Image Analyzing System For Agricultural Inspection Robot In Orchard

Crop growth monitoring is an important part of agricultural production,which can provide basic data for agricultural planting management.Aiming at the problems of image distortion,different scales of visible light and infrared thermal images,an agricultural inspection robot image analysis system based on outdoor dynamic monitoring and multi-content production environment is proposed in this paper.By integrating a variety of algorithm modules,the problems of image correction of agricultural production environment,extraction of plant canopy temperature by matching and fusion of visible light images and infrared thermal images,and large view design of agricultural production environment are solved.The main contributions of this paper are as follows :(1)Aiming at the problem of image distortion caused by optical imaging and distortion,the basic parameters of the camera are analyzed,and Zhang Zhengyou's correction method,camera correction method based on Halcon library,and camera parameters optimization based on improved particle swarm algorithm are studied.Through the self-built agricultural inspection robot platform,the correction test and analysis of six visible light lenses carried by the robot are carried out under outdoor conditions,and the correction effects of the above three algorithms are compared.The results show that compared with Zhang Zhengyou's correction method and Halcon correction method,the reprojection error of particle swarm optimization algorithm for robot image correction is reduced by 0.0319 pixel,which achieves the purpose of effective lens correction.(2)Aiming at the problem that infrared thermal images and visible images are different in scale,it is difficult to register with high quality and thus lead to inaccurate extraction of canopy temperature.Firstly,three methods of image registration based on SIFT(Scale Invariant Feature Transform),multi-scale PIIFD(Partial Intensity Invariant Feature Descriptor)and contour corner feature are studied.By applying the multi-algorithm fusion of this module to the self-built inspection robot platform,the test and analysis of image registration and temperature extraction were carried out under outdoor conditions,and the registration efficiency of the three registration methods in different plants was verified.The results show that the error of the contour corner feature matching method at the inspection point of kiwifruit park is 1.665 pixel,and the error of the contour corner feature matching method at the inspection point of vineyard is 2.837 pixel,which achieves the purpose of extracting canopy temperature by registration and fusion of infrared thermal image and visible image.(3)In view of the limited field of vision obtained by the lens and the problem of blind field of vision,the camera parameters optimized by improved particle swarm optimization algorithm are used to calibrate the camera.By building a video stream server environment,the test and analysis of video push and pull flow are carried out under the condition of school LAN,and the transmission effect of video stream is verified.The results show that the method of this study uses the Web interface to combine seven cameras to obtain a large horizon through push and pull flow.In order to serve the users timely,the image analysis system of agricultural inspection robot is designed and developed.The system realizes the function of large view interface display,collecting images at inspection points and matching fusion,extracting canopy temperature by temperature detection algorithm and displaying.