Abstract:

In order to avoid the problem of serious production accidents caused by the conveyor belt running off the mine road during the sorting operation in the coal processing plant, a video detection algorithm fusing Gaussian edge projection based on semantic segmentation was proposed, which monitored the conveyor belt deviation in real time. First, the algorithm was preprocessed by Gaussian filtering to reduce the influence of video image noise on the neural network. Then, a radial basis neural network structure with fused Gaussian projection was constructed to enhance the edge information of the image, and the Gaussian projection was fused into the network to realize semantic segmentation, which ensured the recognition and detection effect while reducing the computational consumption of the network as much as possible. The results show that the algorithm can accurately recognize and detect the conveyor belt and then realize the conveyor belt deviation warning through the experimental scene verification and analysis of the working video of the mine road in the coal processing plant. Meanwhile, the algorithm consumes fewer computational resources and has higher computational efficiency than the typical deep learning model of semantic segmentation while realizing the effective function.

Key words: semantic segmentation, conveyor belt deviation, Gaussian filter, video detection, deep learning