基于颜色和区域的运动目标识别

针对全自主足球机器人目标识别受光强变化的影响,实时性、准确性和鲁棒性存在的不足,提出了一种基于动态窗口HSI色彩空间模型的阈值向量位与及区域合并算法,并通过动态窗口减小搜索范围加快分割速度,实测目标识别的平均运行时间约26ms;另外,根据目标的矩不变及卡尔曼滤波的方法,提高了跟踪目标的准确性和鲁棒性,实验结果表明,目标跟踪准确率约99.3%,使整套系统具有很高的实时性和很好的识别效果。

The moving target identification based on color and region

In view of drawbacks of autonomous robot soccer in aspects of real-time, precision and robustness due to influence of light intensity in the process of target recognition, a kind of algorithm based on regional merger and threshold vector bit-and by dynamic window HSI color space model is presented. Dynamic window is used to narrow scope of search and accelerate speed of division with average operation time of approximate 26ms for recognition of actual detected target. Furthermore, moment constant of target and Kalman Filtering Method is used to improve precision and robustness of traced target. The results show that precision in tracing of target is approximate 99.3% which endow whole set of system with high real-time performance and excellent effect of recognition.