Research On 3D Reconstruction Of Cotton Canopy Based On 3D Point Cloud

With the rapid development of science and technology,a large number of high-precision,high-efficiency,fast and non-destructive acquisition of threedimensional point cloud data equipment and applications have been developed.In order to achieve rapid and efficient extraction of individual phenotypic parameters and dynamic monitoring of cotton phenotypic parameters,Plant Eye was used to obtain multi-time series cotton phenotypic parameters and point cloud data,and manually measured the three phenotypic parameters of plant height,leaf area and leaf area index,studied the denoising,streamlining and three-dimensional reconstruction methods of point cloud data,obtained better cotton canopy reconstruction effect,and evaluated the fitting between the measured values and measured values of the analysis phenotypic platform.The research content and conclusions of this paper are as follows:First of all,in view of the unnecessary point cloud noise that will be affected by the natural environment and operation methods in the experimental data collection process,the radius filtering method is used to remove the noise point cloud on the ground,and for the small noise point cloud around the cotton,a denoising method based on bilateral filtering is proposed,which can effectively remove the points with large differences and retain the edge characteristics,and the application effect in the cotton point cloud data is also significant.Secondly,aiming at the problems of low reduction rate and long running time of traditional methods,a point cloud reduction method based on Voxel Grid grid voxelization is proposed,and setting the appropriate grid edge length is an important step to achieve reduction.It can be obtained through experiments: the number of point cloud data after denoising treatment is 59919,the edge length of the cube grid is set to 0.1f,0.3f,and 0.5f,and the number of point clouds after reduction is 49175,20424,and 4058,respectively;when the edge length is set to 0.5f,the reduction rate is 93.8%,the reduction effect is relatively good,and the surface characteristics of the point cloud data can be effectively retained.Then,the cotton canopy reconstruction takes the method of Poisson reconstruction,greedy projection triangular meshing.Poisson’s reconstruction process is a process of forming a convex package according to the distribution of points,but due to the different densities of cotton point cloud data,the density of point clouds in the central location and the phenomenon of upper and lower layering,the reconstruction effect has a false patch generation,which is not suitable for the point cloud data collected in this paper.Compared with this method,this paper uses the greedy projection triangular meshing method to achieve three-dimensional surface reconstruction,which realizes the establishment of topological three-dimensional space for point cloud data and the three-dimensional model construction of cotton leaves.Compared with the Poisson method,the effect of refactoring is good,and the time of refactoring is also reduced.Finally,due to the traditional greedy projection triangulation reconstruction of cotton leaves,there are holes,edges are not smooth and other phenomena.Therefore,an improved algorithm is proposed for these problems,and the point cloud data is smoothed first,and the method used is the least squares method MLS,which mainly realizes the greedy projection triangulation reconstruction of the cotton canopy through the normal vector calculation of the points.The experimental results show that the number of point clouds after MLS reduction is 3899,the reduction rate is 94.1%,which is about 0.3% higher than the effect of voxel reduction,from the analysis of the reconstruction effect,the point cloud data with the same degree of reduction,the effect of the point cloud smooth denoising treatment will be better than the reconstruction effect without smoothing treatment,the reconstructed cotton leaves are not easy to distort,and the number and time of the improved algorithm reconstruction patches have been improved...