摘要
针对NAO机器人识别目标准确率过低的问题,为降低光照对识别的影响,提出一种基于HSV颜色空间的轮廓信息特征识别的算法,通过融合颜色特征和轮廓特征识别图像中的目标。利用HSV空间模型,通过颜色阈值分割对图像进行预处理,提取红绿色目标;根据目标规则的多边形轮廓,对其形状信息加以约束;利用二值图像的轮廓特征矩加以判决,得到识别目标及其在图像中的中心坐标,实现目标的精确识别。利用NAO机器人采集图像进行模拟实验,改变NAO与目标的相对位置并多次测量,成功定位的准确率可达到92.67%。实验结果表明,NAO机器人采用该算法可以快速稳定地实现目标识别,提高了准确率。
Focusing on the issue of the low accuracy of target recognition of NAO robot, and to reduce the impacts of light on target recognition, an algorithm of contour feature recognition based on HSV color space was proposed. The target was identified successfully in the image through combining the color and contour features. Color threshold segmentation based on HSV color space was adopted to achieve image preprocessing, to extract the red and green targets. The shape information of the target was restrained according to regular polygonal contour. Contour moment feature of the binary image was used to judge and determine the contour of target, and the target and its center coordinates in the image were got, thereby accurate target recognition was achieved. In the simulation experiments based on the images collected from NAO robot, many experimental data were measured through changing the relative positions of the NAO and the target. The successful positioning accuracy of the proposed algorithm can reach 92.67%. Experimental results show that, by the aid of the proposed method, NAO robot can realize the target recognition rapidly and stably with higher accuracv.
作者
梁付新
刘洪彬
张福雷
常发亮
LIANG Fu-xin LIU Hong-bin ZHANG Fu-lei CHANG Fa-liang(School of Control Science and Engineering, Shandong University, Jinan 250061, China)
出处
《计算机工程与设计》
北大核心
2017年第8期2235-2239,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61673244
61273277)
关键词
NAO机器人
目标识别
颜色识别
轮廓特征矩
图像预处理
NAO robot
target recognition
color recognition
contour feature moment
image preprocessing
