摘要
煤矿监控图像的对比度低、灰度不均匀,使监控图像的处理和识别难度增大。在巷道环境下采集多角度的矿工监控图像,根据脸部肤色区域像素的统计特征,用上下限阈值的方法在HSV颜色空间分割出各种角度的矿工脸部,用半径为2的圆形结构元素进行数学形态学的开启和闭合操作,去除被误分为脸部区域的背景像素;根据脸部灰度分布特征构造了平均脸模板,用相似度函数作为脸部模板匹配的判别函数,检测矿工脸部位置。结果表明,用肤色分割和平均脸部模板匹配方法,可以快速检测出矿工脸部的精确位置。
The miner surveillance images are of low contrast and unevenness gray level, which make the progress of processing and recognition very hard. Miner surveillance images from many angles are sampled in laneway conditions. The miners' face areas of all angles can be segmented in HSV color space with up and bottom threshold value according to statistical characteristic of face color. The pixels which are mis-segmented can be wipped off with the open and close op- eration in morphological with the circular structure element of 2 pixels radius. An average face template based on the distribution feature of face gray level was constructed. A similarity functions was used as the discriminant functions to determine the locations of faces. The results show that the locations of the miners' faces can be pinpointed fleetly with the method of face color segmentation and template matching.
出处
《计算机科学》
CSCD
北大核心
2010年第8期280-282,共3页
Computer Science
基金
中国矿业大学科技基金资助项目(OD090228)
江苏省基础研究计划(自然科学基金)(No.BK2009093)
国家"863"高技术项目(No.2006AA01Z128)
中国科学院智能信息处理重点实验室开放基金(No.IIP2006-2)资助
关键词
矿工监控图像
颜色模型转换
开启
闭合
平均脸模板
模板匹配
Miner surveillance image, Color model transform, Open, Close, Verage face template, Template matching