Catálogo de publicaciones - libros

Storm detection algorithm is a key element of the severe weather surveillance service based on radar image data. 3-D clustering technique is the fundamental part of storm detection. During the clustering process, the connection area between adjacent storms may cause the existing algorithms to identify them as one storm wrongly. Isolating storms from a cluster of storms is another difficulty. To overcome these difficulties, this paper introduces a novel approach which combines the strengths of erosion and dilation in a special way. First, the erosion operation is used to solve the problem of false merger. Then the dilation operation is performed when using gradually increased threshold to detect storms. This keeps the internal structure information of sub-storms well when isolating storms from a cluster of storms. The results of the experiment show that this method can correctly recognize adjacent storms. And when isolating storms from a cluster of storms, this method can also keep the internal structure of sub-storms which will benefit the following tracking task.

Intelligent traffic surveillance systems are assuming an increasingly important role in highway monitoring and city road management systems. Recently a novel feature was proposed to improve the accuracy of object localization and occlusion handling. It was constructed on the basis of the strong shadow under the vehicle in real-world traffic scene. In this paper, we use some statistical parameters of each frame to detect and segment these shadows. To demonstrate robustness and accuracy of our proposed approach, impressive results of our method in real traffic images including high congestion, noise, clutter, snow, and rain containing cast shadows, bad illumination conditions and occlusions, taken from both outdoor highways and city roads are presented.

Face recognition is one of the most successful applications in biometric authentication. However, methods reported in the literature still suffer from some problems which prevent the further development in face recognition. This paper presents a novel robust method for face recognition under near infrared (NIR) lighting condition based on Extended Local Binary Pattern (ELBP), which solves the problems produced by variations of illumination rightly, since the NIR images are insensitive to variations of ambient lighting, and ELBP can extract adequate texture features form the NIR images. By combining the local feature vectors, a global feature vector is formed and as the global feature vectors extracted by ELBP operator often have very high dimensions, a classifier has been trained using the AdaBoost algorithm to select the most representative features for better performance and dimensionality reduction. Compared with the huge number of features produced by ELBP operator, only a small part of the features are selected in this paper, which saves much computation and time cost. The comparison with the results of classic algorithms proves the effectiveness of the proposed method.

In this paper we propose a new technique for modeling three-dimensional rigid objects by encoding the fluctuation of the surface and the variation of its normal around an oriented surface point, as the surface expands. The surface of the object is encoded into three vectors as the surface signature on each point, and then the collection of signatures is used to model and match the object. The signatures encode the curvature, symmetry, and convexity of the surface around an oriented point. This modeling technique is robust to scale, orientation, sampling resolution, noise, occlusion, and cluttering.

We propose an event extraction method from traffic image seque-nces. This method extracts moving objects and their trajectories from image sequences recorded by a stationary camera. These trajectories are mapped to 3D virtual space and physical parameters such as velocity and direction are estimated. After that, traffic events are extracted from these trajectories and physical parameters based on case-frame analysis in the field of natural language processing. Our method facilitates to describe events easily and detect general traffic events and abnormal situations. The experimental results of actual intersection traffic image sequence have shown the effectiveness of the method.

In order to address the challenges of occlusions and background variations, we propose a novel and effective rule-based multiple object tracking system for traffic surveillance using a collaborative background extraction algorithm. The collaborative background extraction algorithm collaboratively extracts a background from multiple independent extractions to remove spurious background pixels. The rule-based strategies are applied for thresholding, outlier removal, object consolidation, separating neighboring objects, and shadow removal. Empirical results show that our multiple object tracking system is highly accurate for traffic surveillance under occlusion conditions.

This paper presents an effective approach for iris recognition by analyzing the iris patterns. We propose an iris classification method that divides the normalized iris image into several regions to avoid the iris image with several noise factors (eyelids and eyelashes) and reduce the error rates. In every region, effective features are extracted by the proposed method of local edge pattern (LEP) for edge and corner detection. Feature vectors are linearly combined into a two dimensional matrix that represents every iris image for further recognition. Then 2D linear discriminant analysis (2DLDA) is used to identify the person. We use two public and freely available iris image databases for evaluation, organized in training and test sets respectively. Experimental results show that the recognition rate of the two iris image databases have achieved similar performance more than 98% and the proposed method has an encouraging performance and robustness.

A novel method for automatic registration based on Iterative Closest Point (ICP) approach is proposed. This method uses geometric bounding containers to evaluate the optimum overlap rate of the data and model point sets.

Traditional classification algorithms are not suitable for feature extraction on high resolution satellite images, given the heterogeneity of the pixels of this type of imagery due to a great amount of detail. Most of this type of imagery is taken by the satellite in several bands and at different resolutions, and the method presented in this paper takes advantage of this situation, merging information provided by the multispectral bands with the panchromatic band. An Ikonos image of 2 x 2 km of the university campus of Alcala has been used for obtaining a classification with seven land use classes. A comparison is carried out between the traditional maximum likelihood method and the method developed here. The latter using context information obtained by the texture from the band with the maximum resolution, the panchromatic band. The results show how texture information improves maximum likelihood classification of the multispectral bands for smooth-textured classes.

Certain components in images can be recognized with high accuracy, for example, backgrounds such as leaves, grass, snow, sky, water. These components provide the human eye with context for identifying items in the foreground. Likewise for the machine, the identification of background should help in the recognition of foreground objects. But, in this case, the computer needs explicit lists of object and background co-occurrence probabilities. We examine two ways of deriving estimates of these a priori object co-occurrence probabilities: using an online social network of people storing annotated images, FlickR; and using variations on co-occurrence frequencies in natural language text. We show that the object co-occurrence probabilities derived from both sources are very similar. The possibility of using non-image derived semantic knowledge drawn from text processing for object recognition opens up possibilities of mining a priori probabilities for a much wider class of objects than those found in manually annotated collections.