Catálogo de publicaciones - libros

This paper presents an approach of target word selection for Korean verbs based on lexical knowledge contained in a Korean-English bilingual dictionary and WordNetWe focus on deciding which translation of the target word is the most appropriate using the measure of semantic relatedness between possible translations of target word and some indicative clue words. With five Korean ambiguous verbs, we report an average accuracy of 51% that outperforms the default baseline performance and previous works.

A novel segmentation algorithm for natural color image is proposed. Fibonacci Lattice-based Sampling is used to get the symbols of image so as to make each pixel’s label containing color information rather than only as a class marker. Next, Region map is formed based on Fibonacci Lattice symbols to depict homogeneous regions. On the other hand, by applying fuzzy homogeneity algorithm on the image, we filter it to acquire Edge map. To strengthen the ability of discrimination, both the weighted maps are combined to form Region-Edge map. Based on above processes, growing-merging method is used to segment the image. Finally, experiments show very promising results.

This paper proposes a novel method for the recognition of passports based on a FCM-based RBF network. First, for the extraction of individual codes for recognizing, this paper targets code sequence blocks including individual codes by applying Sobel masking, horizontal smearing and a contour tracking algorithm on the passport image. As the last step, individual codes are recovered and extracted from the binarized areas by applying CDM masking and vertical smearing. This paper also proposes a FCM-based RBF network that adapts the FCM algorithm for the middle layer. This network is applied to the recognition of individual codes. The results of the experiments for performance evaluation on the real passport images showed that the proposed method has the better performance compared with other approaches.

This paper describes a vision system for extracting and recognising partially occluded 2D visual landmarks. The system is developed based on the traditional template matching approach and a memory feedback modulation (MFM) mechanism. It identifies the obscured portions and selectively enhances non-occluded areas of the landmark, while simultaneously suppressing background clutters of the bottom-up edge processed input images. The architecture has been tested with a large number of real images with varying levels of landmark concealment and further evaluated using a vision-based navigating robot in the laboratory environment.

This paper proposes a two-phase diversity control approach to prevent the common problem of the loss of diversity in Genetic Programming with Automatically Defined Functions. While most recent work focuses on diagnosing and remedying the loss of diversity, this approach aims to prevent the loss of diversity in the early stage through a refined diversity control method and a fully covered tournament selection method. The results on regression tasks suggest that these methods can effectively improve the system performance by reducing the incidences of premature convergence and the number of generations needed for finding an optimal solution.

This paper describes the personal locating system in image sequence using a vision-based augmented reality technique which allows the user to navigate an unfamiliar and unknown place in an office environment. For identifying personal location in image sequences, the system uses a color histogram matching method and location model. The results are overplayed on the user’s view through AR technique. This system is applicable to guide an application.

In machine vision, template matching is key component and used usefully in various tasks such as pick and place, mark identification, and alignment. In this paper, we propose fast template matching algorithm using edge projection. Proposed algorithm reduces the search problem from 2D into 1D using edge projection within the 2D template area. By this, it could effectively reduce the computational burden. Also, it gives comparable discriminating power compared to template matching using intensity. In this paper, rotation and translation search is implemented to cope with typical machine vision application where the height between camera and target object is fixed.

In this paper we propose a clustering algorithm called s-Cluster for analysis of gene expression data based on pattern-similarity. The algorithm captures the tight clusters exhibiting strong similar expression patterns in Microarray data,and allows a high level of overlap among discovered clusters without completely grouping all genes like other algorithms. This reflects the biological fact that not all functions are turned on in an experiment, and that many genes are co-expressed in multiple groups in response to different stimuli. The experiments have demonstrated that the proposed algorithm successfully groups the genes with strong similar expression patterns and that the found clusters are interpretable.

In this paper, we present a segment-based stereo matching algorithm using adaptive variable windows and dynamic programming with a robust disparity. We solve the problem of window shape and size using adaptive line masks and adaptive rectangular windows which are constrained by segments and visibility that reduces ambiguity produced by the occlusion in the computation window. In dynamic programming, we also propose the method that selects an efficient occlusion penalty.

Auto-playing programs are often used on behalf of human players in a MMORPG(Massively Multi-player Online Role Playing Game). By playing automatically and continuously, it helps to speed up the game character’s level-up process. However, the auto-playing programs, either software or hardware, do harm to games servers in various ways including abuse of resources. In this paper, we propose a way of detecting the auto programs by analyzing the window event sequences produced by the game players. In our proposed method, the event sequences are transformed into a set of attributes, and various learning algorithms are applied to classify the data represented by the set of attribute values into human or auto player. The results from experiments with several MMORPGs show that the Decision Tree learning with proposed method can identify the auto-playing programs with high accuracy.