复杂场景下自适应视频前景提取算法

针对目前基于背景建模的前景提取算法在复杂场景中误检率高以及鬼影融入背景模型慢等问题,提出一种复杂场景下自适应视频前景提取算法。在前景检测阶段,利用背景模型中样本之间最小欧式距离的均值衡量背景动态波动程度,自适应调整像素点的半径阈值,从而抑制在光线变化,树叶晃动等场景中产生的拖影和噪声点；在更新背景模型阶段,根据物体的运动速度自适应选择一次更新背景模型中样本个数,加快因首帧存在运动目标和物体运动状态变更而产生的鬼影融入背景模型。实验表明,相比其他代表性算法,改进算法在加快鬼影融入背景模型和抑制背景动态干扰方面均有较好的表现,且提升了准确率、召回率,降低了假正率。

Adaptive video foreground extraction algorithm in complex scenes

Aiming at the problems of high false detection rate of the current foreground extraction algorithms based on background modeling in complicated scenes and the slow integration of ghost images into the background model, an adaptive video foreground extraction algorithm for complicated scenes is proposed. In the foreground detection stage, the average value of the minimum Euclidean distance between samples in the background model is used to measure the dynamic fluctuation of the background, and the radius threshold of the pixel is adjusted adaptively to suppress the smear and noise generated in the scene such as light changes and leaf shaking. In the background model stage, the number of samples in the background model is adaptively selected and updated according to the movement speed of the object, so as to speed up the integration of ghost images caused by the change in the state of motion of the target and background objects in the first frame into the background model. Experiments showed that compared with other representative algorithms, the proposed algorithm has better performance in accelerating the integration of ghost images into the background model and suppressing background dynamic interference, and can improve the accuracy and recall rate, and reduce the false positive rate.