论文部分内容阅读
针对发展型机器人自主学习过程中特征提取涉及的增量计算和实时性问题,结合已有的CCIPCA(直观无协方差增量主成分分析)和BDPCA(双向主成分分析)算法,提出了一种增量的BDPCA算法.采用了迭代的计算方式,具备增量的计算能力,并且将2维原始图像矩阵直接作为处理对象,有效地降低了计算量,缩短了程序运行时间.以机械臂待抓取的物块作为实验样本,利用支持向量机进行分类,验证上述算法.实验结果证明了该算法的有效性,平均分类率可达90%,算法处理速度大约26帧/秒,基本满足了发展型机器人的实时处理需求.
Aiming at the incremental computation and real-time issues involved in the feature extraction of autonomous learning of developmental robots, a new algorithm based on CCIPCA (direct covariance non-covariance incremental principal component analysis) and BDPCA (bidirectional principal component analysis) Incremental BDPCA algorithm uses an iterative calculation method with incremental computing power, and takes the 2-D original image matrix directly as the processing object, which effectively reduces the computation and shortens the running time of the program. The results show that the proposed algorithm is efficient and the average classification rate is up to 90%. The processing speed of the algorithm is about 26 frames / second, which basically meets the requirements of development Real-time processing needs of robots.