Abstract:
A novel algorithm for moving objects detection based on pixel classification is proposed. This algorithm preprocesses the images with illuminate standard in order to decline detection mistakes caused by illuminant changes. The pixels are classified by its characteristics into the single-model and the multi-model. The former uses median method for background modeling while the later uses GMM. The experiments show that this algorithm, compared with GMM, is computational efficient and can be used for real-time systems easily.
Key words:
background subtraction,
GMM,
median method,
moving objects detection,
pixel classification
摘要: 针对复杂环境下运动目标检测提出一种基于像素分类的运动目标检测算法。该算法通过亮度归一化对图像序列进行预处理,用以降低光照变化造成的误检,根据场景中不同像素点的特点,对图像进行分类处理,单模态类的像素用中值法进行背景建模,多模态类的像素用混合高斯模型建模。实验结果表明,该算法与传统的高斯建模法相比,减少了运算量,更易于应用在实时系统中。
关键词:
背景差分,
高斯混合模型,
中值法,
运动目标检测,
像素分类
CLC Number:
XU Yi-mei; GUO Bao-long; ZHANG Jin. Moving Objects Detection Algorithm Based on Pixel Classification[J]. Computer Engineering, 2008, 34(23): 205-207,.
徐以美;郭宝龙;张 晋. 基于像素分类的运动目标检测算法[J]. 计算机工程, 2008, 34(23): 205-207,.