基于MobileNet和文本识别匹配的证件图片分类算法  

License Image Classification Algorithm Based on MobileNet and Text Recognition Matching

作　　者：曾鹏 李曦[1,3] 赵璐 杜彦辉 ZENG Peng;LI Xi;ZHAO Lu;DU Yanhui(School of Information and Cyber Security,People's Public Security University of China,Beijing 100038,China;Xiamen Public Security Bureau,Xiamen 361008,China;Informatization Corps,Sichuan Provincial Public Security Department Science and Technology,Chengdu 610031,China;Xiangyang Railway Public Security Department,Xiangyang 441002,China)

机构地区：[1]中国人民公安大学信息网络安全学院,北京100038 [2]厦门市公安局,福建厦门361008 [3]四川省公安厅科技信息化总队,四川成都610031 [4]武汉铁路公安局襄阳公安处,湖北襄阳441002

出　　处：《中国人民公安大学学报（自然科学版）》2023年第3期52-58,共7页Journal of People’s Public Security University of China(Science and Technology)

摘　　要：利用计算机视觉技术进行图片分类是公安部门打击违法犯罪的重要手段。由于存在不同种类图片特征相似的问题,单纯基于视觉特征的证件类图片分类效果有待进一步提高,因此提出一种融合MobileNet和文本识别匹配的证件图片分类算法,用于证件类图片的分类任务中。该方法把图片分类及文本匹配算法相结合,提高了证件类图片分类的准确性。在证件类图片数据集上的实验结果表明,算法在损失一定召回率的情况下,准确率有显著提升。Utilizing computer vision techniques for image classification plays a vital role in law enforcement agencies'efforts to combat illegal activities.However,the effectiveness of solely relying on visual features for license image classification is limited due to the problem of similar characteristics among different types of images.To address this issue,a novel algorithm that combines MobileNet with text recognition matching is proposed for license image classification tasks.This approach integrates image classification and text matching algorithms,thereby enhancing the accuracy of license image classification.Experimental results on a dataset of license images demonstrate that the accuracy rate of algorithm is significantly improved with the loss of certain recall rates.

关 键 词：MobileNet 文本匹配 证件图片 图片分类 

分 类 号：D918.92[政治法律—法学]