Font Size: a A A

Researches On The Multiscale Modeling Method For A Stochastic Process

Posted on:2007-10-13Degree:MasterType:Thesis
Country:ChinaCandidate:J J ShiFull Text:PDF
GTID:2120360185953954Subject:Applied Mathematics
Abstract/Summary:PDF Full Text Request
In modern high-performance, multi-level, complex system, the problems about the multi-sensor data fusion are often encountered, which effectively integrate the information obtained by different types sensors in different scales. In recent years, to solve this kind of problem more effectively, many scholars research thoroughly in depth on multis-cale systems theory and related multi-scale representation methods and processing algorithms for the random process, this study has become the hot spot in many scientific research fields.Multi-scale stochastic model has been applyed broadly in many practical issues, such as image fusion, image denoising, boundary detection, texture classification, image segmentation, medical image analysis, as well as oceanographic, geophysical remote sensing and voice science and so on. The established model based on the multi-scale dynamic tree structure for the phenomena or processes, is not only an important way to acquire the data analysis or signal processing with the multi-scale characteristic, but is used to induce the highly efficient parallel iteration algorithms to the state variable for optimally estimating the stochastic process. In the actual project, because of the diversity and the actual complexity in the sensors, the sampling rates between the sensors usually present the indefinite proportional relationship. Therefore, based on the regular tree modeling method, this dissertation has done the following several aspects works:1. Propose the multi-scale representation and modeling method on the irregular tree. The multi-scale modeling method is mainly aimed at the observation systems with different sampling rate sensors, and is completed by determining the state transition matrix between the father and son nodes on the tree according to the probability - weighted approach, then determining the irregular multi-scale representation, and finally inferring the state transition matrix, disturbance matrix, the initial state and corresponding covariance matrix in the multi-scale model.2. Apply the multi-scale stochastic modeling idea in the image fusion, and...
Keywords/Search Tags:Multi-scale modeling, multi-scale representation, data fusion, Markov process
PDF Full Text Request
Related items