Improved Aircraft Flared Tube Defect Detection Algorithm of YOLOv4 Network Structure

Abstract When detecting aircraft flared tube defect with the YOLOv4 network structure, tiny object defects will be missed, and resulting in a high missed detection rate low mAP (mean Average Precision) value. This paper proposes an improved algorithm of structure. Firstly, order to improve feature extraction capability for defects, convolution operation is added SPP (Spatial Pyramid Pooling) PANet (Path Aggregation Network) Secondly, representation pyramid enhanced utilizing PANet. Thirdly, decoupled head utilized model performance. Finally, we construct dataset by labeling samples, experiment network. Experimental results show that value task increased from 91.26% 95.31%, average time 346.17ms 278.61ms.