Soccer Static Video Summarization System Based On Genetic Algorithm

With the popularity of the network and video capture such as the development ofmultimedia technology, a larger number of digital video comes. How to search adesired video clip effectively in a massive video database for saving user s time andeffort becomes an important research topic. Soccer video in the digital video accounting for alarge proportion, thus the abstract system for soccer video is significant.This paper presents a genetic algorithm (GA) based model for soccer videosummarization. Based on this model, the main work that has been done is described asfollowing:1. We study the concept, classes and characteristics of video summarization. Accordingto the specialty of the soccer video, we introduce the structure of this kind of video.2. We present the basic model for video summarization. First, the least similar imagesfrom original video were picked out by measuring the differences with the technique of colorhistograms. Then we introduce a novel fitness function, which is defined by length,commonality, and precedence factors. Based on the reduced set, the model using geneticalgorithm to search the optimized keyframes. Experimental results show that the proposedmodel can capture more information than the traditional method and avoid redundancy.Moreover this algorithm has a distinct advantage in terms of time.3. To improve the model, we introduce the audio features to get the better performance.Based on this fitness function, the model employs crossover and mutation operators to searchfor a meaningful summary in the video search space. This model shows a good performancefor producing a static summarization of soccer videos.4. We present a model using video time density function (VTDF). VTDF is employed tomodel the video sequence. Experimental results show that the GA based method with VTDFand audio information is more suitable for extracting abstracts from soccer videos.5. We investigate the performance of the Binary Genetic Algorithm (BGA) method andthe Decimal Genetic Algorithm (DGA) method. Based on the above model, we compared theBGA and DGA method about the time issue and repeat performance. Experimental results and comparisons are presented to show the high quality of BGA method for static videosummarization.