柑橘果园机器人仿生眼的目标跟踪与定位系统

针对传统柑橘园林机器人云台视野范围小、定位精度低等问题,在仿生眼平台上,提出一种基于连续自适应均值偏移(CAM shift)算法和变结构PID控制器相结合的目标跟踪方法。具体包括效仿人类眼颈的结构特点,设计一个4自由度的仿生眼视觉平台；建立仿生眼的成像模型,提出一种相机外参数自适应标定方法,以解决仿生眼平台外参数会随着相机位姿改变而不断变化的问题；提出基于CAM shift的仿生眼目标跟踪算法,成功实现了对于目标的持续跟踪；为提高仿生眼系统在饱和非线性特性下的动态响应性能,提出采用基于抗饱和(Anti-windup)的变结构自适应PID控制器来设计仿生眼控制系统。以柑橘为研究对象,在设计的仿生眼平台上开展试验。结果表明,本文方法左眼平均中心位置像素误差(CPE)为7.2,右眼CPE6.1,均小于设计所要求的CPE10,且每帧图像处理过程小于0.1s。系统能满足实际应用,具有较高的鲁棒性和准确性。

Target Tracking and Positioning System for Citrus Orchard Robot Bionic Eye

Aiming at the problems of small field of vision and low positioning accuracy of traditional citrus garden robot head, a target tracking method based on continuous adaptive mean shift (CAM shift) algorithm and variable structure PID controller is proposed on the bionic eye platform. Specifically, a 4-dof bionic eye vision platform is designed based on the structural characteristics of the human eye neck. The imaging model of the bionic eye was established, and an adaptive calibration method for external parameters of the camera was proposed to solve the problem that the external parameters of the bionic eye platform would change with the change of camera pose. The algorithm based on CAM shift is proposed to realize the continuous tracking of the target. In order to improve the dynamic response performance of the bionic eye system under saturation nonlinear characteristics, a variable structure adaptive PID controller based on anti-windup was proposed to design the bionic eye control system. With citrus as the research object, experiments were carried out on the bionic eye platform. The results show that the average center position pixel error (CPE) of the left eye is 7.2, and that of the right eye is 6.1, both of which are less than CPE10 required by the design, and the processing process of each frame is less than 0.1s. The system can meet the practical application and has high robustness and accuracy.