Analysis Of Typical Object Infrared Spectrum Characteristics And Study On Infrared Image Preprocessing

Nowadays, Development requirement of whole day geographical space informationguarantee is increasing， infrared detector technology is becoming more and more crucial amongall image acquisition methods for its all-day detecting ability. Based on study actuality of typicalobject radiation features and other correlative technologies, the paper introduces object infraredradiation temperature distributing model, and then makes use of objects spectrum surveying data,analyzes typical object spectrum characteristics, at last, the article shows infrared imagepreprocessing, includes non uniformity corrects and image enhancement. The paper’s mainstudying contents are as follows:1. Based on typical object infrared radiation physical foundation, the paper shows theprocedure of creating bare ground surface Heat Radiation Temperature Distributing Model andmodel solving, and using finite difference way to calculate ground surface temperature andradiation, both using Monte-Carlo to simulate temperature distribution. At last, validating modelvalidity is made by the method of comparing surveying data and theoretic data.2. Field Spec3Spectrometer is used to surveying object spectrum. The article mainlydiscusses object spectrum curve characteristics about different object types, surveying time,object position. Then analyzes and summarizes effect factors of object spectrum characteristics.3. A BP neural network method for correcting the no uniformity of an infrared focal planearray (IRFPA） based on median filtering and weight changing is proposed. The result showsthat this algorithm owns its advantage of higher correction precision, rapider convergent speedand hold edge details.4. The article shows infrared image enhancement by the methods of histogramequalization, wavelet transform, Multi-Scale Retinex and maximal fuzzy entropy theory.compared and analyzed the adaptability of these algorithms and process results, it shows thatthe method of enhancing image based on Multi-Scale Retinex and maximal fuzzy entropy ismore suitable for low contrast enhancement processing than others.