A New Algorithm Designing for Detection of Moving Objects in Video

摘要

Now a day's video motion detection is implemented targeting a wide class of applications such as in autonomous video surveillance strategies for security and vision analysis, detecting human presence in destructed environments, etc. This paper proposes a method for detection of moving objects in highly secured environments where it can be deployed either on a robotic vehicle or at a static permanent position. The robot acquires information about its surroundings through a camera mounted on it in real time. Another objective of this paper is to increase the efficiency of moving objects detection in offline and online video processing mode. In offline mode, an AVI file is read and it is decomposed into frames. Noise removal is done to improve the image quality and segmentation is performed to detect the moving objects in the foreground by separating it from a known static background. Various operations are carried out and the moving object is identified by marking a rectangular box around the detected object in each frame. When a movement is spotted, alarm is activated. The distance between the centroid of the object in the video file is found and thus the velocity of the movement is determined. In online mode, by comparing each and every frame the presence of moving object is checked. When a prowler is detected, the proposed algorithm triggers the alarm. At that instant, the snapshot of the object is generated and from this, the distance to the object is identified. The proposed algorithm is tested with input AVI format video file of 320 x 240 frame size and frame rate 15fps in offline mode. In online mode, video file of 640 x 480 frame size and frame rate 25fps is captured in real time using a webcam. MATLAB is used for the execution of motion detector algorithms. The result is demonstrated in a different real sequence and analysis of algorithm on the basis of its performance is evaluated.