相机抖动场景中数据驱动的背景图像更新算法

Data-driven background representation algorithm in camera jitter scene

中文关键词  运动目标检测  特征检测  数据驱动  无模型自适应控制  背景图像表达  背景减除法

英文关键词  moving object detection  feature detection  data-driven  model free adaptive control  model free adaptive control  background subtraction

中文摘要

      视频序列中的运动目标检测是计算机视觉领域的重要研究课题. 背景减除是运动目标检测的有效方法, 但 相机抖动会对背景提取带来极大干扰, 从而可能造成传统基于模型的图像处理方法模型失真. 本文提出了相机抖动 场景下前景图像提取的数据驱动背景图像更新控制算法. 首先利用Harris特征检测进行背景补偿以消除抖动干扰. 然后利用无模型自适应控制方法, 建立单入单出控制系统来表示背景图像并进行实时更新. 最后运用背景减除法提 取运动目标前景图像. 本文方法与传统基于模型方法进行了不同视频序列的对比仿真. 实验结果表明, 本文方法可 以有效处理相机抖动场景下的运动目标检测问题, 目标前景图像分离效果更加接近真实场景.

英文摘要

      Moving object detection in video sequences is an important topic in the field of computer vision. Background subtraction is an effective method for moving object detection. Camera jitter will bring great interference to the extraction of background image, which could make the traditional model-based methods more likely image distortion. In this paper, a data-driven background representation algorithm in camera jitter is proposed. First, the Harris feature detection is used for background compensation to eliminate jitter interference. Then, the model free adaptive control method is used to establish a single input single output control system to represent and update the background image in real time. At last, the foreground image of moving object is extracted by background subtraction. The proposed method is compared with the traditional model-based methods in different video sequences. Experimental results show that the proposed method can effectively solve the problem of moving object detection in camera jitter scene and the separated foreground is closer to the ground truth.