针对混合高斯和多模态均值模型在阴影、噪声、扰动、计算量和存储空间等方面的优缺点,提出一种基于区域纹理的目标检测方法.该方法分析场景纹理分布,制定区域复杂度分类准则,对复杂区域采用混合高斯,对简单区域采用基于运动历史的多模态均值进行目标检测.实验表明,该方法能在多模态环境中克服噪声与扰动,实时转换前/背景,优化速度和存储量,准确检测出目标.%The mixture Gaussian and multimodal mean model with respect to shadow, noise, background disturbance, computation and memory space have some advantages and disadvantages, this paper proposed a method of motion detection based on region textures. First,it analysed texture distribution of scene. Second, made a rule of classification to judge regional complexity. Lastly, adopted the mixture Gaussian model in the complex region and the multimodal mean based on motion history in the simple region to detect moving target. Experimental results show that the algorithm can overcome noises and background disturbances in multimodal environment, and real-time update foreground and background. It optimizes the speed and memory space, and detects moving target accurately.
展开▼
