Abstract: Aiming at the hand feature detection and recognition problem for the augmented assembly guide process,an improved algorithm based on Continuously adaptive mean-Shift (CamShift) was proposed.In this algorithm,the hand was traced and detected in the assembly environment,the search box was obtained by iterative computing the center distance,and complete information of hand dynamic optimization detection area to get.To get the features of hand,by using hand outline with image enhancement processed,the feature points with equidistance was sampled,the curvatures for different position of feature points were calculated,and a group fingertip points was selected after clustering analysis.A prototype system was developed by using 3D registration of augmented assembly as an example to verify the effectiveness,feasibility and better robustness of proposed method.

Key words: hand feature, tracking detection, augmented assembly, guides

摘要： 针对增强装配引导过程中手部识别检测的问题,提出一种改进自适应均值漂移的特征检测算法。该算法对装配环境中的人手进行跟踪检测,迭代计算中心距获取搜索框,动态优化检测区域,得到包含手指信息的手部完整信息。为得到手指特征点,利用图像增强处理过后的手部轮廓进行等距特征点采样,对不同部位的特征点计算曲率信息,聚类分析得到指尖特征点。以增强装配系统中的三维注册为例开发原型系统,验证了所提方法具有可行性、有效性和较好的鲁棒性。

关键词: 手部特征, 跟踪检测, 增强装配, 引导