Abstract: Coal weight measurement facilities at the main transport face of belt conveyor in coal mine are not perfect. In the complex underground environment of coal mine,weight measurement method of belt weigher has high requirements for its precision and is easy to be operated by human,which makes the measurement error larger and difficult to obtain the real coal production data;weight measurement system of laser three-dimensional scanning is expensive and difficult to be widely used. According to the shortcomings of coal quantity measurement of main transport surface of belt conveyor in coal mine,a dynamic irregular object weight measurement method based on binocular vision is proposed. Firstly,left and right top view of the main transport working face of the belt conveyor in the coal mine is obtained by using the binocular camera,preprocessing method suitable for the image characteristics of the coal mine is studied. Then,according to the principle of binocular vision,an improved semi global block matching （Semi-Global Block Matching，SGBM）algorithm is used to calculate dense parallax map,holes in the parallax map are filled to eliminate the wrong matching points. Finally,parallax map after filling is collected the improved watershed segmentation algorithm is used to separate the raw coal target in the parallax map;depth information of each point is recovered by the method based on triangulation to calculate the three-dimensional information of corresponding points;on the basis of the three-dimensional information,a new dynamic irregular volume measurement method is proposed to calculate the raw coal volume on the surface of belt conveyor;finally,calculation formula of coal quantity is applied Coal forming capacity 〖JP+1〗measurement. In this paper,we〖JP〗〖LM〗use binocular camera to obtain real coal mine images to test the effectiveness of the proposed scheme. Research shows that：hardware equipment of the system is a binocular camera with light weight and small volume,which can collect the target image and distance in real time;error between the real weight and the measured weight is less than 10%,which provides an effective scheme for coal quantity measurement in coal mine,which can meet the estimation demand of actual coal production in coal mine.