Camera Response Model Based Moving Detection under Auto Exposure

To deal with the problem of the dynamic change of the background caused by camera automatic exposure, a camera response model based moving detection algorithm is proposed in the paper. In the initialization stage, through design of energy function based on penalty term and screening of mass data, the response function of the camera is solved. In the on-line detection stage, firstly, by modeling the current frame and background frame to a single Gaussian distribution which contains outliers, Wiener filtering is used to obtain real-time exposure coefficient ratio between current frame and background frame. Then, when the automatic exposure is detected, the background reference frame which is consistent with exposure coefficient of current frame is obtained by response model of the camera, thus wrong detection of motion caused by automatic exposure is eliminated. Experimental results demonstrate that the proposed algorithm stratifies the demands of real-time processing, and it eliminates better the false motion detection caused by automatic exposure compared with the classical method.