The Research On Filtering Method Of Three-dimension Al Laser Scanning Point Cloud Data Based On BP Neural Network

With the mature development of computer technology and the wide application ofgeographic information technology research and development and the concept of digital citypeople in the study of3d information become more deeply, so as to encourage people to howto better access to space3d information demand becomes more urgent, and with lasertechnology, the development of semiconductor technology, computer technology is more andmore mature and more pressing to the requirement of these technologies, which makes3dlaser scanning technology is becoming more and more advanced, such as scanning speed andscanning accuracy become faster, higher, and the range of scanning also becomes more andmore big, so the measurement personnel can in the case of without contact with the terrain,can realize the complex area and complex objects such as irregularly shaped objects morehigh-speed and efficient all-round scan, obtain the three dimensional space information andmore accurate. However to3d laser scanning technology can become more perfect in the field,the development space is more and more big, how to through the3d laser scanner to scan forthe huge amounts of point cloud data for more efficient processing has become the hot topicsin the study of many scholars today. Point cloud data preprocessing is one of the key steps inthe process of data processing, point cloud data filtering is the data preprocessing stage focusand hotspot issues, so on the surface of3d laser scanning point cloud data filtering, forming asystem of ground3d laser scanning point cloud data filtering theory is very necessary.This paper briefly expounds the3d laser scanning technology at home and abroad andthe point cloud data filtering theory, the research status and development according to therules of the arrangement of point cloud data to its classification, filtering methods commonlyused for different types of point cloud data is analyzed, the research content of this paper isbased on the BP neural network is built after the point cloud data filtering model and usingother methods to deal with point cloud data after constructing the model of the contrast, thisarticle USES the other methods for using the regular grid mathematical morphology filteringand the use of point cloud data Geomagic Studio software own filtering tools of point clouddata filtering. The results of three kinds of filter method to carry on the quantitative andqualitative analysis, through the analysis of the test results show that the BP neural networkfiltering method has better filtering effect, the effective to reduce the noise in point cloud dataat the same time, also very good to keep the topography, the constructed model is morerealistic.