Environmental protection is an increasingly important problem, especially in coal ports. A large amount of dust is generated during the unloading of bulk coal. To achieve environmental protection, the first step is to detect coal dust. But, it is difficult to recognize dust from images due to variance of dust color, texture and shape. To solve this problem, we propose Coal dust Segmentation based on a priori information. Our method is a four-channel segmentation network model which are guided by a priori information. In our semantic segmentation algorithm, the input consists of RGB channels and a dust channel. The dust channel is obtained from dehazing model, which is used as a priori information to guide the segmentation network. Compared with other target segmentation methods, our method can accurately segment the dust region, It has higher accuracy and stability.