Tracking Multiple Variable-Sizes Moving Objects in LFR Videos Using a Novel genetic Algorithm Approach

摘要: All available methods for tracking motion objects in videos have several challenges many situations yet. For example, Particle filter cannot track that variable sizes within frames duration efficiently. In this work, a novel multi-objective co-evolution genetic algorithm approach is developed can efficiently the size low frame rate videos. We test our method on famous PETS datasets 10 categories with different rates and number of each scene. Our proposed robust tracker against temporal resolution changes it has better results accuracy (about 10%) lower false positive rate(about 7.5%) than classic particle GA which contain small objects. Also uses only 5 second instead 15 or more frames.