The Research On Forest Fire Detection Technology Based On YCbCr Color Space

Traditional forest fire detection system using physical sensor devices to identify the fire, it send out alarm signals according to the received signals such as the chemical and physical properties in air, there is some problems such as reaction lag and poor reliability. With the development of image processing technology, forest fire detection based on image technology becomes a research hotspot in recent years. In this pepper, the research status of forest fire detection based on digital image are analyzed, summarized the framework of corresponding algorithm, and the fire flame segmentation, feature extraction and recognition were studied also in this paper.Firstly, through the relevant literature, the advantages and disadvantages of forest fire detection algorithm based on digital image processing are researched and analyzed. The advantages are such as through the YCbCr color space which can be better to distinguish luminance and chrominance, under the condition of clustering algorithm is more noticeable to the flame can precisely extract the fire area, and so on. Also, the shortcomings are such as low recognition rate when in strong light or under the interference of strong fog, used for the analysis of the scene are not comprehensive, and so on.Secondly, an new segmention algorithm of forest fire flame is proposed which is based on color space. After comparison test from several different segmentation methods, this paper proposed a fire recognition algorithm based on image to reduce the rate of positive error. There are several steps of this method. At first, the image color space was transformed from RGB to YcbCr. Then set up a coordinate system of the Cb and Cr values of the fire sample pixels, and the flame pixel values just focused on an elliptical area. At last, coupled with k-means clustering algorithm based on L*a*b*color space to segment the fire flame, and extracted the flame area whether extraction area has the flame dynamic features. This new method enhanced the fire accuracy of judgment. Research statistics found that the Cb’s and Cr’s distribution probably close to normal distribution, and the distribution in YCbCr color space is also in line with two dimensional normal distribution. Through the actual test, the segmentation effect is good. Thirdly, the forest fire image flame characteristics are analyzed, mainly analyzed the two aspects of the flame color feature and shape feature. For the flame color features, made the method of circulation area division to analyze it; for the flame shape feature, the split image should binarization processed before it was extracted, and then analyzed its features such as perimeter, area, and the circular degree. At last, combined the various characteristics, to ensure the flame high detection rate and reduce false detection rate.Finally, using C++code to realize algorithm, and carried on the simulation of forest fire flame recognition. The experimental results show that the proposed identification method can accurately identify the forest flame, and misjudgment rate is low. The method also ruled out the interference of approximate flame color objects and high light. For the forest fire monitoring and detection, it has good practical application value.