개선한 분할 알고리즘 및 ADBM 정보를 사용한 비디오 기반 화재검출

Abstract

About 65% of the whole Korea’s land are surrounded by mountains. The majority of whole forests is in general pine forests which is likely to be burnt with ease. Recently, the fire is frequent and getting larger. The fire monitoring application systems can reduce the damage of properties by providing earlier warning for possible fire situation. Most of the fire alarm system using sensor-based detector. But these sensors, after the fire happened, when fire smoke spread to certain degree, can be detector fire. Instead, video-based fire detection system, does not have to wait for the spread of smoke, through the camera can immediately detector fire. Although, used a advantages, there are few related research. In order to overcome the existing problems, In this paper, we proposed a new vision-based fire detection algorithm. First, color-based segmentation using the RGB and YCbCr color space. Second, The absolute value of the successive inter-frame pixel value difference using average for the Average Difference Background Model to generate. Third, segment the image into foreground and background regions using the Average Difference Background Model. Fourth, acquired the degree of change in pixel values using the Texture Filter Functions. Fifth, clean up the raw segmented image using connected-components analysis. This form of analysis takes in a noisy input mask image. It then uses the morphological operation open to shrink areas of small noise to 0 followed by the morphological operation close to rebuild the area of surviving components that was lost in opening. Thereafter, we can find the contours of the segments. The experimental results using real world videos show that our system can indeed improve detection performance compared to previous research.The detection rate achieved was 99.9%.