基于机器视觉感知的交互式虚拟装配研究

Research on Interactive Virtual Assembly Based on Machine Vision Perception

融合机器视觉感知,研究了交互式虚拟装配技术。首先建立虚拟装配平台,构建零件与虚拟手仿真模型,根据行为推理规则实现交互;然后基于颜色空间分割手部图像,通过SVM分类器识别手势命令,并根据三角测量原理计算手部位置信息;最后应用该系统进行视觉系统验证以及交互式虚拟装配试验。经试验可知,所构建SVM分类器的分类成功率为92.45%,F1值为92.41%;算法耗时分布于(65~80)ms;零件装配成功率为90%。结果表明系统可以快速准确地跟踪手部,完成交互式虚拟装配。

Combining machine vision perception,the interactive virtual assembly technology was studied.Firstly,the virtual as⁃sembly platform was built.The part and virtual hand simulation models were built,and the interaction was realized according to the behavior inference rules.Then,the hand image was segmented by the color space,the hand command was recognized by the SVM(Support Vector Machine)classifier,and the position information of the hand was calculated by triangulation.Finally,the system was applied to visual system verification and interactive virtual assembly test.According to the test,the classification success rate of the SVM classifier was 92.45%,and F1 score was 92.41%.And the algorithm took time to distribute between(65~80)ms.And the interactive virtual assembly of parts had a success rate of 90%.The results show that the system can quickly and accurately track the hand and realize the interactive virtual assembly.