【作者】 刘杰； 秦永强； 史元春；

【Author】 Jie Liu, Yongqiang Qin,Yuanchun Shi Key Laboratory of Pervasive Computing Dept. of Computer Science, Tsinghua University, Beijing, China

【机构】 清华大学计算机科学技术系；

【摘要】 本文为基于背部红外照明方式的交互桌面上的触点识别技术提出了改进算法。首先针对红外光照射不均匀的问题,采用动态化参数方法以提高桌面边缘触点的检测效果;然后采用双阈值方法消除噪声影响,同时也提高了触点识别的准确度和速度。文章最后与当前常用的多点触摸软件Community Core Vision(CCV)所用触点识别算法进行了对比,结果表明本文所提出的算法无论是在处理速度、准确度还是在抑制噪声能力上,较CCV的算法都有很大的改进和提高。更多还原

【Abstract】 This paper studies the current blob recognition technique used in diffused illumination interactive tabletops. In order to splice several tabletops into a large scale tabletop, we use Fresnel lens to enhance the display effect, which also brings in large bright areas where blob cannot be recognized. So cameras can only be place at the side in the tabletop and this leads to the non-uniform spread of infrared light in the image. The existing methods cannot address this tough issue. We did many researches and came up with several algorithms to tackle it and at the same time improve its accuracy and speed. First, we use ROIs to deal with the problems caused by the non-uniform spread of infrared light. But we find that the more ROI are used, the less frame rate we get. The ROI method is not practical. Then we adopt an adaptive parameter method, which has significantly improved the recognition result, especially around the edge of the tabletop. Second, we use hysteresis threshold in the binaryzation process, which not only eliminates background noise,but also efficient in both accuracy and speed.We have made a comparison between our algorithm and Community Core Vision(CCV),a popular multi-touch software.The results show that ours outperforms CCV in speed,accuracy and suppression of noise.Our algorithms can successfully recognize98.8%of the total blobs,raises80%frame rate and66%recognition rate.更多还原