To ensure assembly quality， traditional methods rely on manual visual inspection， which often suffers from limited stability and accuracy. Addressing the limitations of existing assembly inspection techniques， a vision-based compliance inspection method for wing cable bracket assembly is proposed. To mitigate missing image information caused by strong reflections from metal brackets， polarization imaging technology is employed to enhance the imaging quality. To resolve the challenge of mixed foreground and background information in the assembly scene， a two-stage instance segmentation algorithm is developed by integrating YOLOv8 with SAM， enabling rapid deployment and precise part segmentation. To compensate for viewpoint discrepancies resulting from repeated positioning errors between the actual standard and the assembly drawing， invariant features are extracted for image alignment. Finally， pixel-level mask intersection over union （IoU） metrics between corresponding parts are calculated to identify incorrect or missing assemblies. Compared to manual inspection， the proposed method substantially improves inspection efficiency. Experimental results validate its effectiveness， demonstrating an accuracy of 96.08% and a recall rate of 100% for misassembled parts， thereby confirming its high deployment efficiency， accuracy， and robustness.