Research on Similarity Algorithm between Vehicle-borne Laser Point Cloud Images and Panoramic Images
Aiming at the obvious difference between the depth or intensity images obtained by projecting the laser point cloud onto the plane and the true color images. This paper presents an algorithm for calculating the similarity between vehicle-borne point cloud pro-jection images and panoramic images. The algorithm first converts the point cloud data into a plane projection images, then calculates the gradient direction of each pixel in the projection images and panoramic images respectively, obtains the histogram of the gradient direction according to the gradient direction, and finally calculates the correlation coefficient between the point cloud projection image and panoramic image according to the gradient direction histogram. The size of correlation coefficient reflects the similarity of matc-hing. The innovation of this algorithm is to improve the calculation of gradient direction, normalize the 360 degree gradient direction angle to less than 180 degrees, which can effectively deal with the inconsistency of gradient direction caused by different pixel attrib-utes between point cloud projection images and panoramic images, and improve the matching accuracy.
Laser point cloudpoint cloud depth imagespoint cloud intensity imagesgradientmutual information
万方数据
维普
