基于改进的混合高斯背景模型的运动目标检测

Moving object detection based on improved Gaussian mixture background model

混合高斯模型在应对背景中存在扰动的情况具有优势,而其不足之处主要表现在对光线变化比较敏感和当场景中前景与背景之间发生转换时容易产生较长时间的虚影。针对上述问题,提出一种融合相邻帧差法和背景减法的算法。采用了循环周期和动态更新相结合的背景重建机制,通过运用Matlab对视频图像某个像素点的S值和V值的变化情况分析来体现背景更新和重建的过程,并对背景变化前后分别采用传统算法和改进算法进行对比分析。该改进算法解决了背景模型对光线变化敏感以及容易产生虚影等问题,实验结果表明了算法的有效性和鲁棒性。

Gaussian mixture model has a clear advantage of dealing with the existence of the background disturbance, but it has some disadvantages, inciuding sensitivity to the light change and long-time ghost in the case of the change between foreground and background. For these problems, an algorithm is proposed, which combines adjacent frame difference method with background subtraction method, and a background rebuilding mechanism that combines cycle counter with dynamic updating is adopted. The change situation of S and V value of some pixel in video images are analyzed to show the process of background updating and reconstruction, then we employ the traditional algorithm and improved algorithm in changed background and unchanged background respectively, and the results are analyzed and compared. The experiment proves that these improvements can solve the problems including sensitivity of light change and ghost, the validity and robust are also tested.