical path can be converted and calibrated for many times, so that the changing position of the optical element after each switch of the measurement diameter is fixed, and the real-time status is displayed in the interactive interface based on MFC(Microsoft Foundation Classes). The results show that the measurement repeatability of PV10 and RMS can reach 0.004 λ and 0.000 4 λ under 450 mm measurement diameter, and the measurement repeatability of PV10 and RMS can reach 0.000 8 λ and 0.000 16 λ under 100 mm measurement diameter. Finally, the experimental results show that the system ensures superior measurement efficiency and repeatability, which provides reference value for the research and development of dual-optical path interferometers.
The traditional interferometry method of large optical element relies on artificially changing the beam expanding lens and optical path structure according to different test samples. This method inevitably introduces some systematic errors. Therefore, a set of corresponding dual-line control scheme is proposed according to the functional requirements of the dual-optical path interferometer and the simulation experiment. Through the cooperation of Bluetooth communication, serial communication and mechanical structure, the optical path can be converted and calibrated for many times, so that the changing position of the optical element after each switch of the measurement diameter is fixed, and the real-time status is displayed in the interactive interface based on MFC(Microsoft Foundation Classes). The results show that the measurement repeatability of PV10 and RMS can reach 0.004 λ and 0.000 4 λ under 450 mm measurement diameter, and the measurement repeatability of PV10 and RMS can reach 0.000 8 λ and 0.000 16 λ under 100 mm measurement diameter. Finally, the experimental results show that the system ensures superior measurement efficiency and repeatability, which provides reference value for the research and development of dual-optical path interferometers.
, Available online
Abstract:
An automated measurement setup of laser-induced damage threshold (LIDT) according to ISO 21254-1,2,3,4:2011 standards is presented. The main blocks of the setup, include pulse laser system, laser beam parameters measurement system, damage detection system, specimen holder and control system, are described. In order to control the LIDT measurements, software based on LabView programming package was created. Energy density of 0.1 J/cm2～100 J/cm2 is achieved. LIDT using the 1-on-1 test on 1064 nm AR coatings and Al reflective coatings on K9 substrates are respectively 27.09 J/cm2 and 3.21 J/cm2, relative measurement uncertainty are respectively 3.91% and 5.61%.
An automated measurement setup of laser-induced damage threshold (LIDT) according to ISO 21254-1,2,3,4:2011 standards is presented. The main blocks of the setup, include pulse laser system, laser beam parameters measurement system, damage detection system, specimen holder and control system, are described. In order to control the LIDT measurements, software based on LabView programming package was created. Energy density of 0.1 J/cm2～100 J/cm2 is achieved. LIDT using the 1-on-1 test on 1064 nm AR coatings and Al reflective coatings on K9 substrates are respectively 27.09 J/cm2 and 3.21 J/cm2, relative measurement uncertainty are respectively 3.91% and 5.61%.
, Available online
Abstract:
In order to measure the resonance wavelength of plasma grating, the sensitivity of grating parameters to stress is studied, a new type of stress-sensitive polydimethylsiloxane (PDMS) thin film plasma grating is proposed. Based on the principle of finite difference time domain (FDTD), a simulation model of periodic plasma grating structure is established. With the help of periodic boundary conditions, by applying stress to the grating and changing the parameters of the plasma grating (i.e. period, duty cycle and Au film thickness) to achieve the measurement of the resonance wavelength, the sensitivity of the grating parameters to the force is studied; and compare the simulation result with the theoretical value to get the relative error. The relative error is calculated by comparing the simulation result with the theoretical value. The results show that when the grating period is 0.7 μm, the duty cycle is 55%, and the gold film thickness is 0.02 μm, the response to force is most sensitive at this time; Secondly, comparing the resonance peak wavelength at different periods obtained by simulation with the theoretical calculation value, the results of the two are consistent; When the period is 0.7 μm, the wavelength of the resonance peak is 1.251 μm, and the relative error obtained by theory and simulation is less than 2%, and the result is more accurate. This method plays an important role in the fields of monochromator, spectrometer and sensor.
In order to measure the resonance wavelength of plasma grating, the sensitivity of grating parameters to stress is studied, a new type of stress-sensitive polydimethylsiloxane (PDMS) thin film plasma grating is proposed. Based on the principle of finite difference time domain (FDTD), a simulation model of periodic plasma grating structure is established. With the help of periodic boundary conditions, by applying stress to the grating and changing the parameters of the plasma grating (i.e. period, duty cycle and Au film thickness) to achieve the measurement of the resonance wavelength, the sensitivity of the grating parameters to the force is studied; and compare the simulation result with the theoretical value to get the relative error. The relative error is calculated by comparing the simulation result with the theoretical value. The results show that when the grating period is 0.7 μm, the duty cycle is 55%, and the gold film thickness is 0.02 μm, the response to force is most sensitive at this time; Secondly, comparing the resonance peak wavelength at different periods obtained by simulation with the theoretical calculation value, the results of the two are consistent; When the period is 0.7 μm, the wavelength of the resonance peak is 1.251 μm, and the relative error obtained by theory and simulation is less than 2%, and the result is more accurate. This method plays an important role in the fields of monochromator, spectrometer and sensor.
, Available online
Abstract:
A general model of fiber Bragg grating (FBG) acceleration detector has been established for the two-point package model. The sensitivity and the resonance frequency's analytical expressions of the acceleration detector have been deduced theoretically, and the influence factors which affecting the sensitivity and resonance frequency have been researched deeply, the acceleration detector’s response characteristics have been discussed, which would be affected by the ratio, that the package fiber stiffness to the structure stiffness. Based on this basis, the restrictive relation between the sensitivity and the resonance frequency have been analyzed, among the equivalent mass and the ratio have been in the range of 0~100 g, 0~1 and 0~100, respectively. And the resonance frequency’s change rule within the scope of 0~500 Hz (low-medium frequency) and 0~1200 Hz (medium-high frequency) also has been studied. Furthermore, the package fibers have been chosen 10 mm and 60 mm as examples to analyze the influence to the above two mentioned, the sensitivity of each simulation would reach up to ~1 000 pm/G, and the quality factor has been introduced. This study has played a significant role in the design and comprehensive performance evaluation of the acceleration detector, and provided a theoretical reference for the optimization of structural parameters to it.
A general model of fiber Bragg grating (FBG) acceleration detector has been established for the two-point package model. The sensitivity and the resonance frequency's analytical expressions of the acceleration detector have been deduced theoretically, and the influence factors which affecting the sensitivity and resonance frequency have been researched deeply, the acceleration detector’s response characteristics have been discussed, which would be affected by the ratio, that the package fiber stiffness to the structure stiffness. Based on this basis, the restrictive relation between the sensitivity and the resonance frequency have been analyzed, among the equivalent mass and the ratio have been in the range of 0~100 g, 0~1 and 0~100, respectively. And the resonance frequency’s change rule within the scope of 0~500 Hz (low-medium frequency) and 0~1200 Hz (medium-high frequency) also has been studied. Furthermore, the package fibers have been chosen 10 mm and 60 mm as examples to analyze the influence to the above two mentioned, the sensitivity of each simulation would reach up to ~1 000 pm/G, and the quality factor has been introduced. This study has played a significant role in the design and comprehensive performance evaluation of the acceleration detector, and provided a theoretical reference for the optimization of structural parameters to it.
, Available online
Abstract:
In order to solve the problem of image clarity and contrast degradation in fog scene image restoration, a single image defogging algorithm based on residual learning and guided filtering was proposed. The residual network was constructed by using foggy images and corresponding clear images. Multi-scale convolution is used to extract more detailed haze features. Taking advantage of the anisotropy of the guided filter, the image after the residual network is filtered to maintain the image edge characteristics, and a clearer fog-free image is obtained. The experimental results show that, compared with DCP algorithm, CAP algorithm, SRCNN algorithm, DehazeNet algorithm and MSCNN algorithm, On synthetic foggy images, the PSNR reaches 27.840 3/dB at the highest, the SSIM value reaches 0.979 6 at the highest, and the running time on natural foggy images reaches 0.4 s at the lowest. and the subjective evaluation and objective evaluation are better than other comparison algorithms. Proposed to fog algorithm not only to the fog effect is better, and faster, with strong practical value.
In order to solve the problem of image clarity and contrast degradation in fog scene image restoration, a single image defogging algorithm based on residual learning and guided filtering was proposed. The residual network was constructed by using foggy images and corresponding clear images. Multi-scale convolution is used to extract more detailed haze features. Taking advantage of the anisotropy of the guided filter, the image after the residual network is filtered to maintain the image edge characteristics, and a clearer fog-free image is obtained. The experimental results show that, compared with DCP algorithm, CAP algorithm, SRCNN algorithm, DehazeNet algorithm and MSCNN algorithm, On synthetic foggy images, the PSNR reaches 27.840 3/dB at the highest, the SSIM value reaches 0.979 6 at the highest, and the running time on natural foggy images reaches 0.4 s at the lowest. and the subjective evaluation and objective evaluation are better than other comparison algorithms. Proposed to fog algorithm not only to the fog effect is better, and faster, with strong practical value.
New website launched!
陕公网安备 61011302001501号