Study Of Laser Scanner Calibration And Vehicle-Borne Laser Point Cloud Classifacation & Vectorization

Laser scanning, which is rapidly developing in recent years, is seen as a new technique revolution in surveying & mapping field. It can get density and definition point cloud of a large scene in a short time. Vehicle-borne laser data collection is a new technique in laser scanning, and the researchers begin to study it. Vehicle-borne laser data collection system integrates many sensors such as GPS, IMU, LS and CCD. It can get point cloud and image of the scene on the two sides of the street. Because the system integrates many sensors, and the point cloud has a large amount of information, there are many key techniques need study for data collection and data processing.The main work in this paper is laser scanner calibration, the classification and the vectorization of the point cloud got by vehicle-borne laser scanner. The innovation works include:1. Propose and realize an algorithm to calibrate a laser scanner and a camera. The algorithm is based on the homologous lines in the same plane. Design the calibration scene and made the calibration models. Analyze the algorithm result and give some advice to get better result.2. Propose and realize an algorithm to classify the point cloud got by vehicle-borne laser scanner. The algorithm is based on multi-resolution and step-classification. Because there are various objects and disturbing information in the point cloud, it is difficult to extract the objects directly. In the algorithm, the advantages of high resolution image and low resolution image are used, and the objects are extracted in several steps.3. Propose an algorithm to realize the vectorization of the point cloud got by vehicle-borne laser scanner based on point-line-face mode. The point cloud processing method in surveying & mapping field is different from that in reverse engineering field. The point cloud in surveying & mapping field has a lot of disturbing points, and the methods must remove them. The point cloud got by vehicle-borne laser scanner has more complex object structure and more disturbing points than the point cloud got by air-borne laser scanner or fixed station scanner. The vectorization in this paper is based on scan line. The algorithm of point-line-face vectorization includes the following steps. Firstly the points are sorted in two directions and the scan lines are formed. Then line segments are fitted in each scan line. Finally faces are fitted by the line segments.