Abstract:Based on the analysis of the noise data,a method based on density clustering and fast bilateral filtering was proposed to improve the three-dimensional reconstruction accuracy.Firstly,the density clustering algorithm was used to deal with large-scale outliers,and then the small-scale noise was smoothed by fast bilateral filtering.Three-dimensional point cloud data of rice and cucumber obtained by multi-view stereo matching were denoised.Experimental results show that the denoising time of the proposed method is 48.1％ and 45.7％ of the traditional bilateral filtering and good effect is achieved.The proposed method can quickly and accurately remove the noise of different scales,preserve the integrity of the plant edge,and get good three-dimensional point cloud data.

Denoising method of 3D point clouds based on multi-angle stereo matching