摘要 |
A sequence of video images is generated of a pavement area of an airport which contains one or more objects. A processor accesses a background model of the pavement area and determines in a current image a single cluster of foreground pixels that is not part of the background model and assigns a first value to each foreground pixel in the cluster to create a foreground mask. The background model is updated by learning new conditions. A convex hull is generated from the foreground mask. A ratio is determined from pixels captured by the complex hull and pixels in the foreground mask. A ratio higher than a threshold value indicates an object not being an airplane and an alert is displayed on a computer display. Images may be thermal images. A surveillance system based on the calculated ratio is disclosed. |