Research On Point Cloud And Image Registration In Vehicle Moving Measurement System

With the introduction of new concepts such as "Digital Earth" and "Smart City",the construction of urban informatization has put forward new requirements for multi-resolution massive data collection and processing,such as 3D data visualization and other basic geospatial information.Mapping industry,one of the cutting-edge technology,automotive mobile measurement system technology for the emergence of the wisdom of the city has provided a strong technical support.The mobile measurement system is widely used in the application of digital map,basic surveying and mapping,urban 3D modeling and so on,which is based on its rapid three-dimensional spatial information acquisition and automatic processing technology.The vehicle motion measurement system has thus become a research focus in the current context.In the field of automotive mobile measurement system research,point cloud and image of the precise registration and integration problems are still not well resolved,on the one hand,point cloud data itself is not high enough accuracy,on the other hand,the use of wide-angle lens,Large field angle lens distortion is difficult to correct,the image is difficult to accurately match with the point cloud data.Based on this,this paper studies the point cloud and image data registration of high precision vehicle motion measurement system as follows.First of all,according to the principle of vehicle movement measurement system acquisition,from the time registration and spatial registration point of view,in-depth exploration of data acquisition is completed after the acquisition of multi-source data organization and correspondence,in order to better point cloud data and Image data registration and integration,this paper design data file organization directory,and the use of C + + programming language development of three-dimensional laser point cloud and image registration software,the two data display and organization and management.Secondly,a high precision point cloud and image registration algorithm based on the same feature pair is proposed.This method divides the point cloud and image registration problem into camera parameters.The parameters include the internal parameters of the camera and the external parameters of the camera in the panoramic camera system,and the exact pose of the panoramic camera in the carrier coordinate system.Respectively,to give a solution.The camera is calibrated by setting up the calibration field,and the parameters in the camera and the external parameters of the camera in the panoramic coordinate system are solved based on the equations of the same name feature points in the calibration field.Then,the cloud is segmented according to the time alignment.The distance between the panoramic coordinate system and the platform coordinate system is constructed and solved,and the three-dimensional coordinate points of the world coordinate system are solved by the two steps.The relationship between the pixel coordinates on the image photo,thus completing the registration of the cloud and the image.Finally,the algorithm is tested and verified by collecting multiple actual data(such as turning road area and straight road area)in complex urban road environment,and the accuracy is evaluated from both qualitative and quantitative aspects.The experimental results show that the high precision point cloud and image registration algorithm based on the same feature point pair proposed in this paper can accurately combine the point cloud with the image information to verify the effectiveness of the algorithm and have great practical significance.