Study Of Video Smoke Detection Based On Motion Block Of Track

Video fire detection has the advantages of fast response, non-contact, strong anti-interference, etc., and is easy to combine with the security system to realize the integration of the security and fire monitoring, which is a hot research direction in the field of fire monitoring in recent years. Video fire detection can be divided into the flame detection and smoke detection. When the fire breaks out the smoke is produced firstly before the visible flame and the fog is not like the flame which can be blocked easily, so the video smoke detection can detect fire earlier.Based on the summary of existing video smoke detection methods, this paper proposed a video smoke detection method based on motion block of tracking. The smoke movement characteristics, the concentration features and gray consistency of the image block were studied and then these three characteristics of smoke image block were combined to recognize suspicious smoke image block.First of all, this paper summarizes the commonly moving object detection methods, and compares and analyzes the advantages, disadvantages and applicable scope of the finite difference method, background subtraction and optical flow method. Combining with the need of this algorithm, it uses Gaussian mixture model to set up background model and extract the motion area. After the preprocessing operations such as filtering de-noising, the extracted motion regions are divided into image blocks, which will be screened preliminarily based on the gray distribution of smoke.Analyze the smoke video and then extract the smoke movement characteristics, the concentration features and gray consistency features. These features include static and dynamic characteristics of smoke. In the process of the smoke movement it presents obvious regular distribution in the direction of motion, mainly reflecting in upward movement trend. Since a certain concentration is a essential characteristics of fog, the paper analyzes the relationship between the concentration of smoke, background image, the current frame image and the environmental parameters, and gives solving method of the smoke image blocks’light transmittance and reduction coefficient, which is a characteristics used to distinguish smoke image block. By analyzing the changes of the smoke density in the movement, it was found that there was consistent trend in the gray-scale variation in the video image during the movement of smoke. In this paper this feature was extracted and the description was given. Combining these three characteristics of smoke image block it can distinguish whether the various suspicious image blocks extracted are smoke image blocks or not, so as to complete the smoke detection.Finally the video smoke detection algorithm proposed in this paper was tested by using different scenarios of smoke videos and interference videos and the test results showed that the proposed algorithm could effectively identify smoke image blocks accurately, and complete the video smoke detection.