Research On Multi-arm Strawberry Harvesting Robot Based On Dynamic Recognition And Localizationg

Strawberry is a popular fruit in our country and is highly demanded.The cultivated area has increased year by year which was 113.3 thousand hm2 in 2014.However,strawberry harvesting is labor intensive which takes a long time to bow down.It is necessary to use a robot to replace human labor.Since ripe strawberry fruit is red and significantly different from background and separates from stem and leaves,it is highly feasible to harvest strawberry with robot.Aiming at the growth and distribution feature of the ridge planting strawberry,this research proposes the multi-arm continuous harvesting method.The mechanical structure design,Dynamic Positioning Identification,picking task assignments is finished.And finally develop the prototype.The main research contents and results are as follows:1.A collecting fruits end-effector was designed to nondestructively and efficiently harveste fruits.It gathers the fruit down to up to avoid the obstacle and then clamp and cut off the stem.The mechanical structure is comprised of gather-cutting mechanism and ridge contour following mechanism.The test shows that it is can be harvested successfully when the location error is in the±7mm and the width of the fruits' area is in the 10 cm.2.Automatic threshold segmentation based on local histogram was applied on strawberry images to solve low robustness of global threshold segmentation.A "mini-batch k-means" based algorithm which optimizes the initial clustering center according to statistics of the peaks,and the histograms were used to further improve the efficiency of the algorithm.A method of roughly separating contacted fruits was proposed,which firstly detects sag-regions as well as their bump as the underside-concave-points of mutual-contacted strawberries,and calculates the general direction and rough edges of the overlap edges by the contour of strawberry near the concave point.3.Contour matching method was applied to stereo vision to reduce complexity;the position of walking mechanism was recorded with high speed triggered by the exposure signal of the binocular camera to get its position in the world coordinate without time lagging.The visual servoing with eye-in-hand the system was applied to locate the position of fruits in horizontal direction.During the localization process,the position of the servo motor encoder was recorded with high speed when the motion controller was triggered by the exposure signal of camera to solve the problem that the position at which the camera capturing image inconsistencies with the image caused by the lag of visual feedback.Real-time position information from servo motor encoder was taken to compensate the low frequency of visual feedback.And the depth of target fruits was obtained on the basis of structure from motion.4.The centralized architecture control system with multiple arms picking robot was designed.It can meet the requirement of controlling the robot system,which was built by a main controller,a walking system controller,a vision system controller and other robot controllers.A dynamic task allocation algorithm for multi-robots was proposed,which can be applied in the situations with a changing picking task and an uncertain single picking time.5.A prototype of multi-arm strawberry harvesting robot was built,and the indoor picking experiments were conducted.The robot consists of a binocular camera,three Groups 2 DOF picking robot,a walking mechanism and a control system.The experiment results show that the robot can recognize the dynamic positioning and multi-objects picking when the strawberries distributed over the surface with a density of 10 fruits per meter along the ridge,the fruits could be picked completetly at the robot traveling speed of no higher than 100 mm/s,the average harvesting speed could be 1.7 s for each picking.