Defect Detection Method of Interface Surface for Automobile Brake Pipe via Double Layer Cascade and Multi-feature Scale Fusion

To address the challenges of uneven distribution, small size, and diverse morphology of interface surface defects on automotive brake pipes, a novel defect detection algorithm utilizing a dual-level cascade and multi-feature scale fusion Faster Region-based Convolutional Neural Network (Faster-RCNN) is proposed. This approach incorporates region of interest (ROI) alignment pooling and a balanced loss function to enhance detection accuracy. Experimental results demonstrate that the enhanced algorithm achieves high precision and robustness in detecting defects on automotive brake pipes.