Automatic Registration Between Remote Sensing Image And Vector Map

The rapid progress of photogrammetry and remote sensing makes digital image acquisition of high resolution, accuracy, multi-temporal resolution, multi-spectrum and real time images coming to truth, but the corresponding theory and methodology of processing can not catch up with the rapid development of data acquisition. There is a sharp contrast between mass of remote sensing imagery and the theory and methodology of data processing. The automatic acquisition of the spatial and attributive information of the ground object from remote sensing images is one of the important research areas in photogrammetry and remote sensing, which is also main data source to geography information system (GIS). Quickly collecting and automatically updating geographic data have become a difficult task in the areas of photogrammetry, remote sensing and GIS. At present, many countries have accomplished the schedule of topographical mapping around the countries with the basic scale. The need of making map revision arises due to the addition or removal of terrain features. The rapid changes in the map contents and the need for up-to-date maps have compelled surveyors to focus their attention to the development of faster and more economical map revision processes. Photogrammetric techniques and satellite techniques are main means for the map revision, in which remote sensing imagery are effective and economical data source, especially for high-resolution satellite image. In the processing the captured image, the first step is calculating the parameters of image position and orientation, based that the image is registered to the map. The precision of map revision is determined by the precision of the geographic registration. Up to now the conjugate points are measured manually for registration between the map and the image because of the difference between the images and corresponding maps. The manual measure of the conjugate points is a time-consuming and tedious task and the precision of the conjugate points is low. This has become a bottle-neck to map revision and the automatic Aero-triangulation in the geometrical process of mass remote sensing images. How to use the exiting map and ortho-image as knowledge and initial position to automatically extract the linear primitives from remote sensing images, which can be used as control information for image orientation and registration, has become a new way to photogrammetric automation. This is a new idea for solving the bottle-neck problem of remote sensing images processing.In this dissertation, many conjugate lines are automatically generated by automatic extraction of linear primitives from new remote sensing images based exiting vector maps. The exterior orientation parameters of the images are calculated based on the theory and the arithmetic of the extended line photogrammetry, in which the conjugate lines are used as the observational values. For many overlapping images in a block, the images are registered to the corresponding maps by the hybrid bundle block adjustment based on multi-conjugate primitives with different geometric characteristic. This lays a foundation for map revision by automatically registering the new image to the exiting map. Main content of the dissertation are discussed as following.This dissertation summarizes the exiting methods and arithmetic about the registering remote sensing images with vector map and the automatic orientation of the images in the literatures. The pro and con of the methods and arithmetic are analyzed and the research trend is discussed in the field. In this dissertation the registration between spacial data is defined. The software and the strategy proposed in this dissertation can realize the registration between the remote sensing images and the vector map by maidng fully use of many linear conjugate lines with the analyzing the line primitive's advantage over point one.Automatic calculating the exterior orientation parameters of remote sensing images is based on building the imaging projective geometry. The imaging projective geometry of common optical remote sensing images and their conditional equations are outlined in this dissertation. For satellite images, the disadvantage of the traditional collinear conditional equations is analyzed and the new strict geometric model is adopted from the literatures. So the problem of the relativity of image parameters calculation is solved completely.Automatic extracting many conjugate lines, which have obvious advantage over the conjugate points in automation and reliability, is the base of registering image to map and calculating of image's exterior orientation parameters. The method of linear primitives extraction mainly based on active contour model is dwelled on according to the image characteristics by combining the advantage of automatic extraction and semi-automatic extraction of linear primitives at present. So do the precisely locating of linear primitives based on the extended least squares template matching (LSTM). Different strategies and methods of extraction are adopted based on the image spatial resolution. The self-adaptive LSTM is extended for the extraction of edges with non-equal width and centerline of ribbon primitives by introducing a "scale" parameter in the LSTM. The extracted linear primitives are used as conjugate lines with the corresponding linear primitives on vector map, which are used as observational values for the adjustment in the image exterior orientation and the registration.Line photogrammetry, which has better automation and high reliability, is new research field with the aim to accomplishing a series of photogrammetric tasks by using linear primitives, such as relative orientation, spatial resection, spatial intersection and block aero-triangulation. With the progress of the satellite remote sensing technology, more and more high-resolution remote sensing images can be afforded. Most remote sensing images are captured by linear array scanner with charged coupled device (CCD) because of the technology of sensor and economical fact, in which every image line has individual exterior orientation parameters. The points on the linear primitive in the image have different exterior orientation parameters although every point is captured on the condition of perspective geometry. So the coplanar condition equation in the line photogrammetry for frame image cannot apply to the image captured by the linear array scanner. To circumvent this problem, the coplanar condition equation should be extended. In mathematics, all lines and curves consist of points. Same as visual point, co-linear equation can be used to the points in line feature, and all feature points can come down to the points in order to fit the co-linear equation, that is so called as generalized point photogrammetry. Consequently, it is easy to reduce the points, lines and curves into one mathematical model, which can apply to all kinds of remote sensing images. On the otherwise, the corresponding conclusions about the collinear condition equation can keep on adopting.To assess effectiveness the proposed arithmetic in generalized point photogrammetry, one is always required to understand its sensitivity and robustness, i.e., how perturbations in input data affect the output result. Through a series of test, this dissertation analyzes the elimination of blunders, and the effect of the precision and convergence speed of adjustment produced by the distributing and error of the conjugate lines. The redundancy and relativity of the observation are analyzed, so do the stability of the adjustment. The conclusion is drawn that to accomplish the spatial resection in generalized point photogrammetry, three unparallel conjugate lines each other are at least needed. The software system built for test the proposed arithmetic can realize the task of automatic single image (model) registering to vector map. The addition of conjugate lines makes the registration better and more rational at the distributing area of many linear primitives. The changed linear primitives suchas roads and rives can be check out on the based of the results of extraction and the adjustment, which can be also used for map revision.For many overlapping images in a block, the images are registered to the corresponding maps by the hybrid bundle block adjustment of generalized point photogrammetry based on multi-conjugate primitives with different geometric characteristics. In the block adjustment, the extended collinear equation is used as the conditional equation and the strategy of line translation is proposed for tie lines or pass lines. All images' exterior orientation parameters are calculated on the restriction of minimizing the error of all observations, which realize the whole registration between the images and corresponding maps. The block adjustment not only can circumvent the problem of the deficiency in conjugate lines such as the block water area, but also eliminates system error between the adjacent images at best, which can improve the precision of the registration between images and corresponding maps and the mosaic between the adjacent images.By the strategies and arithmetic mentioned above, some experiments are carried on for a lot of images including aerial images, SPOT and Landsat TM images with different resolutions. All of the results are quite perfect by analyzing the precision. The conclusion can be drawn that the automatic registration can be realized between remote sensing images and corresponding vector maps by addition of conjugate lines in generalized point photogrammetry. The precision can meet the need of the map revision with corresponding scale, which has the relationship with images' scale and the precision of vector maps and digital elevation model (DEM).With the pressing need for map revision, it is more and more important to do research in the field of automatic exterior orientation for remote sensing images, which can be used for rectifying the images and registering the images to the corresponding vector maps. The operational strategy and research content in this dissertation might be a help to the automation in photogrammetry and remote sensing at a certain extent.