Person Identifying in Video Surveillance

Abstract

The purpose of this research is to introduce a method to identify a person from a video surveillance. CCTV surveillance systems have been widely used in many places to address security issues. Those systems record images 24/7 and are monitored at strategic times in a local monitoring suite. Continuous and careful attention is required to recognize the activities of a large crowd. Considering large premises, more cameras and operators are required to handle the situation and also the number of monitoring are also increased. There is no proper mechanism to search and retrieve important information whenever it is required. Moreover it very costly to maintain such systems in public places and also specific staff have train to maintain these systems in order to retrieve important information. This research introduces a method which is capable of addressing those issues using different image processing algorithms for CCTV videos. Analyzing video footages for detecting and recognizing people is an active area of research aiming at fulfilling this goal. The proposed detecting people approach needs static background. By using image processing techniques and algorithms subtract the static background from the objects to detect moving objects and generate a binary image for a video frame. Then track the detected objects radiuses which are greater than to a specific threshold value in binary image. By subtracting original color image with binary image detect people in that frame without background. Then the features extractions happen for those interesting objects. The first feature extraction is doing to select an individual from the CCTV video frame. This can be done by marking on person object in a video frame, which is useful to get the visible timeline in the video for this person. The second feature extraction is doing to each and every person object in a video frame and also every frame for the video. The proposed feature extraction method generates color histograms for upper part, below part and full part of the body for R, G and B channels. By using those channels values calculate the Mean, Mode and Standard Deviation color variation for R, G and B channels. After fulfilling of extracted objects features, both feature extracting modules has been compare to get the person visible timeline of the CCTV video. Results of the evaluation show that the proposed mechanism has given acceptable precision level.