Power line detection based on an improved RCF and the UAV images
In order to solve the problems such as edge blur when richer convolutional features(RCF)algorithm detects power lines,feature maps contain too much noise,and multi-scale information is lost when fusing feature maps,RCF algorithm was improved in this paper.Firstly,the down-sampling tech-nique with translation invariance was used to enhance model robustness.Secondly,convolutional block attention module(CBAM)mechanism was introduced into the convolutional block attention of RCF trunk network to enhance the power lines characteristics.Thirdly,the cascade network is added into the side output network of RCF,and the feature map is fused with the multi-scale feature fusion module using the channel attention mechanism,to obtain better details.The results showed that the optimal dataset scale,the optimal image scale and average precision of the improved RCF increased 0.7％,1.3％and 1.7％,respectively.The detection results of the improved model are less noise,and the power line is more clear and accurate.
power linesedge detectionricher convolutional features(RCF)unmanned aerial vehicleattention mechanismmulti-scale fusion
万方数据
维普
