The Defect Detection Method for Automotive Parts Based on Improved Single Shot MultiBox Detector Using Feature Enhancement

The application of machine vision systems holds significant potential for detecting the processing quality of automotive parts in industrial intelligent manufacturing. To overcome the limitations of traditional Single Shot Detector (SSD) algorithms in identifying defects in automotive parts, we propose an improved SSD method. This enhanced approach integrates a Convolutional Block Attention Module (CBAM) into the feature extraction network. Within the CBAM, the spatial attention module increases the model’s sensitivity to features of the machined part of the workpiece, while the channel attention module enhances the model’s generalization in detection tasks. By up-sampling and fusing shallow and deep feature layers, we construct multi-scale feature information, significantly improving the detection capability for small targets such as holes and welded joints on workpieces. Additionally, we introduce Focal Loss to address sample category imbalance and reduce the model’s false alarm rate. Compared to the traditional SSD algorithm, our experimental results demonstrate that the mean average precision (mAP) of our model in workpiece defect detection is improved by approximately 10.3%, and the detection accuracy is increased by 16.1%. This advancement offers a robust solution for quality detection in the automotive parts processing industry.