Research Of Intelligent Environment Detection Technology With TOF Camera

With the rapid development of computer and artificial intelligence technology, intelligent mobile robot has become the research focus. Intelligent environment detection is one of the key problems of intelligent mobile robot. It includes object localization, obstacles identification,3D reconstruction and 3D map automatic generation in unknown environment. These researches have very important theoretical significance and engineering value to improve the research level of intelligent mobile robots in unstructured environments.The dissertation proposes a 3D environment detection method with TOF camera, including point cloud acquisition, point could segmentation and localization,3D scene reconstruction. Firstly, a point could acquisition method is proposed with TOF camera SR4500, a point cloud acquisition software is designed to acquire 3D point cloud and 2D gray image. Secondly, the points beyond area-of-interest are removed from the 3D point cloud by a pass filter, and the isolated noise points are removed using point cloud gridding filter. The quality of point cloud was improved by these two filters. By comparing the experiment result of three different segmentation algorithms, the segmentation algorithm based on Euclidean clustering is chosen as the most effective segmentation method. After segmented with this method, point cloud clusters of different single objects are setup. For the single object point cloud, ground is detected with the height distribution histogram, other objects' position and size are analyzed with the convex hull method.3D reconstruction is realized with point cloud sequence splice when the camera is moving in the environment. In order to reconstruct the 3D scene, the reflected light intensity is standardized into a gray image. The characteristics of adjacent images are extracted and matched with SIFT operator. The corresponding 3D matching point set are found out and used to calculate the camera motion matrix R, T with ICP algorithm. R and T are then used to align and splice two adjacent point cloud. The 3D scene reconstruction is realized by continuous stitching.Experiments of 3D point cloud segmentation, object localization and 3D reconstruction in different scene verified the effectiveness of the algorithm.