Solving Combinatorial Problems with Time Constrains Using Estimation of Distribution Algorithms and Their Application in Video-Tracking Systems
Antonio Berlanga, Miguel A., Jess Garca, Jos M. · InTech eBooks · 2010
In this chapter, the association problem for real-time tracking in video was formulated as search in a hypotheses space. It is defined as a combinatorial problem, constraining the computational load to allow image processing in real time of the sequence of frames. Evolutionary Computation techniques have been applied for solving this search problem, in particular Estimation Distribution Algorithms (EDA) that shows an efficient computational behaviour for real-time problems. The authors have done an exhaustive analysis of EDAs algorithms using several KP0/1 problems. These experimentations help the authors to know the complexity degree of the association problem and to find out the most suitable parameters for real-time video tracking problem. From the parameters obtained in analyzing the KP0/1 problems, the author have been carried out a wide comparison among standard Genetic Algorithm, Particle Filtering based on Mean-Shift weight and several EDA algorithms: CGA, UMDA and PBIL. Three video recordings of different complexity and problematic characteristics have been used to analyze the algorithms performance. Results show the efficiency of EDA algorithms to solve the combinatorial problem in real time and the capacity to be applied in video tracking systems.