一种基于空间信息的运动目标检测方法

Improved Moving Objects Detection Based on Spatial Information

针对现实生活中由于光照变化、背景噪声干扰、摄像机抖动等因素对运动目标的检测与识别存在巨大挑战的问题,提出了一种基于空间信息的运动目标检测算法。通过对像素点及其区域的亮度和角度差分等信息提取特征,建立背景模型,去除光照因素的干扰,比较当前帧和背景模型的相似系数确定前景区域,并且采用了自适应阈值的方法二值化前景图。实验证明,该方法能克服光照突变等复杂背景的干扰,实现对运动目标实时准确检测。

There is a big challenge in detecting the moving objects in complicated environment because of the variable light,noise and camera shake.A new background subtraction was proposed to distinguish the moving objects in unconstrained environments,using both the single pixel value and its spatial information.The new background model combines the angles and intensity in two vectors,comparing the current image with the background image.An efficient method of obtaining the binary foreground through the comparison of image and mean vector of training sequence works well in removing the noise of illumination change and improving the accuracy.The experimental results demonstrate the proposed algorithm can overcome interference from sudden change of background light and achieve exact detection of moving objects.