Yongmei Cheng

Fast algorithm and application of Hough transform in iris segmentation

Hough transform is a common algorithm used to detect geometry shape of objects in computer image processing, some points satisfied lines, circles and other curves can be detected easily by using Hough transform in image. In iris recognition system, both the inner boundary and the outer boundary of a typical iris can approximately be taken as circles, so the two circles of iris can be obtained by using Hough transform. In this paper, Hough transform algorithm is introduced and is adopted in iris segmentation. A modified fast algorithm is presented to solving low speed of Hough transform. Simulations and research results show that Hough transforms are satisfied in iris segmentation.

Analyzing Human Movements from Silhouettes via Fourier Descriptor

Human movement analysis has recently gained growing interest for computer vision researchers. In this paper, a simple but efficient algorithm using Fourier descriptor to represent spatial-temporal silhouette for human movement analysis is proposed. For each image sequence, motion detection and segmentation methods are used to segment and extract the moving silhouettes of people. Then, Fourier descriptor(FD)is used to describe the moving silhouettes. HMMs and Haudorff distance are applied to the time-varying distance signals described by FD for the activity classification and gait recognition. Experimental results have demonstrated that the proposed algorithm greatly improve the recognition rates.

Multi-Camera-Based Object Handoff Using Decision-Level Fusion

Object handoff is the key technique of the multi-camera system. Since the feature extraction is not accurate and not integrated in the object handoff algorithm based on feature fusion, we propose an algorithm based on decision-level fusion. This algorithm can be described as: firstly, every feature of the object is defined as an evidence, whose basic belief assignment (BBA) is computed; secondly, the Dempster combination rule is used to fuse multiple evidence to get the final BBA; lastly, we can use decision rule to determine whether the current object is the same one with that in the identification framework, and also give the current object a unique identifier. After the three steps, the object handoff in the multi-camera system is finished. The simulation results show that this object handoff algorithm can achieve better performance.

Particle Filter Based Visual Tracking Using New Observation Model

A new tracker based on particle filter is proposed in this paper. In our framework, colour cue and edge cue, which are represented as colour histogram (CH) and improved histogram of oriented gradient (IHOG) respectively, are adaptively fused to represent the target observation. Colour histogram is robust to shape variation and rotation etc, but sensitive to varying illumination and easy to be confused by distractions from background due to loss of spatial information; whereas for IHOG, the situation is reversed. With the help of the complementary nature of the two kinds of image features, the proposed tracker is more robust to pose variations, illumination changes and distractions from background. As the second contribution of this paper, an improved model update scheme is proposed to address the varying appearance. The new scheme makes our object model has a better resistance to template drift. Experimental results demonstrate the high robustness and effectiveness of our method in complex environments.

Face detection using SVM trained in independent space

The classical face representation method, such as eigenface, extracts covariance based on low-order statistics feature of image. However, high-order information represents image details, which are necessary for pattern recognition. Hence, PCA is first used to reduce its dimension; then the independent component analysis (ICA) is applied to further obtain independent feature vector instead of low-order statistics; finally support vector machine is used as a classifier that has demonstrated high generalization capabilities for face detection. The feasibility and correctness of this new face detection method are shown in CBCL Face Dataset.

A Multipath Viterbi Data Association Algorithm for OTUR

Target tracking of sky wave OTHR inevitably faces the problem of multiple propagation modes, through modeling target movement in ground coordinate and implementing data association in radar coordinate, a multipath viterbi data association (MVDA) algorithm is proposed, which extends viterbi data association (VDA) from association between measurement and track to association among measurement, propagation mode and track. The simulation results show that MVDA is superior to multipath probability data association (MPDA )

Image recognition based on invariant moment in the projection space

This paper proposes a projection-based invariant moment for image recognition. A set of features invariant to image translation and scaling are obtained in the 1-D projection space. For getting rotational invariance, rapid transform is employed. After obtaining the invariant feature vector, threshold analysis is used for feature data optimization, and principle component analysis (PCA) is applied for feature data length compression. Experimental results show the superiority of our method over Hu and other invariant moments.

A kernel-based bayesian classifier for fault detection and classification

A kernel constructed by Shannon sampling function was utilized for kernel Fisher discriminant analysis (KFDA). And kernel-based Bayesian decision function was implemented for fault detection. Simultaneously, Bhattacharyya distance was introduced as a criterion function for separability comparison. The proposed Shannon KFDA was compared with Gaussian KFDA on Tennessee Eastman Process (TEP) data. The results show that Shannon KFDA has lager Bhattacharyya distance and detects more faults more quickly than Gaussian KFDA.

Estimation of Markov Jump systems with mode observation one-step lagged to state measurement

The estimation of Markov jump systems (MJS) is widely used in target tracking, fault detection, signal processing and digital communications. However, the above researches all assume that state measurement and additional mode observation are synchronous which means both state measurement and mode observation at each sampling time arrive at the fusion centre at the same time. The problem of estimation of MJS that mode observation is one-step lagged to its corresponding state measurement is considered. Along state-augmentation approach and the derivation of image-enhanced interacting multiple model (IE-IMM), a new generic estimation algorithm is proposed. It is shown by simulation result that the proposed algorithm is effective.

A new detection algorithm of multiple traffic parameters under complex background

Real-time and precise detection of the multiple traffic parameters was an important problem in the traffic surveillance and management. By analyzing the frame to frame difference of gray value in a certain part of the image, this algorithm could detect those traffic parameters such as the vehicle speed, the number of the vehicle, the density of the vehicle and so on, under complex conditions without background update. Also, this algorithm was able to classify the moving shadow and the vehicle automatically, which could do great help to the traffic management. Compared to other methods of this problem, the experiment results show that this method is a real-time multiple traffic parameters detection algorithm, and its performance are satisfied.