Design And Implementation For Point Cloud Denoising Software Of Corn Leaf

The 3D reconstruction of crop model and visualization are the Key matters in agricultural informatization research fields at home and abroad.In order to get the high quality 3D point cloud model,point cloud denoising has become a popular research area.For the existing,the image denoising method is applied to the point model,however a single algorithm is difficult to satisfy requirements of various morphological point cloud denoising.In this paper,the maize leaf algorithm and parameter are studied,and we designed and developed corn leaf point cloud denoising software based on the PC and mobile device.The main content of this paper is as follows:(1)A method of mobile platform and PC combined with point cloud denoising was proposed.Because of excessive point cloud and number of outlier noise,which results in slow efficiency.And we are not able to denoise in mobile terminal directly.Thus,we analyze the point cloud simplification algorithm and denoising algorithm,and combined with the characteristics of maize leaf and platform.was used Voxelization grid down sampling simplified algorithm to simplify the point cloud.And was used K nearest neighbor distance statistical denoising algorithm to denoise on PC,which can remove apparent noise by adjusting the times of mean square deviation and the number of K-neighboring points.Bilateral filter denoising algorithm was used on mobile device,which can remove small scale noise points while maintaining the edge feature of leaf.(2)Design and implement software for point cloud denoising.Aiming at this problem that the scope of parameters of denoising and simplification algorithms was too wide to choose appropriately for ordinary users.,we used the different number of point cloud to denoise,experimental results show that the threshold which is the times of mean square deviation has great influence on point cloud denoising,if the threshold is too big,the result of denoising is not obvious;if too small,which will cause the emergence of a large number of hole,and the point size within a K nearest neighbor value will protect the blade shape.When the number of point cloud of corn leaf is about 80000,the range of voxelization grid is between 1.2? 2.6.Reduced rate is60% ? 30%.Threshold can be changed from 1.2 to 1.6.The result is better when the number of neighboring points is about 80,thenthe small-scale noise removal is carried out on the mobile device,Finally,normal vectors which are disorganized becomed ordered after denoising.the denoising time is less than 15.3s,and the ideal result is obtained in the effect and time.(3)Software testing.For the potential problems of this software,detailed test cases were design.By using data testing and functional testing,the software did not break down when inputting the wrong or illegal data.What's more,good results were shown when inputting reasonable range of data.Alarm will occur if input data beyond the scope.Therefore,all the functions of the software were implemented and tested as well as in mobile device in this paper.