A Typical Classification Method Based On Vehicle Laser Point Cloud

Geo-spatial information technology is one of the hottest field in the world, whose three main research topics are Information acquisition, processing and applications. The ideal of digital earth and digital city accelerated the progress and development of Geo-spatial information science. Building Digital city needs a large number of images and point cloud data. Therefore, how to acquire and process those data quickly has become the core and forefront technology in this filed.In recent years, Laser scanning technology has been more and more attentional in various fields. It can obtain information data of the target surface quickly, which has characteristics of high accuracy, authenticity and non-contact measurement. Especially in field of mapping and remote sensing, laser scanning can get high accurate three-dimensional surface information of target objects in a wide range. As a new method of data collection acquisition, vehicle-based scanning technology has been widely used in the geographic information industry. The data obtained by laser scanning from object surface called point cloud, in which contains rich information of feature surface. To extract value information from point cloud, the classification of point cloud is the premise and key study. The current typical methods of classification are presented for airborne laser scanning, and the classification technology of vehicle-based laser scanning data is not mature. Airborne laser scanning mainly get information of object in top, and vehicle-based laser scanning obtains high-precision information of object in the spatial space. Because of the method of point cloud data obtain, many mature classification of airborne laser scanning are not suit for the processor of vehicle-based laser data.To solve these problem, this paper presents a classification method based on space density clustering and typical characteristic feature. This classification method use summary of space regulation and spatial distributing of typical objects, based on space density clustering. This method cannot only complete the classification of different objects, but also extract individual elements feature, which is very importance in extracting feature information and the last modeling. Because point cloud is a very large data set, this paper made an optimization on the calculate of density clustering, which aim to improve effective on calculate, and it also be proved to be correctness and validity by experiments.