Abstract: Precision silicon wafer cutting is a critical process in industrial fields such as solar cell manufacturing, where the accurate inspection of diamond wire grooves and the overall wire mesh layout is paramount for ensuring wafer quality and cutting precision. However, traditional manual inspection methods suffer from low efficiency and poor consistency. Furthermore, pure deep learning-based approaches often suffer from insufficient localization and segmentation accuracy due to feature confusion in such high-density, low-contrast periodic structures. To address these challenges, a hybrid inspection method was proposed. First, a wire groove position encoding model was constructed through frequency domain analysis and curve fitting, achieving micron-level segmentation and localization of the grooves. Next, a lightweight network DARMobileV2 was designed for rapid classification of single wire groove images. Finally, the complete wire mesh layout was reconstructed by integrating the position encoding information. Experimental results under typical detection conditions showed that the proposed method achieved a detection accuracy of no less than 99.6% on edge computing devices, with a detection time under 1 s. Compared with the traditional manual inspection method, the false detection rate was reduced 97.74%, and the speed was increased by 52.4 times. This method provides an efficient and reliable automated inspection solution for precision silicon wafer cutting process.