Research On Three-Dimensional Point Cloud Processing And Reconstruction Method Of Corn Plant

The digitalization and visualization of crop growth process have important theoretical significance and application value for scientific and precise field management. The reconstruction of three-dimensional point cloud model is the base of the visualization of crop growth process. In order to solve such problems as low computational efficiency caused by large volume of point cloud data and imprecise reconstruction result because of the influence of noise, this paper takes corn plant as the research object to carry out the research of three-dimensional point cloud processing and reconstruction method and develop a system for corn point cloud simplification, denoising, smoothing and reconstruction. Firstly, the system used octree to establish the spatial relationship between points, based on this, simplify the three-dimensional point cloud data, then adopted a step-by-step processing strategy to denoise and smooth the point cloud model, finally, Delaunay triangulation method was used to reconstruct three-dimensional point cloud model. The main contributions of this paper are as follows:(1) The original point cloud model contains much redundant data which would take up more storage space and lead to low computational efficiency, in order to solve these problems,this paper proposed a point cloud simplification method based on octree partition. This paper simplified the three-dimensional point cloud data based on the spatial relationships between points established by octree. In each voxel of octree, the point which was composed of average number in all dimensions of all points was used to replace all the points in the voxel.In point cloud simplification experiments, different simplification thresholds was set to simplify the point cloud model, and we found that when the simplification rate reached95.88%, the point cloud model can still maintain its original shape characteristics.(2) The noise in corn point cloud model would influence the reconstruction effect, even lead to the distortion of reconstructed model, to solve these problems, this paper proposed a point cloud denoising and smoothing method based on adaptive density clustering and bilateral filtering method. The method adopted a step-by-step strategy to denoise and smoothpoint cloud model. Firstly, adaptive density clustering algorithm was used to remove points which are far away from the target point cloud, and then bilateral filtering method was used to smooth three-dimensional point cloud model. Through the denoising and smoothing experiments of corn point cloud model, it can be found that denoisng and smoothing method could be applied faster on corn point cloud model after simplification, and the consumed time reduced from 101.162 s when simplification threshold was 10 to 1.779 s when simplification threshold was 80. Besides that, the method can effectively eliminate the noise in the three-dimensional point cloud model, as well as make the point cloud model surface more smooth.(3) This paper realized three-dimensional point cloud reconstruction, and analyzed the reconstruction efficiency. Firstly, the original three-dimensional point cloud model and the model after simplification, denoising and smoothing were reconstructed respectively, then the we analyze the model reconstruction effect and compare the reconstruction efficiency. By comparing the experimental results, we found that the proposed simplification method could improve computational efficiency significantly, and the construction of point cloud model after simplification with simplification threshold 80 consumed 0.7s less than the construction of the original point cloud model. And the model after simplification, denoising and smoothing can be reconstructed faster and better, as well as keep the original shape of three-dimensional point cloud model characteristics.(4) This paper developed a processing and reconstruction system for point cloud. The system realized such functions as the reading three-dimensional point cloud data in txt format,adjust the point cloud model to show better display effect including translation and rotation,octree partition, simplification of point cloud model, denoising andl smoothing and Delaunay triangulation.