Abstract: Due to the influence of dust, water mist and low illumination environment in coal mine, it is very difficult to accurately identify the monitoring images of belt transportation system. Aiming at the problem of poor image processing results and efficiency of existing dust and fog removal methods, a dust and fog removal and enhancement method for low-illumination environment images based on dark-bright channel segmentation and fusion is proposed. Firstly, the channel difference is corrected by threshold segmentation combined with gamma transform to solve the problem that the difference of pixel values between the regions with large dust and fog concentration and other regions is not obvious due to the influence of low illumination environment. After correction, the global atmospheric light intensity which is more in line with the actual situation is obtained by guiding the original image to do guided filtering. Then, in order to solve the problem that the dark channel prior fails in the area with large dust concentration, the bright channel prior is introduced to supplement, and the channel component is used to assist the fusion of dark channel and bright channel transmittance, so as to avoid the problem of edge pixel attribution caused by multiple segmentation. Finally, the RGB image after dehazing is transferred to HSV space, the brightness component is histogram equalized and the brightness component before and after equalization is weighted and fused. The objective index evaluation is used to select the optimal aggregation weight for aggregation. At the same time, considering the saturation loss in the dehazing process and the correlation between the brightness component and the saturation component, the saturation adaptive correction function is proposed to correct the image saturation and keep the tone component unchanged. Then the image is transferred back to RGB space to obtain an image with moderate brightness, rich information retention and bright color. In order to verify the effectiveness of the proposed method, subjective vision, objective indicators, and target detection accuracy and confidence are used to compare the algorithms. The experimental results show that the proposed method is superior to the comparison algorithm in the above four indicators, and the image details are retained more abundant and the visual perception is better.