Research on high-precision disparity map utilizing depth-based filter
To address problems such as point cloud trailing caused by the characteristics of laser radar scanning and ambiguous projection in occluded regions,this paper proposed a disparity map filtering algorithm utilizing a depth-based filter to process and analyze unstructured point cloud data.First,the iterative geometry encoding volume(IGEV)was utilized to compute the disparity map of binocular depth estimation.Then,the point cloud was projected onto the corresponding color image to obtain the projected disparity map.With the disparity map of binocular depth estimation as a guided image,consistency comparison was performed on the projected point cloud image at corresponding pixel positions.Due to the subpar performance of the laser radar at close distances,this paper innovatively proposed a threshold judgment method to distinguish between foreground and background.Experimental results on the Karlsruhe Institute of Technology and Toyota Technological Institute(KITTI),a computer vision algorithm evaluation dataset demonstrate a 0.72%improvement in detection accuracy of the proposed method compared to the KITTI algorithm,which verifies the effectiveness of this method.
point cloud filteringstereo matchinglaser radarimage processing
国家科技期刊平台
NSTL
万方数据
