Abstract:

Aiming at dynamic background of camera of camera shake and leaf sway in ViBe algorithm, an improved ViBe algorithm is proposed with high false detection rate and low accuracy, which firstly selects multiple frames by using Canny-based three-frame difference improvement algorithm for background modeling. The background model is updatecd, and the adjustment factor is set according to the background complexity, and then the adjustment threshold and the background model update rate are adapted to dynamic background detection. In order to improve target detection integrity, the three-frame difference algorithm is combined with morphological processing. Experimental results show that the proposed algorithm improves accuracy and integrity of target detection under complex background of branch shaking and camera shake.

Key words: motion object detection, ViBe algorithm, three-frame difference method, Canny, self-adaptive threshold

摘要：

针对ViBe算法在相机抖动和树叶晃动的动态背景下，出现的误检率高和准确度低的问题，提出一种改进的ViBe算法。该算法选取多帧使用基于Canny的三帧差分改进算法进行背景建模；在背景模型更新时根据背景复杂程度设置调整因子，调整阈值和背景模型更新率适应动态背景的检测；为提高检测目标的完整性，改进的ViBe算法得到的前景目标，与三帧差分算法结合并且进行形态学处理完成对运动目标的提取。实验结果表明，改进的算法在树枝晃动、相机抖动的复杂背景下，检测目标的准确度和完整性提高了。

关键词: 运动目标检测, ViBe算法, 三帧差分法, Canny, 自适应阈值