一种改进的ViBe目标检测算法

现有ViBe目标检测算法存在误检率高，得到的结果噪声较大的问题。为此本文提出了一种将判定阈值自适应规则与优化结果反馈机制相结合的改进算法。该算法首先使用样本自身统计值作为阈值进行目标检测；接着对检测结果进行去噪、形态学处理等优化处理；最后将优化后的检测结果反馈给背景模型更新过程使检测更精确。与传统的ViBe算法相比，本文算法可更精确的提取出运动目标，误判率低，过检率较原始算法降低10%-20%。实验结果证明了该算法的有效性。

An improved ViBe Object Detection Algorithm

The misjudge rate of the existing ViBe algorithm is high and the noise in the result is large. To solve the problem an improved algorithm combining the threshold adaptation and the feedback of improved result is proposed in this paper. First, the statistics of the samples are used as threshold for object detection. Second, noise reduction and morphological processing etc are used on the result. Last, the improved target detection results are used as feedback to background update process. In this way, the result of detection is more accurate. Compared with traditional ViBe algorithm, the proposed algorithm can extract objects more effectively and has lower misjudge rate which is 10% to 20% lower than traditional ViBe algorithm. Experimental results show that the proposed algorithm outperforms the traditional ViBe algorithm.