Search

Surveillance on pedestrian flows in crowded areas is of significance for various security tasks. This problem involves two parts: evaluation of crowdedness and detection of abnormality, where a lot of motion information needs to be studied. This paper defines a crowd energy to deal with crowd modeling and processing in the real-time surveillance. With wavelet analysis of the energy curve, a new approach is presented for monitoring two categories of abnormal events of the scene. The result of a metro video surveillance system has demonstrated the effectiveness of the approach.

In this paper, a finite-mixture-model learning bused sparse component analysis (SCA) algorithm is proposed. In this algorithm, a finite-mixture-model learning method is applied for estimating the mixing matrix for SCA. The main advantage of this method is the ability of selecting the number of sources and measuring reliability of the columns of the estimated mixing matrix. That is, it can give us a probability measurement of the recovered sources, which help us to determine which recovered sources are more reliable and significant. The simulation results show the effectiveness of this algorithm.

This paper presents a hybrid controller for hybrid electric vehicles using hybrid system theory. The hierarchical control structure of parallel hybrid electric vehicle (HEV) is introduced first and the drive control problem of the vehicle is formulated using hybrid system theory. State machine theory is applied to solve the problem of vehicle operating mode transitions. Then a fuzzy torque distributor is developed to deal with power splitting problem between the engine and the motor so that these power sources are operated at high efficiency. A backward-looking parallel HEV model was used for implementation and simulation of the controller. Potential fuel economy improvement has been shown by using this hybrid controller.

During the past few years, much effort has been made to enhance the performance of visual tracking by combining multiple features. In this paper, we propose a tracking algorithm based on region correlation descriptor, which can provide an elegant solution to fuse multiple features for representing the object Besides, we suggest a method to find the best matching region by calculating a centroid estimated from a set of candidate points in searching window. The robustness and accuracy of our algorithm have been demonstrated by experiments.