| With the development of space utilization and exploration, many space targets, such as satellites, space shuttles, space stations, space debris are concerned by many countries and researchers. It is a commonly used technique to observe and survey space targets using large aperture ground based tracking and imaging telescope (GBIT). A technical method of GBIT is"infrared imaging" which has the advantage of all-day working and is highly concerned. The direct purpose of space surveillance is to get rich features of space target:orbital characteristics, photometric characteristics, radiation characteristics, spectral characteristics, polarization characteristics and imaging features. As for the imaging characteristics of the target, due to the resolution and detection capabilities of GBIT, targets are generally regarded as points, simple surfaces or simple ellipsoid. With the diversification of demand for space observation mission, in order to promote the research in this field, this thesis will focus on three-dimensional (3D) reconstruction theory of infrared images of the space targets, which is meaningful for the analysis of space target imaging characteristics.Three-dimensional reconstruction is a classic problem in computer vision (CV) field,in which SFS (Shape from shading) three-dimensional reconstruction algorithm was proposed in1970by MIT’Horn for3D surface reconstruction of NASA’s lunar images. In this thesis, we extend SFS algorithm to the space infrared imaging field, and the new algorithm is named IR-SFS (infrared-SFS),which means infrared image three-dimensional reconstruction algorithm. According to SFS theory, there are cues of3D information in the change of image gray value. The traditional SFS algorithm assumes that the light source is a point source or parallel optical infinity, but infrared target has the external light source and self-radiation.The changes of gray value in the infrared image are not only cues for3D information but also temperature information. From information retrieval in the open literature, it has not yet been found that SFS algorithm is applied to infrared images (including space target infrared imaging), and there is no direct useful theoretical model for our research.Therefore, we need to work on theoretical models and the specific implementation method when we try to extend the SFS three-dimensional reconstruction algorithm to space infrared target.The traditional SFS algorithm is based on non-contact3D measurement method, single grayscale image without camera calibration and priori information of the target. The main research contents of SFS include:building radiation model, illumination and reflectivity estimation, underdetermined PDE equations solution, as well as integration with other three-dimensional reconstruction methods. The core theory idea of SFS is the theory of electromagnetic radiation and photometry. The radiation model associates the lighte source, radiance with geometry feature of objects together. Solving the equations of radiation can obtain the surface shape of the object. SFS radiation model is more simplified, compared with the real presence of radiation model, such as light source position and intensity estimates, reflectivity estimates, solving underdetermined PDE equations, smoothness constraint of object shape, and ignoring the sensor imaging process model, and so on. As a result, SFS algorithm have high adaptability and computing speed, but accuracy is greatly reduced, and a complete system of theory and application has not yet formed after extensive study for decades.Firstly, we try to establish radiation model of space target surface, then research the positive and negative effects of self-radiation to the three-dimensional reconstruction with detailed theoretical analysis and simulation. Secondly, the geometric and photometric parameters of the imaging process are detailedly calculated to avoid errors caused by rough estimate. Finally, design detailed IR-SFS algorithm flow, and the simulation and application results show that the algorithm improves the accuracy of three-dimensional reconstruction. The IR-SFS algorithm specific research covers: coordinate transformation between Local coordinate and Target coordinate, detailed analysis and calculation of the position and intensity of the external radiation, self-radiation intensity calculation, the analysis of the influence for3D reconstruction from self-radiation; based on the above theoretical analysis and simulation results, the concept of temperature field of infrared image is proposed, then estimate temperature field shape using DCT(discrete cosine transform). Obtain the infrared residual image by self-radiation energy to external radiation energy ratio, and then the three-dimensional shape can be recovered. A series of analyzes and calculations make IR-SFS algorithm more complex than SFS algorithm, but the results have a higher SNR, which demonstrate the effectiveness of the algorithm.The research of this thesis is to promote the image features analysis of space targets. Reconstructed3D model is in favor of subsequent information extraction and analysis, such as the space target recognition (SOI) based on3D model, the target operating state judgment, the space target posture measurement, and other geometric parameter measurement and calculation of the target body parts. |
