Machine Vision Based Seedling Recognition And Replanting Path Planning Of Potholes

Modern agricultural development technology has gradually stepped into the research stage of high-tech application.The key of agricultural production is how to obtain high-yield crops,which puts forward certain requirements for the physiological parameters of crop seedlings at the early stage.Transplanting operation is one of the key links of seedling production system.When transplanting,it is necessary to monitor the physiological state of seedling leaves,so as to judge whether the seedlings can meet the transplanting requirements.The recognition technology based on machine vision can determine the physiological information such as crop unhealthy seedling and no seedling in the hole,which provides a certain technical basis for the feedback actuator,and is of great significance to improve the transplanting efficiency of seedlings.Under the support of the agricultural research and development project(2015-NY02-00285-NC)sponsored by Chengdu Science and Technology Department,this paper studies the health recognition treatment and path planning of the seedlings of leafy species under the background of the hole disk framework,fertilizer impurities,seedlings crossing the border,etc.For image filtering,an improved iterative median filtering algorithm is proposed to eliminate the interference of noise.In order to improve the recognition accuracy of seedling health status,the search radius of simple connected domain algorithm is improved.Three kinds of intelligent optimization algorithms are used to model and simulate the seedling path of different specifications,and the strategy of path planning based on particle swarm algorithm is proposed to improve the real-time performance of the seedling repair.The main research contents are as follows.1.This paper introduces the principle of transplanting vision system,selects the camera model,image acquisition card,image processing computer and other hardware parts,installs and configures the image processing programming platform and computer image vision database,and builds transplanting vision system.Canny operator with low threshold value of 65 is used to detect the edge of the image of the disk,Hough transform is used to extract the straight line of the contour,calculate the horizontal tilt angle?,establish the correction model of the disk tilt,and projection method is used to extract the key area image of the disk.2.This paper analyzes the imaging geometry principle of images in space,establishes the parameters and distortion coefficient model of vision system through theoretical analysis,and makes a special chessboard calibration template based on Zhang Zhengyou's calibration algorithm theory,and carries out calibration experiments to obtain the parameter matrix and distortion coefficient of vision system.3.The best grayscale factor 1.8g-b-r was obtained by analyzing RGB spatial model,and the image of seedling's disk was grayscale enhanced.An improved iterative median filtering algorithm is proposed to eliminate the interference of noise on the image.The PSNR and SSIM are used to evaluate the image quality.According to the gray level distribution of the image after denoising,the Otus segmentation algorithm is used for binary segmentation to determine the best segmentation threshold.The method of 0-255-0 double fixed threshold is used to extract the frame of the cavity disk,and the projection histogram is used to identify it.The image proportion method is used to measure the leaf area of the seedlings in the hole.Aiming at the problem of the seedlings'leaves crossing the boundary,an improved single connected domain algorithm is proposed to extract the pixels and the area of the seedlings'leaves.The threshold value of leaf area discrimination is set to 50mm~2,which is used to identify the healthy seedlings,the unhealthy seedlings and the lack seedlings.The accuracy of recognition is analyzed according to the area results.4.The spatial layout of the target tray and the seedling tray was established,and the coordinate position of the seedlings in the tray was analyzed.Based on the theory of business travel problem(TSP),this paper analyzes and models the path of mending seedlings.In view of the common 50,72 and 128 hole size of the tray,the fixed order method,particle swarm optimization algorithm,genetic algorithm and greedy algorithm are used as four supplementary planting strategies to plan and simulate the different size of the tray.According to the simulation results,the total path length,optimization efficiency and running time of four different planning strategies are analyzed and discussed.A path planning scheme based on particle swarm optimization(PSO)strategy is proposed to meet the real-time requirements of replanting.