一种基于混合高斯模型的运动目标检测新算法

Moving Target Detection Algorithm Based on Gaussian Mixture Model

针对传统的混合高斯模型算法检测时存在较多噪声的不足,提出了一种基于混合高斯模型与三帧差法相结合的运动目标检测算法。该算法首先分别通过混合高斯模型算法和三帧差法各获得一幅前景图像,然后将两幅图像进行逻辑或运算;再对获得的图像进行自适应中值滤波处理,并利用连通域阈值面积去噪法除去剩余的噪声,最终提取出运动目标。实验仿真结果表明,利用该方法可以有效地提高运动目标的检测率。

Aiming at the shortness of the existing noise in the traditional Gaussian mixture model algorithm, an algorithm for moving target detection based on Gaussian mixture model and three-frame-differencing is proposed. First, it gets two foreground images by Gaussian Mixture Model algorithm and three-frame-differencing algorithm. Then, the foreground image is extracted by applying logic OR operation on the two previous obtained foreground im- ages. Adaptive median filtering on the previous foreground image is applied and used connected domain threshold denoising area to remove the rest of noise. Finally, the moving objection is extracted. The experiment shows that u- sing this algorithm can improve the detection rate of moving objection effectively.