摘要
人脸是十分常见的复杂模式，在复杂景物图片中自动找出人脸是很困难的，但是却有着重要意义，因为这同时也是自动人脸识别系统的前提．提出了一个基于多层感知器（ＭＬＰ）的用遗传算法实现搜索的人脸定位系统，讨论了系统建立的理论基础，即ＭＬＰ直接感知图象和用遗传算法进行快速搜索．实验结果表明，它尤其适合在复杂背景下的低分辨图象中进行操作，鲁棒性好、定位快、适应面宽．
Human face is a very common complex pattern, it is difficult but quite important to detect faces in a complex background. In fact, face detection is also a prerequisite for face recognition. An approach to this issue is presented, which is based on multi layer perceptron and genetic algorithm. The experimental results show that this system is good especially at low resolution and complex ima ges and is robust, quick and widely applicable. However, its applications are limited as it is required that faces in an image must be of approximate sizes. The paper introduces this system and gives a theoretical analysis about the mechanism of this approach, that is, multi layer perceptron for sensing faces directly on gray data and genetic algorithm for searching faces in the scene.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1998年第9期93-96,共4页
Journal of Shanghai Jiaotong University
关键词
神经网络
遗传算法
人脸定位
MLP
人脸识别
neural network
genetic algorithm
face location
multi layer perceptron (MLP)
quick sensing
