基于图像特征的运动目标识别与伺服跟踪

本文研究了一种在室内相对复杂背景下机器人的运动目标识别与跟踪方法.基于Intel的计算机视觉库OpenCV进行图像处理与特征提取,采用图像的区域特征--图像矩作为特征信息,在有干扰的情况下,实现运动目标的形状识别和基于特征的视觉伺服跟踪控制.为提高系统性能,建立曲线拟合模型来预测运动目标的动态趋势和特征.实验结果表明,系统在不需要摄像机精确标定、背景有干扰的情况下,能够实现对运动目标进行实时稳定快速的识别和跟踪.

Recognition and servo tracking of moving target based on image features

In this paper, the research on the recognition and servo tracking of moving target for a wheeled moving robot with a monocular camera is carried out in a relatively complex background. At first the computer vision library from Intel, Open CV, was employed to implement image processing and feature extracting, then the shape of moving target was recognized and the visual servo tracking was achieved based on local features of image moments in noise condition. In order to improve the system performance, a curve-fitting model was built to predict the dynamic trend and characteristics. Experiment results indicate that the system can stably recognize a fast motional target and track it in real-time under noisy background and without precise camera calibration.