虚拟现实飞行模拟训练中的视觉交互技术研究

Research of Visual Interaction for Virtual Reality Flight Training

为实现半虚拟现实座舱环境中的人机自然交互,使用基于表观的视觉手姿态估计方法。针对基于表观方法在大容量高维度手势特征检索过程中所存在的时间效率低、内存消耗大等问题,提出了引入多探测原理和近邻特征表的改进局部敏感哈希(LSH)索引,并提出了索引性能预测模型和基于性能预测模型的索引参数优化方法,提高了索引方法的检索性能。实验结果表明,预测模型能反映实际的索引性能,使用参数优化后的改进LSH索引进行10近邻特征检索,可保证索引召回率基本不变,而使在线实际总耗时减少41.9%。将改进LSH索引应用于视觉手姿态估计,可实现虚拟手可视化,再现用户真实手的各种动作和状态。

Appearance-based hand pose estimation which relies on computer vision techniques is adopted to realize natural interaction in a semi-virtual reality cockpit. To cope with the low efficiency and high memory consumption in large capacity and high-dimension feature indexing, an improved locality sensitive Hashing （LSH） method is proposed in this paper which combines the multi-probe principle with the nearest-neighbor table. Moreover, a forecast model which predicts indexing per- formance and a parameter optimization method are used to achieve better indexing performance. Experimental results show that the forecast model is appropriate for practical indexing performances and the time consumption is reduced by 41.9 % at the cost of a slight recall rate drop. In summary, the application of the improved LSH to hand pose estimations able to up- grade virtual hand visualization and hand posture reconstruction in a semi-virtual reality cockpit.