【摘要】 针对如何从包含大量冗余信息的视频中快速检测目标的问题,提出了一种基于统计分析的目标检测方法。该方法采用改进的直方图均衡化算法对图像做预处理;通过曼哈顿距离计算图像帧之间的差值,并对差值做进一步处理;采用迭代的方法,从图像帧差值中求取阈值,利用阈值判断前景帧和背景帧;在背景帧基础上建立背景模型,通过卡方值判断前景点和背景点;最后利用形态学还原物体真实形状,实现目标的准确检测。实验表明,该方法能快速准确地检测目标,可应用于视频监控的目标检测。更多还原

【Abstract】 Object detection method based on statistical analysis is proposed for detecting objects from the video which contains a lot of redundant information.The method first employs an improved histogram equalization algorithm to process images.Then we compute difference between frames using manhatton distance,and distinguish background and foreground frames using a threshold,which is computed from the differences of frames by iteration method.We establish the background model from the background frames,distinguish background and foreground points according to Chi-square value.At last,we restore the true shape of objects using morphological algorithm.Experiments show that the proposed method,which can be applied to video surveillance,can quickly and accurately detect objects.更多还原