Research On Intelligent Mobile Fruit Picking Robot

Fruit picking robot has very important significance to reduce labour intensity, to ensure harvesting in time and to pretect fruit quality. It also can increase automation levels in orchard operations. Funded by "863Project-study on key technologies for mobile fruit picking robot", key technologies for mobile fruit picking robot were studied by means of integrating with mechanics&electronics, image processing, data communication, computer networks, intelligent control system etc. The main contents and results of the research are summarized as below:1. By the survey of the standard dwarf orchard, the basic design parameters of the picking robot were determined. Compared with the chassis of the four-wheeler electric vehicle and crawler, it was fund that the existing chassis are hard to meet the design requirements. Consequently, a light tracked intelligent mobile platform was designed independently which was found suitable for orchards operation. The platform is integrated with the picking arm and its control system, end-effecter along with its control system, the drive system of the mobile platform, DGPS and binocular camera etc. With industrial PC as the main controller, the entire picking robot motion control system was built and the prototype was independently developed.2. The vision system of the fruit picking robot was built and the image processing software was developed to recognize and locate the fruits. Under the color space of RGB, normalized rgb, HIS, Lab, YIQ and Ⅰ1Ⅰ2Ⅰ3, numerous pictures of apple fruits, leaves and branches were analyzed. To find out the aberration model whose images of fruits, leaves and branches are obviously different, gray images of a variety of aberration model were extracted. The real time character and the effect of actual segmentation of processing of the gray images extracted from the aberration model were analyzed. The result showed that the ideal segmentation effect of the grays images of the R-G aberration model can be acquired with taking adaptive threshold segmentation algorithm. Eventually the method which combined circular Hough transform with mark of the center was adopted to detect the coordinate of the fruit central point. Compared with utilizing the circular Hough transform Alone, the algorithm proved that recognition rate of fruit can be increased by8.5%.3. The collection of the navigation parameters based on machine vision and DGPS were completed. The software and interactive interface of the intelligent mobile navigation system was developed. While collecting the vision navigation parameters, linear Hough transform and least square method were combined to detect the navigation path. Research proved that the method was able to combine the close point set of the straight line detected by Hough transform to fit the navigation path. Compared with utilizing the Hough transform alone, The algorithm proved that the false detection rate was reduced by30%.4. The kinematics analysis and the resolution of picking arm and intelligent mobile platform were carried out, the optimized arithmetic of the picking arm based on "the shortest travel" and "the least energy consume" are raised. Besides, the trajectory planning of the picking arm and intelligent mobile platform were researched in this paper as well.5. The remote video monitoring system of fruit picking robot is built and the functions of this system are as follows:Firstly, it can inspect both situations of the picking robots’ work states and the nearby environment; Secondly, for the purpose of controlling each module’s action, the remote control command are sent to the picking robot via the remote monitoring client by the inspector when the robot meets the gusty situation; Thirdly, the picking robot can inspect the nearby environment by itself, once the robot approached by humans or obstacles, the alarm will be sent to both the site and the remote monitoring client by the main controller of the picking robot.6. Through the experiments in campus and integrated tests in the orchard, the operation performance of the picking robot was validated.The results showed that the mobile platform’s navigation, picking arm’s movement, end-effecter’s grasp and fruit automatic encasement can operate coordinately,and all the related algorithms satisfied real-time and reliability.