摘要
针对机器人视觉系统对抓取物的模糊识别问题,参照人眼-脑识别对象的过程,建立了包括区域分割与模糊识别两个环节的识别算法.以视觉图像的灰度与色度特征作为区域分割与模糊识别的依据,从灰度、色度及形体特征上提取特征集的指标,并根据经walsh变换后图像灰度迅速向低频聚集的特点,提出基于walsh变换的基元模式识别特征的定义方法.在构建识别目标矩阵与关系矩阵的基础上,应用模糊关系合成与最大隶属度原则建立识别算法.该算法可从少量的采样点中识别出对象,具有较好的实时性.
With reference to the object recognition process of human eyes and minds, the region segmentation and fuzzy recognition algorithms were established for robot vision systems in order to deal with the problem of the object-capture recognition. The gray-scale and chromaticity of images were taken as a basis for region segmentation and fuzzy recognition, the index of the characteristics were taken from gray-scale, chromaticity and body features, and the definition method for feature recognition on the basis of walsh transform metamodels was put forward with the sharp low-frequency gathering tendency of the gray-scale. Then, the recognition algorithms was worked out utilizing the fuzzy relation composition and maximum subjection according to the object recognition matrix and relation matrix was established. With the algorithms, the objects within small samplings spots can be recognized, which has practical values in the object-recognition in the capture control of robots.
出处
《北京航空航天大学学报》
EI
CAS
CSCD
北大核心
2009年第2期197-200,共4页
Journal of Beijing University of Aeronautics and Astronautics
关键词
视觉
图像分析
模式识别
WALSH变换
内积
vision
image analysis
object recognition
walsh transform
inner product
