面向钢丝绳微弱损伤智能识别的多尺度注意力网络

Aiming at the problems of weak surface damage, high identification difficulty and small number of damage samples of passenger ropeway under complex and severe working conditions, an intelligent wire rope weak damage identification method based on multi-scale feature extraction and attention mechanism is proposed. Firstly, the generative adversarial network is introduced to remove the ambiguity of the images under running condition, the perspective transformation and random clipping are used to expand the number of samples. Then the multi-scale convolution neural network is used to extract the overall features and local features of the damage images, and the small-scale feature map samples are collected and then spliced with the large-scale feature map to join the features of different scales. On this basis, the key features are enhanced by the attention mechanism. Finally, the coordinates and category of the damage are output by the prediction module. The proposed method is verified by collecting the damage images on the wire rope damage simulation test bench. The results show that compared with the existing methods, the proposed method greatly shortens the training time while the mean average precision (mAP) is not reduced, which reflects the validity of the proposed method.