基于Res2-UNet模型的皮带煤量检测

Belt Coal Quantity Detection Based on Res2-UNet Model

为了能够检测煤矿井下的煤量,预测和提高煤的利用率,同时节省电能,减少人力的监管和资源成本。利用煤矿安装的视频监控系统,采用非接触的方式通过Camshift算法对快速运动皮带上的煤量捕捉和跟踪,然后建立Res2-UNet模型来获得显著性信息,融合灰度、纹理、边缘等特征到单一的网络中,实现了皮带煤量的检测。模型利用U-Net网络的思想以编码器-解码器为架构,编码器以Res2Net网络为骨干网络提取煤流不同层次特征的信息,解码器通过反卷积上采样操作恢复图像尺寸。经过构建皮带数据集训练和测试模型,实验结果表明,提出的方法能够快速准确地检测出皮带上的煤料,精确率达到95.5%,每张图像从输入网络到输出的运行时间为4.8s。表明该方法具有一定的实用性和有效性。

In order to be able to detect the amount of coal in the mine,predict and improve the utilization rate of coal,while saving electricity,human supervision and resource costs are reduced.By using the video surveillance system installed in the coal mine,the coal amount on the fast moving belt is captured and tracked by Camshift algorithm in a non-contact way,and then the Res2-UNet model is established to obtain the significance information,and the gray level,texture,edge and other features are inte-grated into a single network to realize the coal amount detection of the belt.The model uses the idea of U-Net network and encod-er-decoder architecture.The encoder uses Res2Net network as the backbone network to extract the information of different levels of coal flow characteristics,and the decoder recovers the image size through deconvolution upsampling operation.After constructing belt data set training and testing model,experimental results show that the proposed method can quickly and accurately detect the coal on the belt,with an accuracy rate of 95.5%,and the running time from input network to output of each image is 4.8s.The re-sults show that the method is practical and effective.