Abstract: In order to avoid cheating examinations by using camera pan and tilt，a method of detecting examination paper is proposed by using image processing and pattern classification techniques under surveillance environment.Firstly，a thresholding pre-selection strategy is used to detect possible paper pixels，and adaptive Gaussian Mixture Models is trained to adapt different illumination and place conditions；then region growing technique is proposed to remove noises；and finally several shape description features are computed to classify the detected regions to obtain the final detection result.Extensive quantitative and qualitative experimental evaluations validate the performance and efficiency of the proposed method.