Building Feature Extraction From LiDAR Point Cloud And CCD Image

Light Detection and Ranging (LiDAR) is a high-tech developed in recent years which can acquire real three-dimensional ground data quickly and accurately and widely used in city building extraction. It is difficult to extract features directly using LiDAR points cloud because their lacking of texture information, uneven density and discontinuous spatial distribution characters. However, with CCD images, the weakness of texture can be offset, and then the building features can be extracted accurately.Based on expounding LiDAR system's working principle, application and the points cloud data's characteristics systematically, I studied and analyzed algorithms that extract the 3d buildings'characteristics by combining the LiDAR points cloud and image both at home and abroad, and puts forward to segmenting the points cloud data to get buildings' point first, uses irregular triangle nets to interpolate many times and interpolates encryption points cloud data to create images, and then extracts accurate information of building feature with information extracted from projective images. During points cloud data segmentation process, according to the height of the building area to extract initial points cloud of buildings, and this paper puts forward a new points cloud data buildings segmentation method-eight neighborhoods searching method, and realizes fast and accurate segmentation of buildings data, and extracts the points cloud of buildings that has excluded noise points. In the process of CCD image building feature extraction, this paper realized the optimal selection of the LiDAR points cloud data's expression and interpolation method. According to the characteristics of the selected experiment area, combining the characteristics of color features and shape of buildings to segment experiment area, this paper extracted a rough building area. Using the method based on morphology to extract the accurate building feature; after having extracted building features from two different data source, I studied the method that is to combined the image matching method based on gray-level method and the method based on features to extract 3d building features, and optimized to chose matches base units, looked for the essential attribute of base units, established corresponding matching standards (constraint conditions), selected the reasonable matching strategy, designed a good algorithm structure, and finally I used matlab programming language to realize the matching process, and used measured data to carry on accuracy verification.