Target Recognition Of Greenhouse Harvesting Robot Based On Binocular Vision

Machine vision is a method for intelligent robots to perceive their surroundings.The position information of target object is an important information which can be obtained by the application of Machine vision.A target location model based on machine vision is proposed in this paper.In this paper,the object recognition and localization are researched by means of machine vision and image.Pomegranate is one of the fruits widely planted in our country.In the production of pomegranate,33-50% of the labor force is used for harvesting,which has a strong repeatability and a huge demand for labor.If intelligent harvesting robots can be introduced into the pomegranate production process to achieve mechanized and By harvesting,the limited natural resources can be released and the productivity can be improved..This is not only of great practical significance,But it is also a necessary option for the modernization of agriculture.Identifying and locating the fruit is the main task of harvesting robot and designing.Its accuracy determines the working effect of the harvesting robot.Therefore,the research on the object identification and location technique of Pomegranate Harvesting robot is very important.In this paper,based on binocular vision technology,the object recognition and positioning technology of pomegranate harvesting robot is studied to achieve effective recognition and three-dimensional space positioning of pomegranate fruit,which provides necessary information for pomegranate harvesting robot.The main research work and results are as follows: Then,A binocular stereo vision system is built,and its inner and outer parameters are obtained through calibration.(1)The left and right cameras are stereo calibrated,and the rotation matrix and translation vectors describing the spatial position relationship between the two cameras are obtained.(2)The method of pomegranate image segmentation is studied,and several different methods of pomegranate image segmentation are compared.The method of segmentation based on OTSU threshold and HSV color space is selected as the experimental image of pomegranate in this study.(3)The recognition method of pomegranate image is studied,single target and overlapping citrus target are identified,And a new method of an overlapping target recognition based on the color space transformation theory of concave shell is presented,which achieves the positioning of pomegranate harvest center point and the recognition of pomegranate target.(4)The three-dimensional spatial positioning method for the target of the pomegranate image is studied.First,The contour edge of the target pomegranate is obtained by morphological method,and the central coordinate of the target object is obtained by using the minimal outer rectangle.In this paper,the 3D coordinates of the target fruit are recovered by3 D reconstruction.