Mining Image Anomaly Detection Algorithm Based on Faster-RCNN and Self-attention Mechanism
Mine abnormal behavior detection is an important task,which can help for improving the supervision efficiency for mine safety production.A Faster-RCNN algorithm based on self-attention mechanism is proposed to detect abnormal behav-ior in mine images.The algorithm can extract key features effectively and reduce the interference of redundant information by weighting feature maps with self-attention mechanism.Firstly,the RoI region is extracted from the image,and the feature map in the RoI region is weighted by the self-attention mechanism,so that the key features can be extracted more accurately.At the same time,the feature maps in the RoI region are fused to capture more image details.Finally,a classifier is used to classify each RoI region to detect abnormal behavior in the image.The algorithm is trained and evaluated on the self-built data set,and the test results show that the proposed algorithm has higher accuracy and robustness in the abnormal behavior detection task of miners,and the detection accuracy of the proposed algorithm is improved by 4.8%compared with the traditional Faster-RCNN algorithm.In addition,the algorithm has better robustness to changes in illumination and angle,and can effectively deal with complex environments in real scenes.
NETL
NSTL
万方数据
