联合纹理和二维阈值分割的减背景方法

Background subtraction algorithm based on texture and two-dimensional threshold segmentation

为了有效地检测出交通场景中的运动车辆,在背景建模阶段,提出在多个子时间组里建立多个自适应的纹理直方图作为像素的背景描述模型,能更好地适应环境变化。为了充分地描述纹理信息,扩展了LBP(local binary pattern)算子;为了更好地实现减背景方法,采用的二维Otsu阈值分割方法能有效消除噪声的影响;为了解决背景更新滞后的问题,使用帧间差分能更准确地提取出运动车辆。在不同交通场景下的实验比较结果表明了该算法的有效性。

In order to detect effectively the movement vehicles of traffic scenarios, to built a group of adaptive texture histograms in sub-time blocks for each pixel during background modeling is presented, which can adapt to environment changes. And LBP is modified to represent texture better. Then in order to implement the background subtraction better, two-dimensional threshold segmentation is performed to eliminate the effect of noise, meanwhile flame-difference is used to exactly extract the movement vehicles, which can resolve the problem of background updating lag. Experimental results demonstrate the effectiveness of the proposed method under different traffic scenarios.