Abstract
Most of the existing image surveillance systems are provided with limited function of displaying the monitoring area's situation and therefore cannot provide any intelligent ability of detecting intruders. In order to enhance the ability of the existing image surveillance systems, this resarch proposes an intelligent image surveillance system using digital images processing techniques. This system can detect any intruder emerging in the monitoring area by background differencing and using the morphology filter to remove noises. The edge points of the intruder's contour are detected by applying Laplacian operator. The edge points are further processed with scale space filter to remove the toothed noises. Once the corner points of the contour are detected, the least square error method is then applied to approximate the points around the cormer point into a line and to estimate the slope of the line. Finally, a set of the most similar corner points is found by calculating the fuzzy relation of the slope. By using this set of corner points, it can automatically judge whether the two intruders are identical or not. The moving direction and the displacement of the intruder, if any, can also be calculated. The whole system is composed of four software modules, ie., background differencing preprocessing module, intruder detecting module, intruder matching module, and intruder's moving parameters calculating module. The experimental results show that the system can correctly detect any intruder in the indoor environment. It can also trace the motion of the intruder by comparing two consecutive images.