CLASSIFICATION OF HIGH DENSITYMOBILE MAPPING LIDAR DATA IN URBAN LANDSCAPE

Abstract

With the advent of new technologies like laser scanning, methods can be developed to overcome the problems faced in classification of urban areas. LiDAR is an emerging remote sensing technology for topographic mapping. From this LiDAR survey, the datasets generated are of two types. One, scanned image of the area and second, information about the x, y and z coordinates of the scanned points. Eventually, the data used from LiDAR for classification is commonly in form of points. The information about x, y and z of the points can be stored as text files; however the size of these files may be huge. The complexity and sheer volume of points in typical LiDAR dataset makes it difficult to work with. Thus, it's a demanding task to develop an efficient method to handle a large volume of point cloud data and classify the 3D laser scan data accurately in reasonable time. If a method of directly classifying LiDAR data is developed, then it would initiate a new path for various applications. In general, it is important to classify street scenes, in order to understand the environment (image/ scene understanding). The applications of classifying the LiDAR data are diverse, for example, cities may like to have an inventory of their street signs, traffic lights and trees. Driver assistance and future autonomous driving systems may want the background information to interpret the scene they see with their cameras, scanners, radar. Systems which require exact positioning may use this information to prevent the usage of feature points which are located in structures which are variable over time, such as in vegetation.