The Study On The Method Of Recognizing Forest Fire Images Based On SIFT Feature Matching Algorithm

As forest fires occur frequently in our country, the area of forest is reducing year byyear. As a result, forest resources are becoming more and more valuable. As traditionaldetection technology of forest fire has failed to meet the needs of modernization.Therefore, by studying the traditional fire detection and the objective regular of forestfire, the author analyzed the advantages of SIFT features which have the function ofscaling, rotating, maintaining the brightness of images, and stability in changingperspective, transforming affine and noise in a certain degree so as to propose theintelligent detecting system of forest fire based on image processing. The method is ableto overcome the deficiency of the traditional detecting system effectively, and eliminatethe difficulty of large scale fire warning.Based mainly on the systemic hardware platform, this subject adopted the SIFTalgorithm to process detecting images, according to the analysis of the characteristics,judgments, identification of the fire in the forest fire images, to realize the intelligentdetecting of forest fire. The author’s main tasks done in this subject are:1）analyzedSIFT algorithm and its process, and under the environment of Matlab, programmed andtested the algorithm, which can work out the number of feature point of images, andmanaged to match two images;2）conducted ignition experiments for three times in thewoods on campus, and took many pictures to provide the experiment with adequateanalyzing and researching materials;3）based on the main research on SIFT algorithm,complemented several other algorithms, such as the Algorithms of Image ContourMonitoring and HIS Normalization Model Checking Algorithm, Fire ImageRecognition Algorithm of Neural Network, and the RGB Channel Color CumulativeExperiment, which can overcome the deficiency of the single recognition method, andimprove the accuracy of identifying to a certain extent;4）programmed and realized theImage Contour Detection and RGB Color Channel Points Accumulated Experiment, andthe results of the experiment are almost consistent with expected.In this subject, the SIFT algorithm, as the core, is applied to conduct continuousprocessing of the images, combining with some other auxiliary recognition algorithms, which have basically achieved the goal of intelligent identify of forest fire. In addition,proceeded from the reality, the relevant assumptions put forward in this subject havebeen examined, and the results are agreed with expected.