Video Summarization Model Based On Genetic Algorithm

With the rapid development of computer network and widely used of multimedia technology, a large number of video data has emerged. How the user can quickly find the needed video in a large number of video data for saving user’s time availability, efficiency, has become a problem urgently. The key to this question is to make the long video extraction into short video, but you must keep the main content of the original video.This paper presents a genetic algorithm (GA) based model for video summarization. Based on this model, the main work that has been done is described as following:1. We study the concept, classes and characteristics of video summarization, and we introduce the structure of this kind of video.2. In this paper, we propose a new content-based rapid video playback method. We use the temporal quantization by using the motion-based video time density function to find the optimal quanta and partition in time domain. The quanta and partition in time domain they are connected with each other. Based on genetic algorithm using fitness function to determine the best quanta and partition in time domain. Experimental results show that using the genetic algorithm can capture more information, reduce the redundant information, and at the same time can get better fitness function value.3. We improve the traditional genetic algorithm by creating and maintaining the diversity of population. We introduce the standard of population diversity (SPD) and health of population diversity (HPD) measurements to ensure the diversity of population. With the improved adaptive genetic algorithm, which can avoid local convergence, the optimal solution can be ger. Experimental results show that this algorithm can get better fitness function value, but the convergence speed is slower than traditional algorithm.4. We introduce benchmark functions to study the performance of the improved adaptive genetic algorithm and the traditional genetic algorithm. We compare the average population fitness value, the diversity of population index SPD and HPD. The experimental results show that the improved adaptive genetic algorithm in the diversity of population displays the great advantage, but the convergence speed is not very predominate.