Detection Algorithm Research Of Small Target In Infrared Image Based On Empirical Mode Decomposition

The warning system based on image plays a important role for national security,including infrared image warning system attract more attention because of its goodproperties which can work24-hour day and night, but the key technology in the earlywarning system is small and dim target detection. At the same time, small and dim targetusually have only one or several pixels and easily confused with noise. By intensive studyof many scholars’ domestic and international, many algorithms improve continuously, andmany new methods have been proposed at the same time, which make considerabledevelopment in this field. This article makes an in-depth analysis and discussion ofinfrared small and dim target detection, mainly analysis and discusses the existingproblems and the EMD method in the application of infrared small and dim target, andaccording to the characteristic of EMD theory and experimental results, puts forward anew detection method and algorithm research idea.This article analyzed the detection method of infrared small and dim target imagefrom different angles, and discussed the analysis of the EMD algorithm for small targetimage processing features. The image decomposed by EMD has good inhibition to thebackground, highlighting the small and dim target, combined with the better e ffect basedon image segmentation of adaptive SNR threshold, and carry on the segmentation withdecomposed image, and then achieve the target detection of infrared small and dim targetimages.This article makes a abstract on the target detection algorithm of traditional infraredimage, and makes a series of relevant simulation experiments, comparison the detectioneffect of various algorithm, and analysis the problems of different algorithms. Then I willdetailed introduce the empirical mode decomposition of the signal processing method, andanalyzed in principle the feasibility and superiority for the infrared small target imagemanagement after extended to two-dimensional. This article detail introduces the implementation steps of the algorithm, carried out the simulation experiment, anddiscussed all kinds of interpolation methods by comparing. After that, according to theexperiment results, puts forward a new detection method: carries on the empirical modedecomposition by using median filtering, and improved algorithm simulation experimentbased on it. The improved algorithm promote local signal-to-noise ratio effectively, thisprocess has the obvious effect than before.