Research On The Forming Mechanism Of Multi-band Infrared Images’ Difference Features

Multi-band infrared imaging is an important mean of monitoring fire, eliminating faults,controlling industry, exploring deep space, and it is necessary to target recognition and decision making that fusing multi-band infrared dynamic scene images with unpredictable differences features. The effect of image fusion is determined in the effective fused degree of different information among the original images with selected fusion method, and it is formed in the imaging process that the different information and the respective inherent nature among the original images, and the differences and the respective inherent nature are determined together by the various factors of whole imaging process. By constructing the mapping between the imaging difference characteristics and image difference features, fusion algorithm and imaging characteristics will linked closely, and the fusion algorithm will corresponding changes with the changing imaging, so as to get more accurate and reliable results. Therefore, this paper made a research on forming mechanism of multi-band infrared images’ difference features, and the main contents are as follows:(1) Multi-band infrared imaging difference characteristics were explored deeply under different imaging conditions. Under night vision conditions, compared imaging characteristics difference of target’s apparent radiation, atmospheric transmission, optical system transmission and detector response based on different infrared waveband, and focus on exploring the causes of forming differences of point target pixel gray value of different waveband infrared image in the same scene; under the daytime conditions, deeply analysis that rule of short waveband infrared point targets reflect solar radiation with the variation of time based on short waveband infrared imaging mechanism, and constructed the theoretical models of mid- and long- waveband infrared point targets reflect environmental radiation, and laid the foundation for constructing the effective multi-band infrared imaging difference characteristic sets.(2) A three levels of difference features selection model was put forward. The original feature space composed of statistical features and texture features from different infrared waveband images, through the three levels of difference features selection of consistent difference level, significant difference level and to related difference level, get the comprehensive difference features sets. In this way, it is laid the foundation of the effective multi-band infrared images difference feature sets, at the same time, provided a new method for dimensionality reduction.(3) The mapping relationship between the multi-band infrared imaging difference characteristic sets and the image difference features sets was constructed. This paper derived each of the evolution models of imaging factors in the process of infrared imaging based on the method of separating variables, and obtained the approximate function expression between target pixel gray values with each imaging factors, and simulated the changing curves between the two, so as to comprehensive describe the mapping relationship between the imaging difference characteristic and image difference features.