Research And Implementation Of Wolfberry Plants Phenotypic Measurement System Based On 3D Reconstruction

The plant phenotypic parameter is one of the crucial bases for evaluating crop varieties in selective breeding.With the rapid development of computer technology,the application of machine vision in agriculture is becoming increasingly widespread.Apparently,the phenotypic measurement based on two-dimensional image has the superiority of lower cost,easier operation,and nondestructive measurement.Nevertheless,the complexity morphology of plants and the diversified barrier of leaves tend to generate partial information loss,and ultimately resulting in experimental data errors.Relatively speaking,the phenotype measurement based on three-dimensional data can vividly represent plant growth,which is an important way to comprehensively and precisely measure plant phenotype parameters.As one of the specialties of Ningxia Hui Autonomous Region,Wolfberry enriches both high nutritive and economic value.Furthermore,accurately obtaining the phenotypic trait parameters of wolfberry is of great significance for acquainting its growth status,yield estimation and so on.In order to research how to obtain the three-dimensional data of wolfberry efficiently and thoroughly,and how to upgrade the three-dimensional reconstruction of it,so as to make a more precise measurement on its phenotypic information,this paper analyzes the common technical methods and problems in the measurement of plant phenotypic parameters based on three-dimensional data and investigate the key technologies for acquiring point cloud data,point cloud completion,and phenotypic parameter measurement of wolfberry.and then proposes a research framework and some feasible approaches,and research the key technology such as point cloud data acquisition and completion,phenotypic parameter measurement regarding wolfberry plants.The main content of this paper is as follows:(1)To address the problem of missing wolfberry leaves caused by the three-dimensional reconstruction of wolfberry plants,the paper proposes a deep learning-based point cloud complementation algorithm that combines a geometry-aware module with Transformer,which uses a geometry-sensitive Transformer encoder-decoder structure to complete the point cloud complementation task and accurately recover the missing information of wolfberry leaves.(2)The paper proposes a three-dimensional point cloud-based method for extracting phenotypic parameters of wolfberry plants,which uses a point cloud feature segmentation method to complete the stem and leaf segmentation of wolfberry.Besides,using distance-based most traversal,greedy projection and minimum enclosing box to achieve fast and accurate measurement of parameters,such as plant height,stem thickness and leaf area of wolfberry plants,which improves the efficiency of phenotypic measurement of wolfberry plants.(3)A system for measuring the phenotypes of wolfberry plants is designed.After obtaining accurate three-dimensional information of wolfberry plants,the three-dimensional point cloud of wolfberry is automatically processed,which includes point cloud feature extraction,point cloud downsampling,point cloud filtering and extraction of phenotypic parameters of wolfberry plants to monitor the dynamic growth process of wolfberry plants,which is of great significance for monitoring the growth status of wolfberry plants and guiding breeding.