基于改进Sigma-Delta滤波的复杂场景背景估计

背景估计是运动目标检测一项重要的前期工作,在城市交通等复杂场景中,存在大量慢速或暂停运动目标,背景模型很快受到污染,需要进行较多的后续处理或者采用高复杂度算法来检测前景。针对该问题,提出基于Sigma-Delta滤波改进的背景估计算法,融合可选择性背景更新机制和多频Sigma-Delta滤波背景估计方法,处理复杂场景中不同运动目标的运动特征,以获取稳定的背景。通过对典型城市路段和交叉口复杂交通场景序列进行对比实验,结果表明,该算法在保持Sigma-Delta滤波低内存消耗和高计算效率的基础上可获得更好的检测效果。

Complex Scenes Background Estimation Based on Improved Sigma-Delta Filtering

Background estimation is an important preparatory work for moving object detection. In complex scenes,such as urban traffic,the background model is easily contaminated by a number of slow-moving or temporarily stopped moving object,and many subsequent processing steps or higher computational cost algorithms are needed to detect the foreground. To solve this problem,this paper proposes a background estimation algorithm based on the improved SigmaDelta filtering,which is intended to achieve a more stable background model by combining a selective background updating mechanism with multiple-frequency Sigma-Delta background estimation method to deal with different object motion characteristics in complex scenes. The results of comparative experiment on complex traffic scenes sequences of typical urban road and intersection show that the proposed algorithm achieves better detection effects with keeping SigmaDelta filtering high efficiency and low consumption performance.