研究目的
To explore the use of an ensemble classifier approach based on Random Subspace Method (RSM) for palmvein recognition, comparing results with and without hand-crafted features to improve performance under varying environmental conditions.
研究成果
The use of RSM improves palmvein recognition performance, especially when applied to raw images, reducing the need for hand-crafted feature optimization. Results are promising for contactless sensors, suggesting potential for realistic implementations with less user cooperation. Future work will extend to other descriptors and hybrid techniques.
研究不足
The study does not specify technical constraints or optimization areas, but mentions that the effectiveness of hand-crafted features is unclear in large populations or variable environments, and parameter optimization for descriptors like BSIF is not fully explored.
1：Experimental Design and Method Selection:
The study uses an ensemble classifier based on Random Subspace Method (RSM) with feature reduction via 2DPCA and 2DLDA. Two approaches are investigated: using hand-crafted features (LBP and BSIF) and using raw images without feature extraction.
2：Sample Selection and Data Sources:
Two public datasets are used: PolyU MS Database (contact sensor, 250 users, 6000 images) and CASIA Palmprint Image Database (contactless sensor, 100 users, 7200 images). Images are multispectral, with ROIs cropped and preprocessed.
3：List of Experimental Equipment and Materials:
No specific equipment or materials are mentioned in the paper; it focuses on computational methods and datasets.
4：Experimental Procedures and Operational Workflow:
ROIs are extracted, histogram equalization is applied, images are divided into blocks, and features (LBP or BSIF) are computed. RSM generates random subspaces from reduced feature spaces, and nearest neighbor classifiers are used for identification with majority voting.
5：Data Analysis Methods:
Performance is evaluated using Correct Classification Rate (CCR or Accuracy), with experiments varying subspace sizes and number of classifiers.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
