Research On The Technology Of Multispectral Targets Detection And Recognition

Since it could effectively solve the problem of detecting moving dim target in deepspace, target detection based on multi-spectrum plays a significant role in spacesurveillance techniques. Multispectral target detection technique has been widelyapplied in the field such as deep space detection, telemetry, contamination detectionand food security, it promotes the development of national defence industry andnational economy. Based on multispectral imaging, the thesis lays emphasis on theresearch about moving target detection algorithm in sequence star charts, the mainmissions are as follows:Analyze the chromatic dispersion principle of several kinds of spectral imager andlist the superiority of grism. Via Fraunhofer formula, analyze the spectral amplitudeand interval between extreme value by grism imaging. For the acquired multispectralstar image, the thesis adopts8area connected domain analysis method to obtain thestar target and corresponding spectrum stripe. Analyze and contract the classificationmethods of multispectral data, reach a conclusion that the method that supports vectormachine possesses a high level precision.According to the analysis of target properties and background parameters inmultispectral astronomy images, the background gray can be considered to obeyGaussian distribution, the energy distribution of target asterion can beapproximately expressed by Gaussian point spread function, based on the aboveresearch,the thesis establishes reasonable asterion image model.Proceed the detection algorithm’s image preprocessing missions: adopt localstatistical filtering method and adaptive threshold value method to conduct denoisingand false alert elimination respectively. While researching the target centroidextraction and location algorithm, in order to resist the noise influence, the thesisselects second cross projection method and square weighted qualitative method.Comparing the simulation start charts and real one, we can conclude that the methodpossesses a high level accuracy. When researching the star charts registration, the thesis adopts a method based on the feature points include the stars’ interval. Whenresearching the star charts fusion, the thesis proposes a method based on the intensity,hue and saturation (IHS) space and improved wavelet transform.According to the theoretical foundation of moving dim target detection algorithmbased on the maximum projection, the thesis adopts an improved detection algorithm:on the basis of maximum projection, the algorithm utilizes some supplementaryinformation such as track, angle and time. According to the analysis of experimentaldata and algorithm performance, this algorithm which is suitable for various spacescene possesses a high level accuracy and instantaneity.