Rearch On Machine Vision Algorithm Of Grape Recognition For Robotic Bagging

The technology of grape bagging could protect grapes against the pollution of pesticide and being injured by external environment and improved the appearance quality and yield of the grape, which was an important link in grape production process. But the manual bagging operation was so tedious, labor-intensive and low-efficiency, that limited the development of the technology of grape bagging. So the research on grape bagging robot had great significance to the improvement of grape automatic and intelligent operation equipment.In order to resolve the problem, the image of the simple grapevine taken in the pergola trellis was taken as the study object. A grape recognition algorithm combined with color, appearance and texture features of grape image was presented in this paper in order to get the center line and the length of the grape, which lay the foundation of realize the automation of grape bagging in the future. The main research contests and results in thesis were as follows:(1) Combined with the growth characteristics of the grapes in south China, a new cultivation system suitable for automatic bagging was provided. The structural improvements of existing fruit bag were improved to make it more suitable for robot and a overall structure design program of the grape bagging robots was provided.(2) Through the analysis of color features of grape image, the|G-R|+|G-B|chromatic aberration image of the grape was most available to the segment the eggplants. The method of threshold segmentation was used to the|G-R|+|G-B|chromatic aberration image of the grape. Due to the color of grapes was similar to its stem and leaf, the effect of the threshold segmentation was not satisfactory.(3) The sobel operation and thinning operation were used to get the edge of the|G-R|+|G-B|chromatic aberration image. The mathematical model of the berry round was established based on the shape characteristics of the grape. Berries of the grape were detected by Hough transform combined with their edge gradient information. Limited by interference from the complex background and the accuracy of Hough circle detection, parts of the circles detected located in non-grape regions which need to be removed through further judgment.(4) The circles detected by the Hough transform were judged by the feature of the color and texture of grape berries and the concentrated berries. In order to get the center line and the length of the grape, the circles located outside of the grape cluster were removed and the rest circles were used to extract of the grape region. (5) The algorithm of grape recognition was verified through120grape images we collected. According to the evaluation of the recognition algorithm proposed in this paper, the recognition rate can achieve about90%and the time it cost is about3.8s.