摘要
矿井时有安全事故发生,签到管理系统可及时、准确掌握人员出入人员状况,保障矿井安全生产,方便及时救援。针对传统签到管理系统用于矿井,遇到光线昏暗、人脸易附着粉尘、干扰噪音等因素影响,签到识别方法检测率低,提出了一种根据KL变换(Karhunen-Loeve Transform)和TAN分类(Tree-Augmented Naive Bayesian network)相结合的人脸识别,并辅以声音识别的方法。通过形态学滤波变换快速去掉大部分无用背景,使处理更快速,特征点更突出;自动根据具体环境选择图像识别或声音识别,使识别准确率更高。仿真结果表明:结合声音的系统识别方法既减小了计算复杂度,又提高了人员识别率,还增强了适应性。
Coalmine accidents happen sometimes coalmine or outside, which is convenient for rescue. It is significant to know the accurate statement of the miners in When the traditional Sign-in Management System was used in coal mine, the system meets new problems, such as black, hazy face, etc. Aiming at this issue, this paper put for- ward a face recognition algorithm based on the combination of Karhunen-Loeve Transform and Tree-Augmented Naive Bayesian network classifier, which uses the morphological filtering to remove most of useless transform background quickly. In addition, the voice recognition method was addede to that algorithm which makes feature point more out- standing and identification more accuracy, according to the specific environment automatic selection of face recogni- tion or voice recognition. The simulation shows that this algorithm not only reduces the computational complexity and improves the human face recognition rate, but also enhances the adaptability.
出处
《计算机仿真》
CSCD
北大核心
2012年第11期272-275,共4页
Computer Simulation
基金
"211工程"三期创新人才培养计划建设项目(S-09108)
关键词
签到管理系统
形态学滤波
人脸识别
声音识别
Sign-in management system
Morphological filtering
Face recognition
Voice recognition
