Automatic Surveillance: Vision Detection Using Gaussian Processes Analysis

621 Words 3 Pages
Automatic surveillance application has been a subject of extensive research during last decades to develop robust application. As one of its important application, the vision detection and localization interest in detecting moving objects under the challenging conditions of the illumination change, occlusion, shadow and perturbation of the images sources.
In vision detection system, the techniques used are either background subtraction based or feature detection based.
The background subtraction based techniques interests in detecting variation within the scene across several image frames. This approach is based on comparing the current image with a reference one(s) of the background. Pixels of sharp variations are consequently classified into the foreground. Many techniques have been established to work in the frame of this principle among which we can list: the running average [1] [2] that interest in modelling the background each pixel location by fitting a Gaussian probability density function (Pdf) on the last n pixel’s values. At each frame time, the pixel's value can then be classified as a foreground pixel if the difference between its values and the estimated mean values exceeds a given threshold.
The temporal median filter [3]is proposed to address the issues of the running average by using a median value of the last n frames as the background model provides an adequate background model even after processing a couple of frames.
The Mixture of Gaussians [4]…