基于深度学习的智能采摘机器人

Intelligent Picking-robot based on Deep Learning

我国是果蔬生产大国,果蔬的采摘耗时耗力,研究开发自动识别、采摘、收集果蔬的智能采收机器人,可以减轻农业从业者的劳动强度,提高果蔬采摘行业的自动化水平。视觉系统是实现果蔬智能采摘的关键技术,基于深度学习的识别模型有利于提高果蔬的检测和定位精度。主要运用YOLOv3网络训练了采摘识别模型,该模型可以识别场地中的十字标,将十字标中心三维坐标值反馈到无人车控制板,对无人车位置进行了二次校准;该模型还用于果蔬品种和成熟度的识别,并获取成熟果蔬中心点的三维坐标值,将成熟果蔬坐标值带入逆运动学求解得出合适的运动轨迹反馈给机械臂,实现机械臂与相机的“手眼协同”。最后,设计开发好的机器人在实验场地上进行果蔬抓取,验证了智能采摘机器人的可行性。

China is a big producer of fruits and vegetables.The picking of fruits and vegetables is time-consuming and labor-intensive.Research and development of intelligent harvesting robots that automatically identify,harvest,and collect fruits and vegetables can reduce the labor intensity of agricultural practitioners and improve the automation level of the fruit and vegetable picking industry.The visual system is the key technology to realize the intelligent picking of fruits and vegetables.The recognition model based on deep learning is helpful to improve the detection and positioning accuracy of fruits and vegetables.YOLOv3 network is used to train the picking recognition model.The model can identify the cross marks in the field,feed the three-dimensional coordinate value of the cross-mark center back to the unmanned vehicle control board,and calibrate the position of the unmanned vehicle twice.The model is also used to identify the variety and maturity of fruits and vegetables,and obtain the three-dimensional coordinate value of the center point of mature fruits and vegetables.The coordinate value of mature fruits and vegetables is brought into the inverse kinematics solution to obtain the appropriate motion trajectory feedback to the manipulator to realize the'hand-eye coordination'between the manipulator and the camera.Finally,the designed and developed robot is used to grab fruits and vegetables on the experimental site,which verifies the feasibility of the intelligent picking-robot.