Description

In video surveillance applications, the main aim will be identification of the persons i.e., identifying the position of the persons in each frame of the video. Tracking of the targets in the video has many applications in surveillance systems. The specific persons or vehicles in the video can be tracked using multiple kernals tracking process. The persons in the video is tracked all over the video. The positions of the targets in the videos were trained in each frames and the positions of the targets in each frames were identified. Based on the identified target positions the target is tracked all over the frame. There are numerous techniques commonly applied for the identification of the persons, yet identification of the persons in the video remains a problem. A method that uses the multiple kernals tracking method for the identification of selected target position in the video is proposed. The position of the targets that are to be tracked in the video is given as input. The position is leaned in each frame by the use of the prediction and estimation process. In the prediction process the position of the target is predicted by identification of the movements in the specified pixel co-ordinates. The predicted object positions were compared. The movements were detected by means of the employment of particle filtering which denotes the movements in the specified target position. The particle filtering process helps in the generation of the particles and the identification of the movements by analyzing the intensities i.e., the pixel vales in each video frames. The predicted portions were estimated with the help of the trial and error method for the estimation of the target pixel co-ordinates. In the trail and thee error method the pixel values at the particular locations were identified and the positions were compared and if they are same then the predicted portions were chosen otherwise the next pixel positions were compared. The estimated target co-ordinates were then again estimated and if the target co-ordinates change the position is estimated. Thus for the whole video frames the process is employed and the target position is predicted. Thus we can keep in track of the needed persons all over the video frames. The predicted video portions were then denoted by a rectangle to denote the location. The performance of the process is measured by means of measuring the accuracy and error rate. The measured performance of the process shows that the accuracy of the proposed target tracking system is improved compared to other existing methods for target tracking which is due to the employment of the learning method based on multiple kernals method.