机构地区: 广东工业大学机电工程学院
出 处: 《微型机与应用》 2012年第22期38-41,43,共5页
摘 要: 针对当前流行的汽车行人避碰系统中行人轮廓检测效果不够理想的情况,改进行了人轮廓的检测。利用数学形态学操作方法先对视频图像进行处理,然后利用优选出的Canny算子对行人图像进行边缘检测,再使用曲线拟合技术对行人轮廓进行加工,获取完整的行人轮廓曲线,为行人识别与跟踪奠定坚实的基础。试验结果表明,该方法能有效地去除行人轮廓的干扰因子和孤岛,获得较为完整的行人轮廓曲线。 At present, the results of pedestrian edge detection are less ideal using the popular pedestrian edge detection methods. This paper focuses on the improving of pedestrian profile detection. At first, it uses grayscale morphological to process video image. Then, picking up Canny operator for pedestrian edge detection. Finally, it process the pedestrian profile by using crave line fitting technology to obtain smooth pedestrians outline curve for the better foundation of pedestrian recognition and tracking. The experiment results show that the method can he used to eliminate the interference factors and isolated island, and can obtain complete pedestri-an contour curves.
关 键 词: 算子 闭运算 曲线拟合 干扰因子 行人轮廓检测
领 域: [自动化与计算机技术] [自动化与计算机技术]