针对我国载人航天地面虚拟组装训练需求，研究了基于Kinect的视觉手势识别技术，提出了一种使用深度信息进行手势检测及分割的方法，采用手势图像轮廓曲率分析方法进行指尖检测，并结合Kalman滤波稳定跟踪指尖；同时自定义了几种组装所需的手势，根据图像的指尖特征进行特定手势识别。结合OpenNI、Nite和OpenCV，最终在VC2010环境下实现了航天虚拟组装训练系统。实验结果证明，该手势识别方法识别率较高，鲁棒性好，可应用于载人航天虚拟组装训练。%Aiming at the requirements of virtual assembly training for manned space missions of Chi -na, the gesture recognition technology based on Kinect was researched .A gesture detection and seg-mentation scheme based on depth data was proposed .Fingertips were detected through analyzing the curvature of the gesture image contour .Combined with Kalman filtering , fingertips were stably tracked .Then a set of required assembly gestures were defined .The gestures could be recognized with characters of the fingertips .With development tools of OpenNI , Nite, OpenCV, virtual assem-bly training system was implemented in Microsoft Visual Studio 2010.Experimental results showed that there are higher recognition rate and good robustness in this gesture recognition method .It may be used effectively in virtual assembly training for manned space missions .
展开▼
