CAJ | 학술논문

在基于视频的交通流检测系统中，静止背景的快速准确提取是基础。背景的提取主要分为模型法和非模型法，非模型法的算法复杂度低，能满足交通流检测实时性要求。论文比较和分析了均值法、直方图统计法、中值滤波法和背景学习法四种常用的背景提取方法。并通过实际的交通视频测试比较了四种方法。
재기우시빈적교통류검측계통중，정지배경적쾌속준학제취시기출。배경적제취주요분위모형법화비모형법，비모형법적산법복잡도저，능만족교통류검측실시성요구。논문비교화분석료균치법、직방도통계법、중치려파법화배경학습법사충상용적배경제취방법。병통과실제적교통시빈측시비교료사충방법。
Background extraction is a fundamental and critical task in video traffic detection system. Approaches of background extraction can be divided into two groups, that are the methods based on model and the model-free methods. With lower time-consuming feature, model-free methods can satisfy the demand on real-time processing in video traffic detection system. This essay compares four common model-free methods, including medial filtering, histogram method, median filtering, and background learning method by analyzing traffic video, and gains a conclusion on their features.