摘要
针对室外视频监视环境复杂,现有的运动目标检测方法难以在克服背景干扰的同时准确地检测到慢速目标和运动着的小目标,且存在准确性低的问题,提出一种基于三阶中心矩场景区域分类的运动目标检测方法.由于前景区域、扰动区域和背景区域内真实运动、无意义运动、背景噪声像素值变化规律不同,采用三阶中心矩建立区域内像素值变化和区域类型的对应关系,设计了基于三阶中心矩的分类器以完成自适应场景区域分类,最终在区域分类的结果上检测运动目标.实验结果表明,该方法针对室外监视视频区域分类结果良好,能够克服树枝叶晃动、水面波动等背景干扰,可以准确地检测到慢速目标和运动着的小目标.
For the complex outdoor video surveillance, the existing moving object detection methods are dif-ficult to overcome the background interference and accurately detect slow moving objects, as well as small objects simultaneously. These methods have the problem of low accuracy. This paper proposes a moving object detection method based on region classification using three-order central moment. For the different change characteristics of the different pixels in foreground region, cluster region, and background region, which imply the true motion, meaningless motion and background noise, three-order central moment is em-ployed to build the corresponding relationship between the changes of pixel values and the classification of different regions. Therefore, the classifier based on three-order central moments is designed to adaptively classify the scene regions. Accordingly, the moving objects can be detected based on the region classification. The experimental results demonstrate that the proposed method can classify the different regions for outdoor video surveillance effectively. It also can overcome the background interference, e.g., branches swings and water fluctuates, and can detect slow objects and moving small objects accurately.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2015年第5期832-840,846,共10页
Journal of Computer-Aided Design & Computer Graphics
基金
国家"八六三"高技术研究发展计划(2014AA015104)
国家自然科学基金(61370124)
国家杰出青年科学基金(61125206)
国家留学基金(201303070205)
关键词
运动目标检测
区域分类
三阶中心矩
分类器
moving object detection
region classification
three-order central moment
classifier
