Detail extraction of mechanical component image based on depth learning

Abstract Aiming at the problem of feature extraction, recognition and classification machining features various mechanical part models, an image detail extraction method based on deep learning architecture is proposed. We use median filtering to filter impulse noise in initially. After horizontal tilt correction, contour divided into several fan sub regions. Then, structural parameters YOLOV3 network are optimized make it more suitable for small target detection industrial parts basis fully retaining ability depth convolution network. The simulation results show that original YOLOv3 algorithm, improved model has good robustness, problems different scales, low utilization occlusion which obviously improved. Compared with traditional methods images, strategy also shows greater advantages.