Modified RPCA with Hessian matrix for object detection in video surveillance on highways

Video surveillance is an active research topic in computer vision research area to detect the abnormal behavior of vehicle and pedestrian on the highways in order to reduce the collision between them. Statistical methods are helpful in identifying the abnormal behavior of vehicle and human in order to avoid the accident on the highways. To build an effective automatic system that should determine the number of pedestrian and vehicles, if there are any, then their distance and speed needs to be calculated. Detecting object and calculating their speed and distance is challenging task because objects are moving fast on highways, and appear at different scales. In this paper, we propose a Modified RPCA with Hessian matrix for vehicle and pedestrian detection. By using an SVM classifier, it will be able to classify the objects in the current frame. Distance is calculated between the vehicle and pedestrian, speed and their locations. If the distance value is below the defined coverage (50 meters) their performance is evaluated and compared between RPCA and Modified RPCA. The modified RPCA is more efficient than RPCA.