Detection of Solid Material Leakage Based on Deep Convolutional Neural Network
Xiangyu Zhao, Xueyi Zhang, Zipeng Zhao, Ruirui Zhao · 2024
There are many potential dangers associated with the leakage of solid materials. However, there is currently little research conducted on the scenario of solid material leakage. To solve this problem, an automatic leakage detection method based on deep convolutional neural network is proposed in this paper. Firstly, the images in different scenes of industrial sites are taken as the research object, and a solid material leakage detection dataset called SM-Leakage is constructed. Then, the detection model is trained, tested and validated using the constructed dataset. The results show that the model proposed in this paper can effectively realize the automatic detection of solid material leakage. The accuracy of the model on the test set can reach 96.80%. In addition, this paper also visualizes the basis of image classification and finds that it is consistent with the basis of manual detection.