Abstract: In the coal transportation process, it is often impossible to maintain the uniform transportation volume of coal on the belt conveyor, which causes the belt conveyor to run at full speed for a long time, resulting in waste of electricity and ineffective equipment wear. This paper proposes a semantic segmentation-based belt conveyor coal transportation area detection algorithm. Based on DeeplabV3+, the algorithm introduced an attention mechanism according to the interdependence between characteristic channels, and used convolution kernels with different expansion rates to obtain semantic information of multiple scales and precisely segment the coal transportation area on the belt conveyor. The experimental results show that the mean intersection over union (mIoU) of the algorithm is 1.24 percentage points higher than that of the DeeplabV3+ algorithm, which can effectively and accurately segment the coal transportation area, and provide an effective guarantee for the estimation of coal volume.