基于子空间与纹理特征融合的掌纹识别 
下载： 1070次
Palmprint Recognition Based on Subspace and Texture Feature Fusion
1 辽宁工程技术大学电子与信息工程学院, 辽宁 葫芦岛 125105
2 辽宁工程技术大学研究生学院, 辽宁 葫芦岛 125105
图 & 表
图 1. LOBP编码原理
Fig. 1. Coding principle of LOBP
下载图片 查看原文
图 2. LOBP提取掌纹纹理特征的流程图
Fig. 2. Flow chart for extracting palmprint texture features by LOBP
下载图片 查看原文
图 3. 掌纹识别流程
Fig. 3. Flow chart of palmprint recognition
下载图片 查看原文
图 4. 图库示例。(a) PolyU图库;(b)自建非接触图库
Fig. 4. Examples of databases. (a) PolyU database; (b) self-built non-contact database
下载图片 查看原文
图 5. PolyU图库上的实验结果。(a)匹配结果;(b) ROC
Fig. 5. Experimental results on PolyU database. (a) Matching results; (b) ROC
下载图片 查看原文
图 6. 自建非接触图库上的实验结果。(a)匹配结果;(b) ROC
Fig. 6. Experimental results on self-built non-contact database. (a) Matching results; (b) ROC
下载图片 查看原文
图 7. 所提方法与最新方法的EER比较
Fig. 7. Comparison of EER between proposed method and latest methods
下载图片 查看原文
图 8. 所提方法与最新方法的识别率比较
Fig. 8. Comparison of recognition rate between proposed method and latest methods
下载图片 查看原文
表 1RLDA算法总结
Table1. Summary of RLDA algorithm
| Input: data matrix X, parameter λ1 |
|---|
| Initialization:A=0; E=0; Y=0; α=0.1; ρ=1.01;P=argtr s.t. PTP=I;αmax=105;λ=10-4 | | while not converged do1. Update A by using Eq. (17);2. Update P by using Eq. (19);3. Update E by using Eq. (21);4. Update Y, α by using Eqs. (22) and (23), respectively | | end while | | Output: P,A,E |
|
查看原文
表 2单一特征与融合特征识别的EER
Table2. EER for single feature and fusion feature recognitions%
| Database | Subspacefeature | Texturefeature | Fusionfeature |
|---|
| PolyU | 1.5082 | 0.3853 | 0.3440 | | Self-builtnon-contact | 2.5176 | 1.5168 | 1.4922 |
|
查看原文
表 3所提方法与其他方法的EER和识别时间比较
Table3. Comparison of EER and recognition time between proposed and other methods
| Method | PolyUdatabase | Self-built non-contactdatabase |
|---|
| EER /% | Recognitiontime /s | EER /% | Recognitiontime /s |
|---|
| PCA | 3.5064 | 0.3165 | 5.6432 | 0.3091 | | 2DGabor | 2.5242 | 0.8189 | 3.1340 | 0.9136 | | LDA | 3.2340 | 0.2674 | 4.3759 | 0.3010 | | LBP | 3.1754 | 0.2424 | 4.2591 | 0.2911 | | RLDA | 1.5082 | 0.2229 | 2.5176 | 0.2294 | | LOBP | 0.3853 | 0.1628 | 1.5168 | 0.1801 | | LBP+2DLPP[9] | 0.5077 | 0.4844 | 1.6326 | 0.5794 | | GGF[10] | 0.4761 | 0.3505 | 1.6073 | 0.3968 | | WACS-LBP+WSRC[11] | 0.4085 | 0.3239 | 1.5384 | 0.3479 | | ProposedRLDA+LOBP | 0.3440 | 0.3069 | 1.4922 | 0.3127 |
|
查看原文
李新春, 马红艳, 林森. 基于子空间与纹理特征融合的掌纹识别[J]. 激光与光电子学进展, 2019, 56(7): 071007. Xinchun Li, Hongyan Ma, Sen Lin. Palmprint Recognition Based on Subspace and Texture Feature Fusion[J]. Laser & Optoelectronics Progress, 2019, 56(7): 071007.
PDF全文
