Font Size: a A A

Path Recognition And Autonomous Cruise Of Agricultural Greenhouse Robot By Machine Vision

Posted on:2021-09-11Degree:MasterType:Thesis
Country:ChinaCandidate:W B TanFull Text:PDF
GTID:2493306518985049Subject:Master of Agriculture
Abstract/Summary:PDF Full Text Request
With the advancement of science and technology,the influence of intelligent production in the agricultural field is also increasing.The problem of robot movement is the most basic problem faced by intelligent equipment in actual production.The path recognition and autonomous cruise of agricultural greenhouse robots based on machine vision researched in this paper will help the robot improve its autonomous movement ability in agricultural greenhouses.The main content on this paper is as follows:1.Through field investigation of the layout and planting of strawberry greenhouses,the installation position and angle of the robot camera were determined to ensure that the robot can collect effective images.Through the test and comparison,the super-green method is selected for the gray-scale processing of the image,and the Gaussian filter is selected as the image filtering method in the pre-processing to remove image noise.2.Using RGB color space theory,the data analysis of the greenhouse road environment is carried out.Proposed the super green method + OTSU,over red index + OTSU,and custom threshold G-B + X three ways to binarize the image,and finally obtained the super green method+ OTSU processed image is suit for this project.3.Open operation is selected as the morphological processing method of this project,and the nuclear value is set to 50,which can effectively retain the information of the road,and can fill the large area of the crop directly.This method provides a good test sample for the identification of road types.4.The model of the road in the greenhouse is established,and then the image is converted into a four-dimensional vector by the road proportion.This vector will be used as the training set of the KNN algorithm to identify and classify roads.600 images collected through the camera were stretched and rotated to generate 7800 images,which were put into the road recognition model for testing.Finally,it is determined that the appropriate road width is 30%-40% of the image area,and the appropriate yaw angle is between ±4°.Then,by writing a corresponding execution program on the recognized road,the robot can travel on the road in the greenhouse.5.By observing the layout of the greenhouse,design the walking route of the robot on the map.Through the simulated cruise on this route,find the law of its movement and summarize the route.Then the travel strategy is matched with different movement speeds to put the robot into the field for testing.According to the data,it can be concluded that the robot’s running speed at 7cm/s can better complete the full-scene cruise of the standardized greenhouse.
Keywords/Search Tags:machine vision, road recognition, machine learning algorithms
PDF Full Text Request
Related items