一种眼睛状态检测的组合方法  

Combined method of eye states detection

作　　者：刘志波[1] 周石琳 张官亮[1] 闫军伟 

机构地区：[1]国防科学技术大学电子科学与工程学院信息工程系,湖南长沙410073 [2]武警石家庄士官学校通信二系有线通信教研室,河北石家庄050061

出　　处：《现代电子技术》2013年第8期101-103,108,共4页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(60972114)

摘　　要：眼睛状态检测具有广泛的应用。为克服目前单个算法的不足,进一步提高眼睛状态检测的准确率,在此首先介绍了2种眼睛状态检查方法,即模板匹配方法和投影方法,然后对这2种方法进行组合提出了一种新的组合方法。组合方法首先利用模版匹配方法对眼睛状态做出判断,然后通过与预先设定的阈值比较,对第一步无法做出判断的图像再利用投影法进一步判断。通过对300个眼睛图像进行检测,准确率达到了94.8%。试验结果表明了该组合方法较单个检测方法具有更好的可靠性和准确性,同时适中的复杂性也说明该算法的可行性。There are many applications of the robust eye states detection. In order to overcome the shortcomings of the sin- gle algorithm and improve the accuracy of the eye states detection, a combined method of eye states detection is proposed on the basis of two methods： template matching method and projection method. The template matching method is used to identify the eye states, and then before outputting the final result, the result is compared with an pre-set threshold. Some images which can not identify in the first step would have the second detection for eye states by the use of projection method. Experiments are done with 300 different human eye images. The accuracy rate of eye states detection is 94.8%. The experiment result shows that this method is more reliable and accurate, and is practical because it is moderate in both time and space consuming.

关 键 词：眼睛状态检测 模板匹配 水平投影 组合方法 

分 类 号：TN964.34[电子电信—信号与信息处理]