Coal stacking identification method of belt conveyor based on surface reconstructio

The existing coal stacking identification method of belt conveyor has problems of false trigger alarm and high cost. In order to solve the above problems, a coal stacking identification method of belt conveyor is proposed. The method uses infrared structured light technology to quickly reconstruct the coal flow surface of belt conveyor.Firstly, the depth map of the coal stacking is obtained by using infrared structured light technology. Secondly, the depth map is mapped to a point cloud map, and the point cloud data is used to construct a convex quadrilateral network. And the convex quadrilateral network is triangulated by the approximate Delaunay subdivision method to complete the reconstruction of the coal stacking surface.Finally, according to the distance from the triangle vertex to the camera and the proportion of the triangle area whose vertex distance is less than the threshold to the total area, it is determined whether there is a coal stacking accident.The approximate Delaunay subdivision method replaces the insertion sorting process with the traversal process. There is a small probability of not satisfying the Delaunay property, but the algorithm complexity is low. Therefore it can improve the real-time performance of coal stacking identification.The experimental results show that the infrared structured light technology improves the algorithm's robustness to illumination effectively. The success rate of the approximate Delaunay subdivision method is 99.466 1%, and the surface reconstruction time of the approximate Delaunay subdivision method and the classic Delaunay subdivision method under the same conditions is 1.28 ms and 134.93 ms respectively. The approximate Delaunay subdivision method improves the calculation speed greatly when the accuracy meets the application requirements. By setting an appropriate threshold, the number of missed detection and the number of false detection are both 0. The statistics of the processing time of a large number of images show that the processing time of each frame is less than 20 ms, which meets the real-time requirements.