Abstract: In view of problems of low detection precision of existing detection methods of belt longitudinal tear and difficulty to eliminate influence of coal mine complex environment, a detection method of belt longitudinal tear based on support vector machine and infrared image segmentation was proposed. Firstly, the infrared image of belt longitudinal tear is collected. Then, the infrared image is segmented by the method of support vector machine. Finally, the belt longitudinal tear or its tendency is tested accurately by calculating quantity of torn pixels. The test results show that the image segmentation time of the method is short and detection precision is high to 99.1%.