摘要
论述了一种实时的前景-背景分割运动目标检测方法。用中值法获得背景模型;为避免规范化色彩的不确定性,背景图像像素点的亮度被比例缩放为当前帧对应像素点的亮度,在此基础上求出色度偏差,并基于当前帧和背景帧的颜色偏差和亮度偏差来检测目标;最后用数学形态学滤波及连通分量分析的方法进行去噪、区域连通等后处理。方法建立了科学的亮度与颜色的关系模型,充分利用了图像的灰度信息和色彩信息。实验表明,方法具有很好的检测效果。
This paper presents a real - time algorithm of foreground - background segmentation for Moving Object Detection. The background model is generated by using the medium pixel values from an image sequence. In order to avoid the instability of normalized colors, the brightness of a background pixel is geometrically rescaled to the brightness of the current - frame pixel, and in this way the color distortion is calculated. Then subtract the current image from the background model and detect moving objects in color distortion and bright distortion. At last image morphologic filtering and connected component measurement are employed to remove the flecks of noise in the segmented binary image or detect the connectivity of different object area. Evaluation shows that this algorithm is simple and effective.
出处
《计算机仿真》
CSCD
2008年第10期194-196,201,共4页
Computer Simulation
关键词
减背景
色度偏差
亮度偏差
形态学滤波
Background subtraction
Color distortion
Brightness distortion
Morphologic filtering
