基于统计特征的人脸模拟画像检索

Simulacrum face sketch retrieval based on statistical features

目的以刑事侦查和公安网上追逃的迫切需要为背景,配合金盾工程的开展,对人脸模拟画像的检索算法作了初步研究。方法首先,根据网上追逃数据库中的人脸照片设计出一种半自动的归一化算法。其次,研究了人脸模拟画像和人脸照片的边缘检测方法,给出一种自适应的人脸照片边缘检测算法。最后,得到3种用于人脸模拟画像检索的统计特征。结果给出人脸模拟画像检索的具体算法,给出了实验结果,并进行了分析和评价。结论为构建实用的人脸模拟画像检索系统的开发提供参考。对金盾工程的完善可望起到良好的促进作用。

Aim To meet the urgent needs of the police pursuit on net,and promote the implementation of the Golden Shield Project and initiate a preliminary study on the simulacrum face sketch retrieval.Methods Firstly a semiautomatic normalized algorithm of the face image is designed based on the face photoes in the face image database in China Criminal Center.Secondly a self adaptive method is presented to carry on marqinal check of the face photo and the simulacrum face sketch.Thirdly three statistical features that can be used in the simulacrum face sketch retrieval are obtained.Results The simulacrum face sketch retrieval algorithm is designed and the results in our experiments are analyzed and evaluated.And an estimate of the retrieval algorithm is given.Conclusion The study will provide the theoretical foundation of building the retrieval system,and greatly promote the consummation of the Golden Shield Project.