Search

According to the problem that achieves robust human actions recognition from image sequences in computer vision, using the Iterative Querying Heuristic algorithm as a guide, a improved Multiple Instance Learning (MIL) method is proposed for human action recognition in video image sequences. Experiments show that the new method can quickly recognize human actions and achieve high recognition rates, and on the Weizmann database validate our analysis.

We propose an efficient method for automatic detection of suspicious behavior in video surveillance data. First of all, we cluster a set of sequences labeled as normal or suspicious. Then, we assign new observation sequences to behavior clusters. We label a sequence as suspicious if it maps to an existing model of suspicious behavior or does not map to any existing model according to the corresponding HMMs. We evaluate our proposed method on a real-world video surveillance and find that the method is very effective at detecting suspicious behavior.

Human behavior recognition is an active research field in computer vision and image processing. A novel method is proposed for human behavior recognition in video image sequences. First of all, a video sequence is represented by extracting space-time interest points. Then Human behavior is represented by activities through Motion Decomposition. The activity comprises labeled bags that are composed of unlabeled instances comprising to action. Final labeled activities are used to train a strong classifier which is used to predict the labels of unseen behavior bags. Experimental results show the effectiveness of the proposed method in comparison with other related works in the literature and can also tolerate noise and interference conditions.

In this paper, we propose an effective approach for detecting moving vehicles in nighttime traffic scenes. We use Multiple Instance Learning method to automatically detect vehicle from video sequences by constructing the Multiple Instance Learning model at nighttime. At first, we extract SIFT feature using SIFT feature extraction algorithm, which is used to characterize moving vehicles at nighttime. Then Multiple Instance Learning model is used for the on-road detection of vehicles at nighttime, in order to improve the detection accuracy, the class label information was used for the learning of the Multiple Instance Learning model. Final experiments were performed and evaluate the proposed method at nighttime under urban traffic condition, the experiment results show that the average detection accuracy is over 96.2%, which validates that the proposed vehicle detection approach is feasible and effective for the on-road detection of vehicles at nighttime and identification in various nighttime environments.

Micro-structured targets have been widely used in the interaction between ultra-intense laser and target, aiming at improving the electron accelerating efficiency. In this paper, we perform two-dimensional particle-in-cell (PIC) simulations to study the interaction of the ultra-intense laser pulse with the micro-structured foam-attached target (the foam is composed of low density bubbles and high density interfaces between the bubbles). It is found that at the beginning of the laser-plasma interaction, the fast electrons accelerated at the front surface of the foam freely propagate into the target and drive a return current of cold background electrons. These cold background electrons are restricted to propagate along the interfaces between the bubbles in the foam due to the self-generated large sheath field. As a result, small current filaments are generated in the foam, which then leads to the generation of randomly distributed megagauss magnetic field in the foam layer. This quasistatic magnetic field then acts as an energy-selective " magnetic barrier”: the low-energy electrons are reflected back into the laser acceleration region while the high-energy electrons can penetrate through it. If the reflected electrons enter into the laser field with proper phases, they can be further accelerated to higher energy through cooperative actions of the ultra-intense laser pulse and the sheath field generated due to plasma expansion at the target surface. Our simulation results show that many of the laser accelerated low-energy electrons can be reflected back and accelerated several times until they gain enough energy to penetrate through the magnetic barrier. This is termed the " multiple acceleration mechanism”. Due to this mechanism, the electron acceleration efficiency in the foam-coated target with a thickness of several microns is significantly enhanced in comparison with that in the plane target. This enhancement in the electron acceleration efficiency will be beneficial to many important applications such as the fast ignition. Additionally, foam-coated targets with different bubble radii and layer thickness are also studied, and it is found that the yield of the high energy electrons increases with the radius of bubble size more efficiently than with the bubble thickness. In order to understand the physics more clearly, a single particle model is developed to analyze the simulation results.

Key frames play a very important role in video indexing and retrieval. In this paper, we propose a novel method to extract key frames based on information theory. We use improved Bayesian Information Criterion to determining the number of key frames, and then automatically extract key frames to represent video shot based on information bottleneck cluster method. Experimental results and the comparisons with other methods on various types of video sequences show the effectiveness of the proposed method.

In this paper, we introduce a new ear recognition approach including a feature extraction method and the recognition framework. Firstly, we use a supervised kernel neighborhood preserving projection to extract discriminately ear feature, then we use generalized locally nearest neighbor classifier to recognize ear. Experimental results on USTB ear database show the effectiveness of our method.

A convenient method of three-dimensional ray tracing is suggested in the geometrical optical approximation, in which the laser ray propagation is completely based on the concept of optical trajectory tracing rather than depending upon the effective force and propagation time. This tracing has two obvious advantages that the direct application of Snell's law can be avoided when a ray crosses a different density zone and that to any desired accuracy it takes much less computation time than existing tracing. For Gaussian light beam propagation, in a spherically symmetric plasma atmosphere, the results emphasize that the deposition uniformity is strongly dependent not only on the wavelength of the laser but also on the temperature of the plasma.