Based On The Infrared Detection Technology Of Forest Fire Prevention Technology Of Infrared Image Processing Platform

Forest fires produce toxic smoke and dust, causing the greenhouse effect, leadingto destruction of the atmospheric environment and at the same time posing a threat tohuman being; In the traditional sense, using a single criterion to detect and identifyforest fire, which is difficult to overcome the impact of the harsh environment and light,resulting in false positives and false negatives of the fire; Infrared image does notrequire visible light, which will help the early detection and prevention of forest fires.Based on the above, this thesis is to processe the infrared images of the forest fires,including segmentation of suspected area and the fire identification.The main work of this thesis is as follows:1. In the aspects of fire image preprocessing, this thesis uses histogramequalization and median filtering to enhance the contrast of infrared image and toremove interference noise, so that it can be easy to the subsequent processing of flamesegmentation and recognition.2. In the aspects of segmentation of the infrared-fire-image, this thesis makes acomparative discussion of some algorithms including edge detection, the gray levelthreshold segmentation, maximum between-class variance method(OSTU) and themaximum entropy image segmentation. Analyzing the characteristics of thosealgorithms when do the work of extracting the suspected region of flame. At last thisthesis adopts maximum entropy to segment the suspected area of flame.3. In the respect of flame identification, this thesis utilizes flame color, frequency,centroid and roundness to make a comprehensive identification. one characteristic canexclude a certain type of interference. Finally, using weight fuses multi-featuredetermine the occurrence of forest fire.4. To make a judgment of R value in RGB color space on the suspected region ofthe pseudo-color infrared image after the subsequent processing, the pixels which meetthe requirement will be used as the flame identification data, and those do not meet therequirement will be directly removed. Experiments show that this thesis can save9.45%of the total program executiontime, reducing the time of flame identification and improving the judgment speed.