Abstract
Three topics play an important role in modern pattern recognition research: (a) problems where the different elements should be structurally described as networks with connections linking those elements; (b) dynamic problems where the information may evolve along some independent variable (e.g. time in context of the video sequences); (c) problems that present both structural and dynamic aspects, i.e. a network of elements presenting a dynamic evolution along some independent variable. This project, congregating the Vision groups at IME-USP and IFSC-USP, besides other collaborators, aims at studying, developing and applying pattern recognition techniques in the context of the three aforementioned topics. The work will concentrate both on basic pattern recognition research and on applications on computer vision, image processing and bioinformatics. As far as structural problems are concerned, the research will focus on the application of pattern recognition techniques to the analysis of networks in specific applications, as well as on the development of pattern recognition methods based on network (graph) structures. Special attention will be devoted to the use of graphs in structural pattern recognition and spatial reasoning. Image analysis may be carried out using not just object attributes, but also the spatial relation among such objects in the scene. Such spatial relations are often more stable than many object image properties (e.g. due to illumination variations, image processing, noise, etc.) Graph models describing image elements will be developed and applied in the research. The graph model nodes represent the elements in a scene and the arcs store spatial relation information among such elements. Besides the structural approach, dynamic pattern recognition will be also addressed. Problems in this field present some dynamic aspect, i.e. the patterns evolve along some independent variable. An example of important research problem that will be treated in the research is video sequence analysis, where image elements evolve along time. Different sub-projects of the forseen research are based on video sequence analysis. Finally, pattern recognition techniques to analyze network dynamics will be developed. Problems in this context present a network whose behavior evolve along some independent variable. An important example of such a problem arises in bioinformatics: the so called gene expression networks. Genes are modeled as elements that may communicate to each other through gene expression. Such communication channels are represented by links of the gene network model. Therefore, each gene may have influence in the dynamic behavior of other genes, thus creating a dynamic behavior of the gene network. The developed techniques will be used and assessed in specific applications in computer vision and in computational biology. As far as computer vision is concerned, the following applications will be addressed: (1) image processing by W-operators; (2) digital video analysis for tracking, segmentation and recognition; (3) shape classification and structural pattern recognition; (4) analysis of branching structures that form networks. On the other hand, the following computational biology problems will be of interest: (1) gene expression networks identification; (2) development of signal processing and pattern recognition methods to solve classification problems in bioinformatics. (AU)
| Articles published in Agência FAPESP Newsletter about the research grant: |
| More itemsLess items |
| TITULO |
| Articles published in other media outlets ( ): |
| More itemsLess items |
| VEICULO: TITULO (DATA) |
| VEICULO: TITULO (DATA) |