一种用于机器人的物体运动参数快速识别方法
A Fast Parameter Estimation Algorithm of Moving Objects form Sequential Images for Robot Tracking
-
摘要: 为了使机器人能跟踪并抓取运动目标,实时给出目标物的运动参数是首要问题,也是个困 难的问题.本文给出一种物体二维运动的快速估计算法,不需要抽取物体特征,也不需要事先 知道物体的模型,而是通过运动物体序列图像的复数矩与运动参数之间的关系,来恢复物体的 二维运动.该算法与基于傅里叶描述子的运动估计方法进行了比较,证明了算法的快速性和 准确性.Abstract: It is essential to estimate in real-time the parameters of motion of moving objects in visually guided robot tracking and grasping. While feasibility of the task has been demonstrated in laboratories by using for instance optical flow method at the expense of special hardware implementation, dedicated and efficient algorithms are still considered essential and studied as an open research problem. This paper presents a fast parameter estimation algorithm of moving objects in 2-D space, based on our finding of the relationship between the desired parameters of motion and a complex transformation on the images. This algorithm requires no model of the objects and no feature extraction. Comparison of the performances of this algorithm with those of FFT will be given in the paper.
-
Key words:
- robotics /
- visual tracking /
- image processing
计量
- 文章访问数: 2265
- HTML全文浏览量: 100
- PDF下载量: 1003
- 被引次数: 0