Forest Fire Recognition Based On Infrared Image Processing Key Technology Research

Forest fire has been a major problem plagued the protection of forest resources,most early fire detection methods are based on smoke and temperature features of thefire, it leads the judgement too simple and the distorting rate is high. With thedevelopment of computer technology and the invention of the infrared detector, a fireprotection technology that based on infrared image processing began to be used byhuman, the main work of the technology is to process the infrared images of the forestwhich are from infrared detectors and give an opinion whether any fire appears.The main work of this thesis is to research the infrared image enhancement andsegmentation algorithm, by the description of the various algorithms and simulationresults of the experiments, an algorithm that more suitable for forest backgroundinfrared image processing is proposed. Except above work, an effective fire criteria isreached according to the identification of fire on the processed image.Through researching and simulating the classic infrared image airspaceenhancement algorithms prove that a little effect they can be made in removing theforest background infrared image’s noise, and the image will be blurred at the same time,so they not suit to process the forest background infrared image which includes muchnoise and has a bad definition.After survey of the commonly used infrared image segmentation algorithms, itshowed that the threshold-based infrared image segmentation algorithms depend on thethreshold too much, that can lead to incomplete segmentation and over segmentation.The Watershed algorithm has a better segmentation result, but it is hard to realize andeasily goes to over segmentation. When those algorithms are applied to the forestbackground infrared image processing, it may make the follow-up firework’sidentification more difficult.In order to resolve the above questions, a lifting wavelet transform that based onfrequency domain processing and combine with histogram correction to enhance theforest background infrared images is proposed, and then a modified region growing method is used to the suspicious area in the image segmentation and identify whetherthere is any fire in the area, according to the comparison of recognition results ininfrared image sequences of the fire group and interference group obtained a accuratefire criterion.The experiments prove that the algorithm proposed in this thesis makes the forestbackground infrared image’s PSNR improved nearly6dB than the commonly usedairspace algorithm, the contrast of the image both improved certainly. The improvedregional growing algorithm is easy to realize and its segmentation results is morecomplete than its traditional algorithm, it has a good practical. The fire identify criteriaat the end of the thesis can play a reference role in other people’s work.