Research On Target Recognition And Picking Point Determination Technology Of Green Pepper In Greenhouse

China is the country with the highest yield of green peppers and has a vast area of planting.Green pepper is a labor-intensive industry.With the increase of labor cost,it is necessary to research and develop agricultural machinery and equipment to promote the healthy and sustainable development of green pepper industry.In order to improve the efficiency of green pepper picking,it is an urgent problem to realize the automation of green pepper picking.Due to the problem of similar colors of green peppers and leaves,it is an important issue for green pepper picking robots to accurately recognize green pepper targets.On this basis,it is no doubt that the key to improve the success of picking is to determine the position of the picking point.This paper combines the current research hotspots of using machine vision technology in agriculture,and takes the green pepper image as the research target.Through key technologies such as image enhancement,segmentation,and saliency detection based on manifold ranking,the recognition of green pepper targets is achieved,and the research on the determination of picking points is completed using the physical characteristics of green peppers.Then the effectiveness of the proposed method is verified by experiments.The main contents of this paper are as follows:(1)In terms of green pepper image acquisition and preprocessing research,the MX808 camera configured on the green pepper picking robot,which is used to collect green pepper images under different lighting conditions in greenhouse.Then,the green pepper image was analyzed for color features,and the near-color system characteristics were displayed in the form of color component maps.At the same time,the image enhancement method based on fuzzy set theory is used to process the green pepper image,so that the overall intensity of the green pepper image is more uniform.The image enhancement effect is achieved under the premise of retaining the characteristics of the green pepper,ensuring the integrity of the extraction of the green pepper target.(2)According to the characteristics that the target of green pepper is similar to the leaf color,the SEEDS superpixel segmentation algorithm is proposed to segment the green pepper image.The SEEDS algorithm divides the image into several super-pixel blocks while retaining the basic features of the target.Through the effective pixel exchange between each super-pixel block,iterative calculation is gradually performed to reduce the super-pixel block,and the edge of the super-pixel is modified and refined.Finally the internal characteristics of the image are consistent,so as to achieve fast and accurate segmentation of green pepper target image.(3)Aiming at the difficulty of green pepper target recognition,this paper proposes an improved algorithm based on the fusion feature of closed-loop graph manifold ranking salient target detection.This method uses the saliency detection of background features to obtain green pepper targets in the image,and uses the saliency detection of foreground features to obtain green pepper targets at the edge of the image.By fusing foreground and background features,the saliency detection image after fusion obtains the best performance.The experiment results of green pepper target recognition show that this method has good recognition effect.The recognition rate of green pepper reaches 85.6%,which improves the recognition rate of green fruits in near color background,and meets the needs of picking robot for picking green pepper fruits.(4)In view of the fact that the green pepper fruit stalk and the plant stem are very similar in color and shape,it is difficult to identify the green pepper fruit stalk only by image processing.A method of modeling green pepper is proposed to determine the picking point of the fruit stalk.Through the analysis of the physical characteristics of green pepper,the area of interest of green pepper is studied after obtaining the basic parameters of green pepper.On the basis of obtaining the outline of green pepper,the coordinates of the intersection point of the stalk and fruit are obtained by the SUSAN corner detection algorithm.The picking point of green pepper fruit stem is obtained by the correlation between fruit stem and intersection.