A mining belt conveyor coal amount detection method based on binocular vision depth perception is provided. A speed sensor is installed on a belt conveyor to collect speed signals of a conveying belt in real time and transmit the signals to a host computer. Two parallelly placed cameras are installed above the conveying belt to collect conveyed coal material images in real time and transmit the images to the host computer for image analysis and processing. The specific steps are as follows: a multi-resolution wavelet transform algorithm is used to enhance the conveyed coal material images, and a K-means clustering algorithm is combined to segment the images only with the coal material; a binocular vision method is used to acquire three-dimensional point cloud information of the coal material; and a Delaunay algorithm is used to calculate an initial volume of the conveyed coal material, a T-S fuzzy logical reasoning method is combined to correct the volume of the conveyed coal material, and a coal amount formula is applied to achieve detection of the amount of the coal material. The method of the invention can measure the amount of the currently conveyed coal material in real time according to the characteristics of the coal briquette on the surface of the conveyed coal material, has few errors and high effectiveness, has a practical value and is convenient for popularization.