基于PCA和边缘不变矩的车标识别新方法  

New method of vehicle-logo recognition base on PCA and Edge invariant moment

作　　者：王枚[1] 王国宏[1] 高小林[2] 王勇[2] 

机构地区：[1]海军航空工程学院电子通信系,山东烟台264001 [2]烟台职业学院计算机与信息工程系,山东烟台264000

出　　处：《计算机工程与应用》2008年第4期224-226,229,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60541001);全国优秀博士学位论文作者专项基金项目(No.200443)

摘　　要：车标识别技术是进行车辆类型识别的关键内容。在车牌定位准确的基础上,利用车头图像中车牌、车标位置关系定位到车标大致区域;接着对车标进行精确定位、归一化处理,并提出似真度函数对车标进行确认;然后利用边缘不变矩的最小距离进行车标识别。该方法不受原始车标图像大小、变形等影响,能够自适应地进行目标定位确认和识别,具有较强的鲁棒性。通过实测图像测试,结果表明该方法是有效和可行的。This paper presents a method of vehicle-logo location and recognition based on PCA and edge invariant moment.First, the approximate range of the vehicle-logo region can be located depending on the topology structure between vehicle plate and vehlcle-logo.Then the vehicle-logo precise location,reglstration and detection are executed in order.The plausibility function is represented to ensure the real vehicle-logo.At last the vehicle-logo recognition is achieved by edge invariant moment minimum distance,The approach can adapt to complete object location and recognition with high robust.The experimental result on actual vehicle images shows that the method is practical and effective.

关 键 词：似真度函数 边缘不变矩 车标配准 车标识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]