Research On The Three-dimensional Reconstruction Technology Of Plant Based On Binocular Stereo Vision

With the rapid development of digital agriculture and computer vision,the three-dimensional reconstruction technology of plants began to emerge in agricultural application.The real record of the three-dimensional information on the surface of the measured object and the shape characteristics of the object is more conducive to the analysis and processing of the measured object.At present,there are some problems in the research of plant 3D reconstruction technology,such as expensive and bulky instruments.In this paper,the research of plant 3D reconstruction is carried out by using low-cost portable binocular camera combined with binocular stereo vision technology and point cloud processing technology,and high-precision reconstruction results are obtained by flexible data acquisition method,so as to promote the application of 3D reconstruction technology in agriculture with the advantage of low cost.The main research work is as follows:(1)The imaging model of binocular camera and camera calibration method are studied,and the mapping relationship between three-dimensional object points and two-dimensional plane image points is clarified.The checker calibration method is used to calibrate the camera,and the re-projection error is introduced to analyze the calibration results.The calibration results show that the average re-projection error is 0.15 pixels,which meets the experimental requirements.(2)The SGBM algorithm has been improved.Aiming at the visual matching problems of the reconstructed scene such as weak lighting conditions,less texture information,and incomplete camera lenses,the analysis and research found that the matching problem is the problem of obtaining matching left and right disparity maps,and the use is highly efficient.,The improved SGBM stereo matching algorithm with low false matching rate,strong robustness and fast running speed obtains the disparity map,and then obtains the initial point cloud data,and realizes the visualization of the point cloud.(3)In order to solve the problem that noise point cloud data exists due to the occlusion of objects,planting environment and other factors,the processing methods such as outlier cloud filtering,cutting ground point cloud data and denoising are studied based on C++ library function PCL.Finally,the experimental results show that the method is effective and feasible.(4)Based on Visual Stdio 2013 C++,the 3D reconstruction of point plant cloud data is realized.Taking the Phnom Penh tiger’s tail orchid plant as the processing object,combined with the greedy projection triangulation reconstruction algorithm,the three-dimensional reconstruction of the Phnom Penh tiger’s tail point cloud is realized.The experimental results show that the maximum absolute error of the reconstruction is 0.47 cm and the maximum relative error is 6.10%,which can meet the requirements of requirements for reconstruction accuracy.Finally,the reproducible experiment verifies the portability of its algorithm in the process of plant three-dimensional reconstruction.