温室串番茄采摘机器人关键技术研究

In order to improve the automation level of truss tomato harvesting, a robot system for truss tomato harvesting is developed, and two key technologies of picking point location and end-effector design were mainly studied. Firstly, a fruit recognition algorithm based on color difference R-G was presented, and the centroid and boundary of fruit region was extracted. According to the growth characteristics of the stem of tomato truss, an extraction algorithm for picking-point based on color difference G-B and "equidistant block location method" was presented, and then, the 3-D information of picking-point was calculated by sterovision system. Secondly, the hardware structure of an end-effector based on flexible transmission were designed, and the function of cutting and grasping could be realized simultaneously by this end-effector. At last, the task sequence of harvest operation was planned and a harvesting experiment in greenhouse was conducted. Experiment results show that, the success rate for harvesting truss tomato is 86.67%, and the execution time for picking a truss tomato successfully was 37.57s.