基于图像矩与神经网络的机器人四自由度视觉伺服

针对传统的视觉伺服方法中图像几何特征的标记、提取与匹配过程复杂且通用性差等问题,本文提出了一种基于图像矩的机器人四自由度(4DOF)视觉伺服方法.首先建立了眼在手系统中图像矩与机器人位姿之间的非线性增量变换关系,为利用图像矩进行机器人视觉伺服控制提供了理论基础,然后在未对摄像机与手眼关系进行标定的情况下,利用反向传播(BP)神经网络的非线性映射特性设计了基于图像矩的机器人视觉伺服控制方案,最后用训练好的神经刚络进行了视觉伺服跟踪控制.实验结果表明基于本文算法可实现0.5 mm的位置与0.5°的姿态跟踪精度,验证了算法的的有效性与较好的伺服性能.

Visual servoing of 4DOF using image moments and neural network

To avoid the complicated marking, extracting and matching of image features in the traditional visual servoing systems and to improve the universality of the algorithm, a novel visual servoing of 4-degrees of freedom(4DOF) is proposed for an eye-in-hand robot based on image moments and neural network. First, the nonlinear transform relationship between image moments and the robot pose is developed, which provides the theoretical basis for the visual servoing usingimage moments. Then, a back propagation(BP) neural network is designed to map the transformation from image moments variation to the robot pose displacement with 4DOF without the external and internal parameters calibration for the camera. After this, the proposed control scheme can be applied to the robotic visual servoing. The experiment results show that the tracking error is less than 0.5 mm and 0.5°respectively in position and in orientation. This confirms the validity and satisfactory servoing performance of the proposed method.