Catálogo de publicaciones - libros

Small sample is an acute problem in many application domains, which may be partially addressed by feature selection or dimensionality reduction. For the purpose of distance learning, we describe a method for feature selection using equivalence constraints between pairs of datapoints. The method is based on 1 regularization and optimization. Feature selection is then incorporated into an existing non-parametric method for distance learning, which is based on the boosting of constrained generative models. Thus the final algorithm employs dynamical feature selection, where features are selected anew in each boosting iteration based on the weighted training data. We tested our algorithm on the classification of facial images, using two public domain databases. We show the results of extensive experiments where our method performed much better than a number of competing methods, including the original boosting-based distance learning method and two commonly used Mahalanobis metrics.

Several techniques have been proposed to date to build colour invariants between camera views with varying illumination conditions. In this paper, we propose to improve colour invariance by using data-dependent techniques. To this aim, we compare the effectiveness of histogram stretching, illumination filtration, full histogram equalisation and controlled histogram equalisation in a video surveillance domain. All such techniques have limited computational requirements and are therefore suitable for real time implementation. Controlled histogram equalisation is a modified histogram equalisation operating under the influence of a control parameter [1]. Our empirical comparison looks at the ability of these techniques to make the global colour appearance of single human targets more matchable under illumination changes, whilst still discriminating between different people. Tests are conducted on the appearance of individuals from two camera views with greatly differing illumination conditions and invariance is evaluated through a similarity measure based upon colour histograms. In general, our results indicate that these techniques improve colour invariance; amongst them, full and controlled equalisation consistently showed the best performance.

In this paper, foreground detection is performed by scene interpretation. A natural scene in different illumination conditions is characterized by scene context which contains spatial and appearance representations. The spatial representation is obtained in two steps. First, the large homogenous regions in each sample image are extracted using local and global dominant color histograms (DCH). Then, the latent semantic regions of the scene are generated by combining the coincident regions in the segmented images. The appearance representation is learned by the probabilistic latent semantic analysis (PLSA) model with local DCH visual words. The scene context is then applied to interpret incoming images from the scene. For a new image, its global appearance is first recognized and then the pixels are labelled under the constraint of the scene appearance. The proposed method has been tested on various scenes under different weather conditions and very promising results have been obtained.

Service robots need to be able to recognize and identify objects located within complex backgrounds. Since no single method may work in every situation, several methods need to be combined. However, there are several cases when autonomous recognition methods fail. We propose several types of interactive recognition methods in those cases. Each one takes place at the failures of autonomous methods in different situations. We proposed four types of interactive methods such that robot may know the current situation and initiate the appropriate interaction with the user. Moreover we propose the grammar and sentence patterns for the instructions used by the user. We also propose an interactive learning process which can be used to learn or improve an object model through failures.

Floating mines are a significant threat to the safety of ships in theatres of military or terrorist conflict. Automating mine detection is difficult, due to the unpredictable environment and high requirements for robustness and accuracy. In this paper, a floating mine detection algorithm using motion analysis methods is proposed. The algorithm aims to locate suspicious regions in the scene using contrast and motion information, specifically regions that exhibit certain predefined motion patterns. Throughput of the algorithm is improved with a parallel pipelined data flow. Moreover, this data flow enables further computational performance improvements though special hardware such as field programmable gate arrays (FPGA) or Graphics Processing Units (GPUs). Experimental results show that this algorithm is able to detect mine regions in the video with reasonable false positive and minimum false negative rates.

Background modeling and subtraction using subspaces is attractive in real-time computer vision applications due to its low computational cost. However, the application of this method is mostly limited to the gray-scale images since the integration of multi-channel data is not straightforward; it involves much higher dimensional space and causes additional difficulty to manage data in general. We propose an efficient background modeling and subtraction algorithm using 2-Dimensional Principal Component Analysis (2DPCA) [1], where multi-channel data are naturally integrated in eigenbackground framework [2] with no additional dimensionality. It is shown that the principal components in 2DPCA are computed efficiently by transformation to standard PCA. We also propose an incremental algorithm to update eigenvectors to handle temporal variations of background. The proposed algorithm is applied to 3-channel (RGB) and 4-channel (RGB+IR) data, and compared with standard subspace-based as well as pixel-wise density-based method.

Understanding intent is an important aspect of communication among people and is an essential component of the human cognitive system. This capability is particularly relevant for situations that involve collaboration among multiple agents or detection of situations that can pose a particular threat. We propose an approach that allows a physical robot to detect the intentions of others based on experience acquired through its own sensory-motor abilities. It uses this experience while taking the perspective of the agent whose intent should be recognized. The robot’s capability to observe and analyze the current scene employs a novel vision-based technique for target detection and tracking, using a non-parametric recursive modeling approach. Our intent recognition method uses a novel formulation of Hidden Markov Models (HMM’s) designed to model a robot’s experience and its interaction with the world while performing various actions.

We formulate decomposition of two-dimensional shapes as a combinatorial optimization problem and present a dynamic programming algorithm that solves it.

This paper proposes a new texture classification system, which is distinguished by: (1) a new rotation-invariant image descriptor based on Steerable Pyramid Decomposition, and (2) by a novel multi-class recognition method based on Optimum Path Forest. By combining the discriminating power of our image descriptor and classifier, our system uses small size feature vectors to characterize texture images without compromising overall classification rates. State-of-the-art recognition results are further presented on the Brodatz dataset. High classification rates demonstrate the superiority of the proposed method.

Schlick’s approximation of the term is used primarily to reduce the complexity of specular lighting calculations in graphics applications. Since moment functions have a kernel defined using a monomial , the same approximation could be effectively used in the computation of normalized geometric moments and invariants. This paper outlines a framework for computing moments of various orders of an image using a simplified kernel, and shows the advantages provided by the approximating function through a series of experimental results.