摘要
针对场景发生改变的目标检测,提出了一种快速的目标检测算法。该算法将像素点划分为背景点和前景点两类,使用聚类的方法并采取像素级收敛条件分别建立背景和前景模型。背景模型的更新适用于全局发生缓慢的变化,稳定的前景模型向背景模型的转化对光照等其他场景突变具有很好的鲁棒性,使背景模型实时逼近真实背景,同时改进连通域扫描分割算法,提高目标分割速度。实验结果表明,该方法具有很好的鲁棒性,能够快速准确地检测出运动目标,为在DSP等嵌入式系统上实现实时目标检测提供了有利条件。
Aimed at the object detection in scene change a fast object detection method was proposed.The algorithm divided the pixels into background pixels and foreground pixels,the clustering method and the pixel-level convergence criteria was used to establish the background and foreground model.The background model was updated to suit for the global change.When the stable foreground model transformed to the background,it had a good robustness to the light and other scene mutation.The background model could be approximated to the real background,meanwhile,the connected domain scan segmentation algorithm was improved.The experiments demonstrate that this method have a good robustness,can rapidly and accurately detect the moving object.It provides favorable conditions for real-time target detection in the DSP and other embedded system.
出处
《电子测量与仪器学报》
CSCD
2012年第3期261-266,共6页
Journal of Electronic Measurement and Instrumentation
基金
安徽省信息产业厅信息产业发展基金项目(No.2008012)
特种显示技术教育部重点实验室开放课题基金项目(No.2008HGXJ0350)
关键词
复杂场景
目标检测
像素级收敛
光照突变
连通域扫描
complex scene
target detection
pixel-level convergence
light mutation
connected domain scan
