Research On Auto-Fusion Of Point Cloud And Image Based On Laser Reflectance

Terrestrial 3D laser scanneris can obtain high precision 3D coordinates and laser echo intensity reflectivity information of point cloud,and images are visual display of 2D high resolution true color information.The fusion of the two technologies realizes the 3D true color expression of the objective world,promoting the progress and development of numbers of industries and applications,such as virtual libraries,games and entertainment applications,display and restoration of buildings,archaeology,virtual reality and tourism.Therefore,the fusion of both has become an important issue.Based on the point cloud attribute information,such as the laser echo intensity reflectivity and the depth value,the corresponding gray image is projected on the two-dimensional image plane.The physical spacial coordination of the image points is obtained in the way of digital image with the point cloud reflectance image feature matching.Without artificial participation,efficiently and accurately,the method implements automatically.High-precision transformation model of point cloud coordinates and image coordinates is the key step of fusing.However,it is still difficult to match the reflectance image and optical image,because of the different imaging mechanism,laser diffuse reflection and distance attenuation effect.Aiming at improving the matching accuracy of reflectance image and optical image,this paper studies the technology involved in the process of fusion of point cloud and image.The main research contents are as follows:(1)The advantages and disadvantages of different projection models to generate reflectance images and the suitable scanning objects are analyzed.(2)The concept of threshold truncation parameter is proposed,which can reduce the variation range and increase the gradient of reflectance.After preprocessing,the image matching accuracy is improved,and the visual display effect is better;The optimal threshold is determined by comparing the matching points and the spatial distribution.(3)Different feature extraction and mismatch elimination algorithms are studied;An improved RANSAC algorithm is proposed,which is based on the rotation invariance of feature points.(4)The fusion accuracy evaluation method is explored,and the absolute accuracy of fusion is analyzed qualitatively and quantitatively.The fusion accuracy evaluation index is defined as the difference of the point cloud coordinates of the same points of two optical images.