In order to improve orchard automation, an identification and location system of apple picking robot was developed. In order to overcome dependence of existing system on the working distance, human picking process using a web camera and a Kinect sensor was imitated. Compared with manual measurements after calibration, the standard deviation of binocular vision in three directions were 3. 4 cm, 1. 2 cm and 1. 2 cm, respectively. An apple detection and location method was developed based on color, shape and position. The results showed that when the working distance was about 240 cm, the mature apples can still be accurately identified and located, the standard deviation of binocular vision in depth direction was 4. 9 cm. When the working distance was about 150 cm, the standard deviation of binocular vision in depth direction was 2.4 cm. When the working distance was less than 100 cm, the standard deviation in depth direction was less than 1. 0 cm with web camera.%为实现机器人自动化采摘作业,设计了苹果作业机器人识别与定位系统.模拟人类采摘过程,采用单、双目视觉组合传感器系统,克服了现有识别与定位系统对目标到视觉传感器距离的依赖性.实现了对单、双目视觉系统的单独和组合标定,与手动测量结果相比,双目视觉系统标定后3个方向的标准偏差分别为3.4、1.2和1.2 cm.提出了基于苹果颜色、形状和位置特征的识别与定位方法,试验结果表明:当工作距离为240 cm时,双目视觉系统可以准确识别并定位所有苹果,深度方向标准差为4.9 em;当工作距离为150 cm时,双目视觉系统深度方向标准差为2.4 cm;当工作距离小于100 cm时,单目视觉传感器测量目标到传感器距离的标准偏差为1.0 cm.
展开▼
