Anni Cai

Fingerprint matching using ridges

Traditionally, fingerprint matching is minutia-based, which establishes the minutiae correspondences between two fingerprints. In this paper, a novel fingerprint matching algorithm is presented, which establishes both the ridge correspondences and the minutia correspondences between two fingerprints. First N initial substructure (including a minutia and adjacent ridges) pairs are found by a novel alignment method. Based on each of these substructure pairs, ridge matching is performed by incrementally matching ridges and minutiae, and then a matching score is computed. The maximum one of the N scores is used as the final matching score of two fingerprints. Preliminary results on FVC2002 databases show that ridge matching approach performs comparably with the minutia-based one.

Comparing evaluation protocols on the KTH dataset

Human action recognition has become a hot research topic, and a lot of algorithms have been proposed. Most of researchers evaluated their performances on the KTH dataset, but there is no unified standard how to evaluate algorithms on this dataset. Different researchers have employed different test setups, so the comparison is not accurate, fair or complete. In order to know how much difference there is when different experimental setups are used, we take our own spatio-temporal MoSIFT feature as an example to assess its performance on the KTH dataset using different test scenarios and different partitioning of the data. In all experiments, support vector machine (SVM) with a chi-square kernel is adopted. First, we evaluate performance changes resulting from differing vocabulary sizes of the codebook, and then decide on a suitable vocabulary size of codebook. Then, we train the models using different training dataset partitions, and test the performances one the corresponding held-out test sets. Experiments show that the best performance of MoSIFT can reach 96.33% on the KTH dataset. When different n-fold cross-validation methods are used, there can be up to 10.67% difference in the result. And when different dataset segmentations are used (such as KTH1 and KTH2), the difference in results can be up to 5.8% absolute. In addition, the performance changes dramatically when different scenarios are used in the training and test dataset. When training on KTH1 S1+S2+S3+S4 and testing on KTH1 S1 and S3 scenarios, the performance can reach 97.33% and 89.33% respectively. This paper shows how different test configurations can skew results, even on standard data set. The recommendation is to use a simple leave-one-out as the most easily replicable clear-cut partitioning.

Fingerprint Indexing Using Ridge Invariants

Indexing large fingerprint databases is an important and challenging problem. In this paper, an invariant-based fingerprint indexing scheme is proposed. Minutia and surrounding ridges are combined to form a substructure. The invariants describe binary relations between substructures. Experimental results on FVC2002 database demonstrate the validity of the proposed algorithm

Enhanced and hierarchical structure algorithm for data imbalance problem in semantic extraction under massive video dataset

Data imbalance problem often exists in our real life dataset, especial for massive video dataset, however, the balanced data distribution and the same misclassification cost are assumed in traditional machine learning algorithms, thus, it will be difficult for them to accurately describe the true data distribution, and resulting in misclassification. In this paper, the data imbalance problem in semantic extraction under massive video dataset is exploited, and enhanced and hierarchical structure (called EHS) algorithm is proposed. In proposed algorithm, data sampling, filtering and model training are considered and integrated together compactly via hierarchical structure algorithm, thus, the performance of model can be improved step by step, and is robust and stability with the change of features and datasets. Experiments on TRECVID2010 Semantic Indexing demonstrate that our proposed algorithm has much more powerful performance than that of traditional machine learning algorithms, and keeps stable and robust when different kinds of features are employed. Extended experiments on TRECVID2010 Surveillance Event Detection also prove that our EHS algorithm is efficient and effective, and reaches top performance in four of seven events.

A reliable method for paper currency recognition based on LBP

Paper currency recognition with good accuracy and high processing speed has great importance for banking system. How to extract high quality monetary features from currency images is a key problem in paper currency recognition. Based on the traditional local binary pattern (LBP) method, an improved LBP algorithm, called block-LBP algorithm, is proposed in this paper for characteristic extraction. The proposed method has advantages of simplicity and high speed. The experimental results show that this improved method has a high recognition rate, as well as robustness for noise and illumination change.

Fingerprint Matching With Rotation-Descriptor Texture Features

A novel texture correlation matching method for fingerprint verification using Fourier-Mellin descriptor and phase-only correlation function is proposed in this paper. Fourier-Mellin descriptor correlation is used to align the template and query fingerprint images and a matching score is obtained. Matching takes about 1 second in Celeron 2.0 GHz processor, and the experimental results show that EER is 3.8%; fusion with minutia matching gets a better result

Shot boundary detection algorithm in compressed domain based on adaboost and fuzzy theory

A shot boundary detection algorithm based on fuzzy theory and Adaboost is proposed in this paper. According to changes of color and camera motion, videos are classified into six types. By using features in compress domain such as DCT coefficients, the type of the MB, HSV color histogram difference, camera motion difference and so on, videos are segmented into three classes, that is, cut shot, gradual shot and non-change. The results of experiment have shown that this algorithm is robust for camera motion and walk-in of large objects in videos, and have better precision of shot boundary detection compared with the classic double-threshold method and the method of presented by Kuoet al.. There is no problem of threshold selection in our algorithm but it exists in most of other algorithms.

Combining multiple SVM classifiers for adult image recognition

Pornographic image recognition and filtering are of great significance for web security and content monitoring. In this paper, an adult image recognition method based on support vector machine (SVM) and erotic category is proposed. Global color and texture features and local SIFT feature are extracted to train multiple SVM classifiers for different erotic classes. Face detection is used to filter out normal close-up images. Four later fusion schemes are presented to determine the final result. A large scale test on 50,000 web images shows the proposed algorithm achieves 12.32% false positive rate(/p) and 14.17% false negatives rate(/h), which is better than five existing methods.

Fingerprint Representation and Matching in Ridge Coordinate System

Fingerprints are generally represented and analyzed in Cartesian or polar coordinates. In this paper, however, fingerprints are represented and analyzed in a novel coordinate system, called ridge coordinate system (RCS), which is based on a ridge and an oriented point on the ridge. Using RCSs based on lots of ridges, we obtain a robust representation scheme of the ridge (skeleton) image of a fingerprint. Based on this representation, an alignment-based fingerprint matching algorithm is proposed. Experimental results on FVC2002 demonstrate the validity of the proposed algorithm

Datum-Adaptive Local Metric Learning for Person Re-identification

Person re-identification (PRID) is a challenging problem in multi-camera surveillance systems. In this paper, we propose a novel Datum-Adaptive Local Metric learning method for PRID, which learns individual local feature projection for each image sample according to the current data distribution and projects all samples into a common discriminative space for similarity measure. We adopt an approximate strategy based on Local Coordinate Coding to learn local projections. Anchor points are first generated by clustering and the local projection of each sample is then approximated by the linear combination of a set of projection bases, which are associated with the anchor points. Experimental results demonstrate that the proposed approach obtains superior performance compared with state-of-the-art methods on public benchmarks.