Abstract:To overcome the common difficulties emerging in video surveillance

a random sampling background modeling algorithm is presented. Unlike other background algorithm

the random sampling method denotes the background model by using a set of sampling pixels of different time located at a specific position. When deciding whether a new coming pixel belong to background or not

each pixel in the model is compare with the new pixel. If there are enough samples similar to the pixel

it is judged as background pixels. And then

this pixel is used to update the model samples selected out randomly. Thanks to the attributes of random sampling

this algorithm exhibits excellent ability in resisting gradual and sudden illumination changes

dynamic background

camouflage

shadows and video noise. The implementation is completely describes in this paper

and experiments indicate that the random sampling modeling method has higher computing efficiency and percentage of correct classification.