2021 Vol. 42, No. 4
In order to improve the disadvantages that the detection range is close and the probe beam is difficult to be used effectively in the traditional ptychography system, the terahertz ptychography experimental system based on focused Gaussian beam was proposed. Based on the principle of ptychography and using the angular spectrum transmission theory, a focused Gaussian beam was used to replace the diffracted beam generated by the traditional aperture. The feasibility of this replacement was verified through computer simulation experiments, and the corresponding imaging experiments were completed. The numerical simulation results show that it is feasible to carry out the terahertz ptychography using a focused Gaussian beam in the 0.1 THz band, and the simulation and experimental results verify the feasibility of the system. Under the condition of detection distance of 50 mm, the theoretical depth resolution reaches to λ/4, and the experimental system depth resolution is λ/2, which fully demonstrates the effectiveness of this ptychography system.
A measurement system based on binocular vision and industrial robots was proposed to meet in-situ high-efficiency and high-precision measurement of large-scale high reflective components. Through target detection, the system could accurately segment the region of interest where the marking points were located, effectively reduce the false extraction caused by high reflective surface, and improve the robustness and measurement efficiency of binocular vision measurement system. At the same time, through controlling the movement of the robot end effectors, the measurement of the whole component in multi-position was completed, and then the measurement results of different positions were unified in the same coordinate system through the coordinate transformation relationship in multi-position. The experimental results show that in the range of 1.2 m × 1 m, the root-mean-square (RMS) of vision measurement precision can reach to 0.049 mm in nine positions. The whole measurement system can effectively complete the high-efficiency and high-precision measurement of the simulated cabin components.
The external cavity laser (ECL), which is mainly used for coherent light communication, is characterized by narrow line width and low power consumption. In practical application, the output optical power of ECL will decrease when the ambient temperature is the high or low temperature. Through the use of finite element commercial software Ansys, it was found that the mechanical structure had the slight deflection change at high or low temperatures. Compared with the room temperature, the maximum deflection angle at high or low temperatures was 0.04°. The cavity length of ECL was long, combined with the optical path simulation software Zemax, the slight deflection change would cause the big change of optical path, affect the coupling efficiency and then lead to the reduction of output optical power. The change of optical path was confirmed by the experimental measurement, and based on the theory of power reduction caused by the deflection change, the methods of shortening the cavity length and thickening the base were proposed to reduce the variation of output optical power with the ambient temperature from the point of design. The proposed model and ideas can provide quantitative analysis ideas for the correlation analysis of power and temperature of similar products.
In the optical imaging system, the dynamic modulation transfer function (DMTF) is affected by the wave aberration of optical system, the signal resolution and transfer characteristics of detector modules, and the image blur caused by the carrier motion, which has become one of the important parameters of the photoelectric imaging system. Based on the testing principle of dynamic and static modulation transfer function of photoelectric imaging system, a DMTF measuring device was developed, in which the focal length of optical collimation system was 10 000 mm, and the speed control range of moving target was 30 mm/s~5 000 mm/s, which met the DMTF test of photoelectric imaging system with long focal length. A series of experiments were carried out, and the results show that the repeatability of DMTF measurement is better than 0.01, and the uncertainty of measurement can reach to U= 0.05 (k=2).
Aiming at the problem of carrier vibration, system parameter perturbation, friction torque and external disturbance in photoelectric tracking and stabilization platform system that directly affect the stability accuracy of photoelectric tracking system, the deviation in the tracking system was suppressed from the servo controller control perspective to improve the stability accuracy and ensure its strong robustness in the complex environment. By establishing the mathematical model of the photoelectric tracking and stabilization platform system, the influence of disturbance torque, gyro noise and system parameter changes on the photoelectric tracking and stabilization platform was analyzed, and the compound control strategy based on the novel nonlinear extended state observer (NNESO) and sliding mode variable structure control (SMVSC) was designed. The results of theoretical derivation and simulation experiments show that this kind of compound control strategy has full adaptability and strong robustness to disturbance torque, external disturbance and system parameter perturbation. At the same time, compared with the control system without the novel nonlinear extended state observer, it proves that the compound control strategy has higher stability accuracy and faster dynamic response, which has good effect in compensating the nonlinearity and can enhance the anti-disturbance ability of the photoelectric tracking and stabilization platform.
The catadioptric panoramic optical imaging system has the imaging characteristics of the plane-symmetric optical system. Thus, the application of focal power governing equations and sixth-order wave aberration theory of ultra-large field of view optical system were proposed to preliminarily determine the initial parameters of system structure. Based on this structure, the Zemax software was applied to design and optimize, and a catadioptric panoramic optical imaging system consisting of one even aspherical mirror and six refractors was obtained. This designed system worked in the visible wave band. The angle range of the whole field of view is 12°~178°, and the F number is 3.5, thus its aberration is well corrected. The value of modulation transfer function is greater than 0.37 at the spatial frequency of 40 lp/mm, which has better imaging performance and meets the practical application requirements. At the same time, the tolerance analysis of the designed system is carried out.
In the image processing of machine vision, there are many factors affecting the image quality, and one of the main factors is the illumination angle of the light source. In order to get the better image data, it is necessary to carry out the simulation analysis and experimental study on the optimal angle of the light source illumination in the vision system. A single-variable method was used in the experiment to change the illumination angle of the light source only. Through a combination of the theoretical calculation, the simulation, and the experimental verification, the best illumination angle of the light source was obtained. The image edge quality under the optimal angle and other angles was examined in combination with the Sobel algorithm. The results show that the performances of image illumination data and edge image are the best under the illumination angle of 57.17°, which are of guiding significance for deeper image processing and image acquisition.
For the purpose of shaping the low frequency terahertz Gaussian beam into the large-size flat-topped beam, a method of designing a large-size flat-topped beam shaping system using dual-phase plate structure was proposed. Based on the Input-Output algorithm, the system parameters were calculated using angular spectrum theory, and the designed system was numerically simulated and the corresponding experimental system was built to verify whether the lateral intensity distribution of the emitted light field met expectations. The results show that the designed dual-phase plate shaping system can successfully convert the 0.1 THz Gaussian beam into a 100 mm-size flat-topped beam. The simulation results are close to the theory, which proves the rationality of the designed system. In the experiment, the beam diameter reaches 228 mm. Although there is some energy leakage, the flatness is very good, and the relative root mean square is even better than the theoretical and simulation results, reaching more than 89%, which fully demonstrates the actual effectiveness of this design method.
Aiming at the problem that diffraction effect of digital micro-mirror device (DMD) has serious influence on system imaging, the stray light generated by DMD device was studied. Firstly, based on the rigorous coupled wave theory, the DMD diffraction model was constructed and the effects of different bands and incident angles on its diffraction efficiency were analyzed. It was concluded that the diffraction effect of DMD was more obvious in the long-wave infrared band, and the diffraction effect of DMD was more significant with the increase of incident angle. Secondly, the generation methods of diffracted stray light of DMD in the projection system was analyzed. By changing the number of F in the system, the incident angle was controlled, and then the energy proportion of stray light generated by the diffraction effect of DMD was calculated. Finally, the curve of the optical system with different number of F at different wavelengths along with the diffraction effect was obtained. The results show that the larger the F number, the smaller the influence of DMD diffraction effect on stray light.
With the shortages of the large amounts of calculations in speeded-up robust features (SURF) algorithm, and low efficiency of image stitching, the features from accelerated segment test (FAST) corner points were used to instead of the SURF spots in order to extract the feature points in image overlapping area. The SURF descriptor was used to describe the feature points, by using the descriptor dimensionality reduction method, the adaptive nearest neighbor and nearer neighbor ratio method, and the geometric constraint method, the false matching points were eliminated in order to improve the matching accuracy. After matching, the random sample consensus (RANSAC) algorithm was improved by reducing the number of sample set and rejecting the unreasonable parameter models to obtain the homography matrix. Finally, the image transformation, fusion and stitching were carried out. The experimental results show that the total time of image stitching is reduced by 12% compared with the traditional SURF algorithm, and the stitching efficiency is improved significantly.
The infrared dim-small target detection is a key technology in the fields of security surveillance, reconnaissance detection and precision guidance. To improve the accuracy and real-time performance of infrared dim-small target detection under complex background conditions, an infrared dim-small target detection algorithm YOLO-FCSP based on deep learning was proposed. Considering the characteristics of dim-small targets in infrared images, the feature extraction network was designed based on the YOLO detection framework by reducing the number of downsampling and combining cross-stage local modules, Focus and spatial pyramidal pooling structure. The feature fusion network was improved by utilizing the idea of multi-path aggregation. The number of detection output layers was adjusted to enhance the reuse of feature information. The experimental results show that the proposed method has higher accuracy and detection speed in detecting infrared dim-small targets, achieving 91.9% precision and 94.6% recall, the average precision (AP) value to 92.6%, and the detection speed as 170 fps, which meet the requirements of real-time detection in practical applications.
In the process of Fourier ptychographic microscopy, the collected low resolution images will directly impact on the quality of reconstruction images. The existing studies put forward with the image super-resolution reconstruction technology and traditional denoising processing method for low resolution images to solve this problem. However, the super-resolution reconstruction method requires to collect a large number of original images, which can increase the time loss, while the traditional denoising algorithm can cause the loss of original information and seriously affect the quality of reconstruction images. So, the convex optimization algorithm was proposed. The recovery of the noise image could be realized by solving a convex optimization model, and the iterative shrinkage threshold algorithm was used to solve this model. The Barzilai-Borwein (BB) rules were adopted to initialize the line search step length at each iteration, accelerate the convergence speed, and select the soft threshold function to reduce the loss of original information during the image denoising. The PSNR of the final reconstruction image is 27.634 6 dB, the SSIM is 0.926 1, and the required processing time is 5.850 s. Therefore, the Fourier ptychographic microscopy technology based on convex optimization has the advantage of improving the reconstruction image quality without too much time loss.
Aiming at the problem of low illumination degraded image enhancement obtained by remote imaging system, an image enhancement algorithm based on fusion Retinex and discrete wavelet singular value decomposition was proposed. In this method, the adaptive full-scale Retinex (AFSR) was used to coarsely extract the illumination and reflection components, and then the reflection components of the extracted image were decomposed into four frequency subbands by discrete wavelet transform, and the singular value matrix of the low-frequency subbands image was estimated. Finally, the inverse wavelet transform was adopted to precisely reconstruct the image. The experimental results show that the visual enhancement effect of the low illumination degraded image processed by the proposed method is better, which is superior to other classical algorithms in terms of objective evaluation indexes such as image contrast, information entropy, average gradient and edge density.
The three-dimensional face measurement and segmentation have extensive application requirements and are the important research direction at present. However, the rapid development is constrained as the three-dimensional face data is huge and unordered. Firstly, the three-dimensional face measurement system based on the structured light method was developed, and the three-dimensional face data with high precision saved in the form of point cloud was obtained. Secondly, after the conformal transformation was adopted to preprocess the three-dimensional face data and the two-dimensional convolutional neural network segmentation combined with the three-dimensional inverse mapping was adopted, the three-dimensional face segmentation was realized, the disorder and rotation of three-dimensional data were solved, and the time consumption of three-dimensional data segmentation was reduced. The experimental results show that the proposed precision of three-dimensional face measurement system can reach to 0.5 mm, and the average intersection-over-union (IoU) of three-dimensional segmentation can reach to 0.78. The overall efficiency of two-dimensional segmentation combined with the three-dimensional inverse mapping is obviously higher than that of the three-dimensional segmentation.
Aiming at the problem of under-segmentation or over-segmentation in the fault area caused by the shortcomings of more time-consuming, low accuracy of segmentation and mis-segmentation in the multi-threshold segmentation of infrared images in electrical equipment fault diagnosis and location based on the two-dimensional (2D) OTSU image segmentation algorithm, an improved fusion algorithm of glowworm swarm optimization (GSO) and 2D OTSU was proposed to improve the real-time and accuracy of multi-threshold segmentation of infrared images for electrical equipment. In the optimization process, the local optimization was extended to the global optimization, and the nonlinear degressive step size and the new shifting strategy were introduced to optimize and improve the GSO. The experimental results show that the proposed fusion algorithm is more accurate than 2D OTSU and unimproved GSO with 2D OTSU fusion algorithm to segment the image abnormal area of operational electrical equipment in segmentation results, and the segmentation speed can be improved by 19 times and 1.28 times, which lays a foundation for the effective identification and location of early fault in infrared images.
The phase discontinuity caused by the geometric mutation has always been a challenging problem in phase unwrapping. To solve this problem, a fast discontinuous phase unwrapping algorithm based on orientation and transformation was proposed. In the proposed algorithm, the structure tensor of the image was used to estimate the orientation of the wrapped phase diagram. The reliable weight coefficient diagram obtained from transformation and difference calculation of the orientation diagram was taken as the weight of the weighted least square method, and used the preprocessed conjugate gradient method to solve iteratively. This algorithm could quickly find the discontinuous regions, and performed a separate phase unwrapping in the discontinuous regions, which had good recognition and unwrapping effect. The principle and implementation steps of the algorithm were described in detail, and the algorithm was verified by simulation and experimental data. The experimental results show that the root mean square error of phase unwrapping is 0.36, which proves that this algorithm can quickly and accurately unwrap the discontinuous wrapped phase.
The image deblurring technology of dynamic scene is a challenging computer vision problem. The blurry images not only affect the subjective perception but also affect the performance of the subsequent intelligent analysis. An image deblurring method of dynamic scene based on attention residual CODEC network was proposed. Firstly, in the coding stage, many residual modules were used to extract the features, and the spatial attention module was added to perceive the blurry spatial position information. Then, a global-local residual connection strategy in the internet was adopted to fuse the multi-layer convolution features to reduce the information loss. Finally, a restored image with clear edges and structure was generated in the decoding stage. The experimental results show that the peak signal-to-noise ratio (PSNR) obtained on the public data set is 31.76 dB, and the structural similarity index measure (SSIM) value is 0.912. Both the objective and subjective quality evaluations indicate that the proposed method can effectively recover the clear images containing abundant edge contour information, which obtains the optimal performance in the compared algorithm.
The super-resolution reconstruction models represented by the super-resolution convolutional neural network (SRCNN) models usually have high peak signal to noise ratio (PSNR) and structural similarity index measure (SSIM) values, but its visual perception is not satisfactory. And the generative adversarial networks (GAN) models represented by the super-resolution generative adversarial networks (SRGAN) having high perceptual quality is prone to produce a lot of false details, which is manifested in its low PSNR and SSIM values. To solve the above problems, a perceptually enhanced super-resolution reconstruction model based on deep back projection was proposed. The dual-scale self-adaptive weighted fusion feature extraction module was adopted by this model for feature extraction, then the sampling was carried out by the deep back projection, and finally the final output was obtained after the enhanced module was enhanced. The residual connections and dense connections were adopted by the model, which facilitated the features sharing and the effective training of the model. In the index evaluation, the learned perceptual image patch similarity (LPIPS) metric based on the learning was introduced as a new quality evaluation index of image perception, together with PSNR and SSIM as the model evaluation index. The experimental results show that the average values of PSNR, SSIM, and LPIPS of the model on the test data set are 27.84, 0.7320, and 0.1258, respectively, and all the indicators are better than the comparison algorithm.
The axial spacing measurement between optical elements is of great significance to the positioning and adjustment of precision optical systems. Aiming at the problems of complex dispersive objective lens structure and small dispersion range in the existing dispersion confocal measurement system, a method for measuring the axial spacing of optical elements based on the method of polarization diffraction dispersion confocal was proposed. The axial dimension of a traditional refractive confocal lens with a few hundred millimeters and the complex alignment requirements were simplified to a single-chip lens with a few millimeters, which simplified the system structure. The results of lens spacing and thickness measurement experiments show that the spacing measurement error is 10 μm.
In order to accurately locate the focused spot of the laser direct writing system on the surface of the element to be machined in real time, a synchronous phase-shifting microscopic interference focal detection method based on wave aberration criteria was proposed, which realized the real-time detection and adjustment of the defocusing amount. The focal detection optical path was introduced in the direct writing system, and shared the same objective lens with the direct writing optical path. The Linnik synchronous phase-shifting microscopic interference focal detection system was constructed to extract the wave surface phase information containing the defocusing amount. Then the magnitude and direction of defocusing amount was analyzed from the wave aberration data of objective lens with large numerical aperture (NA). The simulation results verify the correctness of defocusing calculation method based on wave aberration criteria, when NA≥0.5, the defocusing detection sensitivity can reach to 4 nm. The experimental results verify the feasibility of synchronous phase-shifting microscopic interference focal detection method, and the precision can reach within 10 nm, which can meet the high-precision measurement requirements of laser direct writing.
The InGaAs photodetector is one of the important optical detection devices in the near-infrared band. It has the characteristics of high quantum efficiency, low dark current and wide bandwidth. It is widely used in photoelectric measurement, optical communication, remote sensing and other fields. Based on the spectral comparison device of super-continuum white laser and double monochromator, a standard InGaAs trap detector was used to calibrate the relative spectral responsivity of a planar InGaAs detector in the 900 nm~1600 nm band. The responsivity calibration results of the two light sources were compared. The maximum relative difference of the spectral responsivity in the 900 nm~1600 nm band under the two light source conditions was less than 0.2%, which achieved good consistency. The measurement repeatability of the super-continuum light source is less than 0.06%. It was much smaller than the measurement repeatability of 1% using the halogen lamp, which reduced the uncertainty component contributed by the measurement repeatability. The feasibility of the super-continuum laser in the calibration of the relative spectral responsivity of the detector was verified. In addition, the measurement uncertainty of the spectral responsivity results of the calibrated planar detector was analyzed.
The compound-eye optical imaging system features outstanding advantages over traditional single-aperture optical system in large field of view detection, image recognition, and target detection. However, with the increase of the field of view, the image distortion of sub-aperture itself and the distortion caused by installation position error of the multiple sub-aperture will directly affect the quality of mosaic images. To address the problem, the distortion measurement and correction of compound-eye system were carried out by using the photoelectric measurement technology, and the multi-mode dynamic electronic distortion measurement target was generated. Then, the distortion measurement and correction model was constructed, and the polynomial fitting algorithm was established. The least square method was adopted to obtain the distortion coefficient, and the images were reconstructed by using the bilinear interpolation model. The experimental results show that the average relative distortion after correction is better than 0.1%, which can meet the accuracy requirements of distortion correction and image mosaic of compound-eye optical imaging system with large field of view.
Aiming at the difficult problem of obtaining the balance towing position of the refueling hose under flight conditions, a set of image-based method for measuring balance towing position of aerial refueling hose was designed. By installing an airborne multichannel and high-definition image measurement system on the tanker to obtain the moving images of the refueling hose in the air refueling process, and researching on the key technologies such as the synchronous acquisition of the multi-camera, the camera calibration with a large depth of field, and the recognition and tracking technology of the refueling taper sleeve without reference point, the accurate measurement of the balance towing position of refueling hose was realized. The simulated experimental results show that the measurement accuracy of this scheme is better than 5 cm, which meets the accuracy requirements of the flight test.
During the working process of high-power LED array dynamic light source, the photoelectric and thermal parameters have the characteristics of uncertainty and time-varying delay nonlinearity. The dynamic kernel principal component analysis method (DKPCA) was used to conduct the on-line state observation and fault diagnosis of the high-power LED array dynamic light source, which could effectively capture the nonlinearity and correlation characteristics of the observation data, realize the fault detection based on the statistical threshold value calculated by the principal component characteristics of the historical data and the statistical characteristics of the online data, and realize the separation of the fault by the reconstruction contribution graph method. The simulation experiments show that the effective monitoring and diagnosis of typical sensor and actuator faults of high-power LED array dynamic light source are more sensitive to faults than kernel principal component analysis method. The fault detection rate is increased by 7.5%, and the false detection rate is reduced by 4.2%.
The optical fiber image inverters are the key components which applied in low-light level night vision device, and the optical fiber image inverters with large aperture can achieve the detection of wide field of view, large view and further visual range. However, the existing twisting operation process cannot produce the qualified products. The twisting operation process with rotation differential velocity combined with double internal furnace heating design was adopted, and finally the large-aperture optical fiber image inverters with required performance indexes were prepared. The test results show that the angle concentration amount in twisting region is decreased to 5.98 °/mm, which significantly improves the edge resolution of the large-aperture optical fiber image inverters.
A flat super-continuum spectrum light source of all-fiber structure based on noise-like pulse pumping was designed. In the dispersion-managed erbium-doped fiber laser, by adjusting the polarization state in the cavity, when the pump power was 450 mW, the stable noise-like pulse mode locking was achieved, the center wavelength was 1600 nm, and the pulse width was 303 fs. When the maximum pump power was 1 W, the direct output power of the resonator was 8.6 mW, and the lower power could not effectively expand the width of the super-continuum spectrum. Therefore, an erbium-doped fiber amplifier was designed for power amplification, and the maximum output power was 338 mW. The power-amplified noise-like pulse was coupled into the highly nonlinear fiber to obtain a super-continuum spectrum. The 20 dB spectral range of super-continuum spectrum is 1 530 nm ~ 2 300 nm, and in the range of 1 736 nm ~ 2 134 nm, the flatness of the spectrum is less than 0.5 dB.
In an optical fiber phased array, the array layout of the outgoing beam has a crucial influence on the coherent combination performance of high-power lasers. Aiming at the problem of low energy synthesis efficiency caused by the difficulty of breaking through half the wavelength of the existing optical fiber phased array element spacing, a retina-like multi-ring optical fiber phase-control method was proposed, and a circularly symmetrical distribution structure was designed. Keeping the positions of the central array element and the outermost ring array element unchanged, a particle swarm algorithm was used to optimize the spacing of the remaining elements between the rings to obtain the best far-field coherent synthesis performance. The simulation results show that, compared with the traditional optical fiber phase-control method, the energy concentration of the proposed array layout is increased from 0.562 to 0.921, and the peak sidelobe level is compressed from 0.212 to 0.043.
An all polarization-maintaining figure-nine cavity fiber laser, which could be self-starting in low threshold state was reported. A phase shifter was used in the resonator structure to reduce the mode-locking threshold. When the pump power reached 120 mW, a self-starting conventional soliton mode-locking could be realized, with the central wavelength of 1530 nm and the pulse width of 614.6 fs. Then the pump power gradually increased to 470 mW, which realized the conversion from soliton pulse to noise-like pulse. In this mode-locking state, the laser output power is 63.2 mW, and the corresponding noise-like pulse energy is 5.69 nJ. The constructed laser has the advantages of low mode-locking threshold, self-starting, and can realize the conversion between ultrafast pulses and high-energy pulses only by adjusting the pump power, which has extensive application values.