基于主成分分析的目标确认方法及其在车标定位识别中的应用.pdfVIP

33 1 ( ) V o l 33 N o 1 2009 2 Journa l of N anj ing U n ive rsity o f Sc ience and T echno logy ( N a tura l Sc ience) F eb 2009 1, 2 1 王 枚 , 王国宏 ( 1 , 264001; 2 , 264 670) : 提出一种基于主成分分析的目标确认方法, 解决小目标定位错误率高, 并由此导致 效 目标识别的问题将已建好的目标模板看成 一组随机向量, 利用主成分分析得到 一组特征目 标; 从原图像中检测可能目标, 并将其映入特征目标空间进行重构; 构造原目标与重构目标的似 真度函数, 根据该函数值可对检测目标进行确认或剔除, 降低误定位率, 确保了进入后续识别的 目标为目标库中对象将该方法应用在实测车辆图像车标定位识别测试, 结果表明: 与不使用 似真度函数验证相比, 目标定位准确度提高了 16 5% ; 使用不变矩最小距离分类器进行车标识 别, 识别准确度比不使用似真度函数确认提高了 20% : 主成分分析; 似真度函数; 目标确认; 车标定位识别 : TP 391 41 : 1005- 9830( 2009) 01- 0042 - 05 ObjectConfirmationMethod Based on PrincipleComponent Analysis and ItsApplication in Vehiclelogo Location and Recognition 1, 2 1 WANG M e i , W ANG Guohong ( 1 D ep artm ent o f E lectron ic and In form at ion Eng ineering, N aval A eronautica l Engineering In stitute, Y antai 264001, Ch ina; 2 Laboratory o f Im age Processing, Y anta i V ocationa l Co llege, Y antai 264670, China) Abstract: A n object conf irm at ion m ethod b ased on pr incip le com ponen t ana lysis( PCA ) is presented to mi prove the accuracy of sm a ll object location and recogn ition T he ob ject sam ples are regarded as a set of random vectors A set of eigenob jects is g iven by PCA The ob ject recon struct ion is execu ted by re flect ing the located actu al object to the e igensp ace The p lausibility function is defined be tw een orig inal located object and reconstru ction object and the function is used to confirm the true ob ject or to de lete the false object in o rder to redu ce the fa