一种基于高频率投影仪—摄像头的人机交互系统

发布时间：2018-07-12 20:49

本文选题：人机交互 + 摄像头—投影仪系统　；参考：《计算机应用研究》2017年03期

【摘要】：针对桌面投影设备,提出了一种基于单投影仪—单摄像头的人机交互系统架构,可以实现在投影面上的类触摸屏操作。提出了一种基于高速投影中互补阴影的方法,在不影响使用者视觉体验的前提下实现触碰点位置检测与触碰判断。针对低强度差分阴影的提取,提出了基于马尔可夫随机场模型的最优化方法,提高了复杂环境噪声下阴影分割的鲁棒性。针对指尖检测,使用了一种基于最小二乘法的曲率估计方法,提高了触碰定位的准确度;触碰判断方面,提出了一种基于双目约束与几何模型的高精度触碰判断准则。实验证明,系统达到触碰识别率99.2%,误识别率0.6%,像素级的触碰点判断准确率1.77像素,超过了同类研究的最先进水平,能够在PC机上达到60 fps的运行速度,这满足了绝大部分实时应用的需求。
[Abstract]:For desktop projection equipment, a human-computer interaction system architecture based on single projector and single camera is proposed, which can realize touch screen operation on projection plane. A method based on complementary shadow in high speed projection is proposed to detect and judge the contact position without affecting the user's visual experience. An optimization method based on Markov random field model is proposed for extracting low-intensity differential shadows, which improves the robustness of shadow segmentation under complex noise. For fingertip detection, a curvature estimation method based on least square method is used to improve the accuracy of touch location, and a high precision contact judgment criterion based on binocular constraint and geometric model is proposed. The experimental results show that the system achieves the contact recognition rate of 99.2, the error recognition rate of 0.6 and the accuracy rate of 1.77 pixels at the pixel level, which exceeds the most advanced level of the similar research, and can run at 60 fps on the PC. This meets the needs of most real-time applications.
【作者单位】：清华大学深圳研究生院;清华大学微电子所;
【基金】：国家自然科学基金资助项目(61373073) 深圳市科技计划资助项目(JCYJ20150331151358146) 清华大学自主科研基金资助项目(20131089223)
【分类号】：TP391.41;TP11

【相似文献】