摘要
为了能在复杂背景图像序列中进行变化目标检测,提出一种基于统计检验的变化检测算法。利用高斯混合模型计算背景像素和观测像素的经验分布函数,根据假设检验理论计算RBJ统计量,通过待检测像素与背景模型的拟合程度判断像素归属,得到粗略的变化目标,采用高斯分裂原则自适应更新背景分布函数,使背景模型能进一步逼近真实背景,从而得到最终变化目标。仿真结果表明,针对复杂背景的图像序列,该算法能够有效抑制恶劣天气对检测的干扰,查准率较高,综合检测性能指标较好。
An algorithm of change detection based on statistical tests for image sequences is proposed.It achieves to change detection of complex background.Gaussian Mixture Model（GMM）is used to calculate the background pixels and observation of empirical distribution function.It calculats RBJ statistics according to the hypothesis testing theory,uses the pixel and the background of the model fitting degree to make judgment of pixel belongs,and gets rough target change.In order to make the background model can be further close to the real background,it uses Gaussian division principle to adaptively update background distribution function to get the final target.Simulation results show that,for the complex background image sequences,the algorithm can effectively restrain the interference of bad weather,get high precision,and can obtain a better detection performance index.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第12期226-230,共5页
Computer Engineering
基金
国家自然科学基金资助项目(61104213)
关键词
拟合优度检验
经验分布
变化检测
高斯混合模型
高斯分裂原则
goodness of fit testing
empirical distribution
change detection
Gaussian Mixture Model（GMM）
Gaussian division principle
