论文部分内容阅读
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。
In order to overcome the shortcomings of real-time, accuracy and robustness of target recognition of autonomous soccer robot, this paper presents a threshold vector bit and region merging algorithm based on the dynamic window HSI color space model. Through dynamic Window to reduce the search range to speed up the segmentation speed, the average running time of the measured target recognition about 26ms; In addition, according to the target moment invariant and Kalman filtering method to improve the tracking accuracy and robustness of the target, the experimental results show that, Target tracking accuracy of about 99.3%, so that the entire system has high real-time and very good recognition.