A 3D Vehicle Recognition System Based on Point Cloud Library

In automatic steel plants, the accuracy for positioning steel coils and grooves is vital to perform a successful loading/unloading operation. With development of sensor technology, point cloud-based methods have been widely applied in positioning system. In this paper, a real-time 3D vehicle recognition system based on Point Cloud Library(PCL) is developed to position coils and steel grooves on the truck floor. The vehicle recognition system's tasks consist of the coil's positioning and the groove's positioning for unloading tasks and loading tasks, respectively. For the coil's positioning, a new mean projection method is proposed to improve the accuracy and efficiency of the center of cylindrical coils recognition. For the groove's positioning, a PCL-based architecture with edge detection and region growing method is introduced to segment point cloud based on edge points and recognize grooves with volume & normal features. Experimental results demonstrate that the proposed mean projection algorithm and grooves recognition method can effectively improve accuracy and efficiency of coils recognition and grooves recognition.