基于机器视觉的圣女果采摘机械臂的动作研究

Research on the Movement of Cherry Tomato Picking Robotic Arm Based on Machine Vision

摘要:

针对当前圣女果采摘机器人无法保证带蒂采摘的问题，提出一种通过机器视觉进行圣女果姿态分析进而生成特定的机械臂采摘动作的方法；该方法通过模拟人手采摘流程，能够使得机械臂末端执行器到达采摘位置时与圣女果的果蒂方向保持一致；整体系统包括对成熟圣女果的目标检测，测距算法的实现，圣女果方向识别以及机械臂动作生成；根据轮廓拟合算法的思想进行算法改进，实现针对圣女果的更加精确、稳定的方向识别算法，从而获得机械臂末端执行器与圣女果果蒂方向一致的目标位姿，进而实现相应机械臂采摘动作的生成；多次实验表明，改进后的对于圣女果方向的识别算法相较于传统轮廓拟合算法而言误差角度更小，对于不同姿态圣女果的方向识别更具稳定性，因此更加适用于实际采摘流程中根据圣女果姿态生成机械臂的特定采摘动作。

Abstract:

Aiming at the problem that the current cherry tomato picking robot cannot guarantee the picking with pedicle, a method of analyzing the posture of cherry tomatoes through machine vision and then generating a specific picking action of the robotic arm is proposed; by simulating the manual picking process, this method can make the end effector of the robotic arm reach the picking position in line with the direction of the fruit pedicle of cherry tomatoes; the overall system includes the target detection of ripe cherry tomatoes, the realization of the distance measurement algorithm, the recognition of the direction of the cherry tomatoes, and the generation of robotic arm movements; according to the idea of contour fitting algorithm, the algorithm is improved to achieve a more accurate and stable direction recognition algorithm for cherry tomatoes, so as to obtain the target pose of the end effector of the robotic arm in the same direction as the saint fruit, and then realize the corresponding generation of robotic arm picking motions; multiple experiments have shown that the improved recognition algorithm for the direction of cherry tomatoes has a smaller error angle than the traditional contour fitting algorithm, and it is more stable for the direction recognition of cherry tomatoes in different postures, therefore, it is more suitable for the specific picking action of the robotic arm generated according to the posture of cherry tomatoes in the actual picking process.