Embedded Smoke Monitoring And Recognition System

Fire is one of the worst disasters which causes a large number of economic losses and casualties.In order to prepare against fire,smoke sensor has been used in many public places.However,the sensor fails to detect fire outdoors or in the large area due to scattering of the substance emitted from the fires.With the evolution of image processing technology and the embedded system calculation accelerating. Many scientists strive for recognizing fire with video surveillance system to give alarm. As the fire produces large mounts of smoke in the early stages to give early warning before the fire grows larger.In this paper, the objective is to create a system which can be able to detect smoke using video frame.We combine embedded system and image processing technology to detect smoke.Initially, we detect the motion in video frame to define the candidate smoke region and reduce the mount of area needed to analyze as well.Subsequently, we extract the color information from the candidate region and compare with the smoke color decision model.In addition,the growing area and irregular contours in smok region are good features to exclude the smoke colored moving objects.To increase the robustness of the smoke detection system, we use an algorithm to detect smoke and fire cuz the smoke can not be detected at night.The hardware of the embedded system in this paper is Dev Kit8500 D based on TI processor which integrates ARM Cortex-A8 kernel at 1GHz with high-level digital signal processing functions. The operation system we adopt is Linux.Finally,we port the algorithm to ARM board and it is implemented to detect smoke and fire.