The time adaptive background extraction and updating algorithm based on clustering

Extraction and real-time updates of the background image is the premise of the moving target detection. This article proposes an algorithm based on clustering time adaptive background modelling. It is removed a pixel value in the same position from the N-frame images, and found the mean and variance, then removed the farthest pixel value until the remaining pixel variance is less than a specific threshold. The mean is the background value of the pixel. It can effectively inhibit a variety of errors and can use few frames to extract the background to save time. It can also automatically predict the next background update time.