Research on the Location of Citrus Based on RGB-D Binocular Camera

• Main Authors: Hu, Xiaomei, Wu, Jun, Yu, Hefang, Lv, Shunke, Wang, Chuan
• Format: Conference Proceeding
• Language: English
• Subjects: backproject; Calibration; Cameras; citrus recognition; depth image repair; hand-eye calibration; Lighting; Maintenance engineering; Manipulators; Robot kinematics; Robot vision systems
• ISSN: 2689-6621
• DOI: 10.1109/IAEAC50856.2021.9390798

Using robots instead of picking fruits by hand is a trend of scientific and technological development.Firstly,the picking robot needs to solve the problem of identifying and locating the fruit, which determines the final success rate of picking. According to the growth characteristics of citrus, a citrus recognition method based on RGB-D binocular camera was proposed in this paper. Backproject and template matching were used to obtain the citrus position area. Then, the influence of noise, illumination and shade of branches and leaves was eliminated by morphological operation and convex hull operation.Finally,the coordinates of 2-D center point were obtained by elliptic fitting and moment calculation. In order to obtain the effective depth information of citrus,a depth image cavity repair algorithm based on improved bilateral filtering was proposed to calculate the more accurate 3-D coordinates of citrus picking points. Finally, the 3-D picking point coordinates in camera coordinate system are converted into manipulator coordinate system by the 9-point calibration method. The identification and picking success rate of the proposed method is more than 90% in laboratory environment.