| 摘要 |
PURPOSE: A visual surveillance system and a method for detecting an object of the visual surveillance system are provided to adopt a layer concept to each mode in a Gaussian mixture model, and update information of layers of each mode, thereby effectively detecting an object, particularly an object in a dynamically moving background or an object being at a standstill after moving for a while, as well as utilizing existing advantages of the Gaussian mixture model. CONSTITUTION: A method for detecting an object by a video surveillance system comprises the steps of: generating a histogram presenting a brightness distribution of pixels which is obtained by accumulating a surveillance target image acquired by a camera for a predetermined time (S300); modeling a time accumulated histogram by mixing a plurality of Gaussian modes (S400); classifying layers of each mode of in the Gaussian mixture model (S500); updating information of the layers if detecting a change from external environments or a motion of an object (S600); and detecting the object from the surveillance target image (S700). [Reference numerals] (AA) Start; (BB) End; (S100) Take an image using a camera; (S200) Create a surveillance target image; (S300) Create a brightness distribution histogram for pixels accumulated for a certain period of time; (S400) Model the accumulative distribution using a gaussian mixing model; (S500) Classify layers of each mode of in the Gaussian mixture model; (S600) Update information of the layers; (S700) Detect objects from the surveillance target image |
