Design And Implementation Of Embedded Fire Smoke Detection System Based On Image Processing

Fire detection based on image processing is a new technology which gets the video information through machine vison and detects fire by using image processing algorithm. Compared with the traditional detection method based on sensor, it has better accuracy and real-time performance. Early in a fire, smoke appeares earlier than open flame, so fire could be alarmed superearly through detecting smoke. This paper designs an embedded fire smoke detection system base on TMS320DM8168 which completes functions such as video capture, fire smoke detection and network transmission, and it has a certain value with small volume and high degree of integration. The main content includes the following parts:(1) The system general plan is designed. Firstly the CCD camera grabs image and sends it to the DM8168. Then DM8168 completes the smoke detection algorithm and H.264 video coding. Lastly it sends coding data to the PC players via Ethernet transmission service.(2) Fire smoke detection algorithm based on image processing is researched. Firstly, the original image noise is eliminated by median filter method, and compared with three commom ways of moving object segmentation, background subtraction based on Kalman filter is finally chosen for extracting the candidate smoke regions.Then this paper amends the candidate smoke regions by means of morphologic processing. Secondly, analyse some smoke characteristics such as color, convex degree, texture features and extract these features through algorithms. Lastly, these features are fused by a BP neural network for pattern recognition to determine there is smoke or not.(3) The DM8168 software platform is designed and implemented. The system establishs a complete clain on control module of ARM based on MCFW architecture to achieve the control of each module. Capture module is designed to collect the video data under the HDVPSS development directory. Fire smoke detection algorithm is transplanted into the image processing module under the DSP development directory. The coding module is designed to realize the H.264 video encoding under the Video M3 development directory. The function of network transmission is realized by using Live555 streaming media server.This paper verifies the feasibility and effectiveness of the proposed algorithm through many experiments. At the same time, this paper also tests the practical application of the system, the results show that the system has a certain practical value which can identify smoke in video and has some anti-jamming capability.