论文部分内容阅读
服务监控对象人体存在检测是室内移动机器人应用中定位、识别与跟踪人的基础,但室内环境的复杂多变性与视觉系统的移动给人体检测带来很大的困难,使得人体检测结果不稳定且有效性差,为此,本文提出一种基于室内移动机器人视觉系统的人体存在检测方法.首先采用多尺度小波变换检测法与边缘连接算子相结合的方法提取图片边缘特征,并提出一种形态学方法去除非目标小区域、不封闭的边缘线或孤立点,利用边缘图片的不变Hu 矩作为模式识别特征向量.然后应用自适应高斯核函数软间隔支持向量机建立两类识别分类器,并与基于不同特征建立分类器的人体存在检测法和基于不同分类方法建立分类器的人体存在检测法进行分析比较,结果表明本文算法是更稳定有效的.
Detection of Human Body Existence of Service Monitoring Objects is the basis for locating, identifying and tracking people in indoor mobile robot applications. However, the complex and variability of the indoor environment and the movement of the visual system bring great difficulties to human body detection and make the human body test results unstable This paper presents a human body presence detection method based on indoor mobile robot vision system.First, the edge feature of the image is extracted by combining the multi-scale wavelet transform detection method with the edge-connected operator, and a morphological The method removes the non-target small area, the non-closed edge line or the isolated point, and uses the invariant Hu moments of the edge picture as the pattern recognition feature vector. Then two types of recognition classifiers are established by using adaptive Gaussian kernel soft interval SVM Compared with human existence detection method based on different features to establish classifier and human existence detection method based on different classification methods, the results show that the proposed algorithm is more stable and effective.