中文摘要：

提出一种光照变化场景建模方法，并利用该模型对目标进行检测。首先分析光照变化场景图像序列的平稳性，建立自回归模型。考虑到光照变化和成像系统噪声的影响，将像素亮度扰动参数引入到自回归模型。该模型在线自学习，跟踪并预报场景中的光照变化情况，在该模型的基础上，提出像素亮度扰动置信区间，对光照变化场景中的目标进行检测。实验结果表明，该模型能够准确描述和跟踪场景的光照变化。基于模型的预报值和观测值及置信区间，可以对场景中的目标进行准确检测；检测结果表明，该方法对光照变化是不敏感的。

英文摘要：

A scene modeling method under lighting variations is presented, and this model is used to detect the objects in the scene. Firstly, the stability of image sequence in lighting variations scene is analyzed. The corresponding autoregressive model is structured. Con- sidering the lighting variations and the noise of imaging system, a set of perturbation parameters ofpixel＇s intensity, which are on-line and self-learn, is introduced to this model. Moreover, lighting variations of the scene are tracked and forecasted. Based on the model, a confi- dence interval with perturbation ofpixel＇s intensity is derived, and object detection in the scene is accomplished. The experimental results show that scene under lighting variations can be exactly described and tracked with the proposed model. According to the forecasting value, the observation value and the confidence interval of this model, objects in the scene can be exactly detected, which reveals the insensitivity to lighting variations.