Methodological Research On Geo-positioning High Resolution RS Imagery On The Basis Of Small Topographic Map

In order to identify land objects and determine their geo-spatial coordinates by the way of high-resolution remote sensing, the captured remote sensing image (RS image) needs to be precisely ortho-rectified. Usually ortho-rectifying RS images require ground control points (GCP), of which geodetic coordinates are known. However in the mission of reconnoitering foreign lands, it is very difficult to carry out GCP field survey and it is also not easy to obtain large-scale topographic maps. Under such circumstance, it is evidently valuable in practice to research on a new way which can ortho-rectify high resolution RS images dispensable of large-scale topographic maps and GCP field survey.This dissertation proposes a multi-source RS images' geometrical rectification method. This method runs in the order of from coarse to fine. At first, take the globally public topographic map (1:1000000 or 1:250000) as geo-referencing basis to rectify low resolution RS image, such as LandSat image, by their homogeneous linear features. In the second, take the rectified low resolution image as geo-positioning basis to rectify next level RS image. Then the newly rectified RS image is used as geo-positioning basis for next higher resolution image rectification. Till the highest resolution image is rectified. At this stage the rectified RS image with highest resolution can be utilized for the identification and geo-referencing of land objects.The main research subjects and innovative spots of this dissertation are as follows:I. The strategy of geo-positioning high resolution RS imagery Geo-positioning RS imagery has always been an important research subject in the field of photogrammetry. Recently, to implement geo-positioning RS imagery using GCP as few as possible has become a hot topic in the sphere. Some high resolution RS systems are confronted with difficulties in geo-positioning their captured images. For example, in foreign lands and bad condition areas, it is hard to carry out GCP field survey, aerial tri-angulation as well as exterior elements measure in domestic high resolution Satellite RS system and Unmanned Aerial Vehicles(UAV) system. This dissertation puts forward a technical strategy to deal with the above mentioned problem, which utilizes small scale topographic map as geo-referencing basis, geometrically registers multi-resolution RS images from coarse level with next finer level till the highest resolution RS image is rectified and got geo-referenced. The practicality and application value of such a strategy in rectifying images either from domestic satellites' mission of reconnoitering foreign lands or from UAVs' monitoring projects is clarified. The practicality and application value of this strategy in multi-resolution image construction of Digital Earth is also illustrated.II. Graphic-Image matching method that ensures low resolution RS image be precisely rectified by small scale topographic map.Image rectification by narrow-sense GCPs is a classic method and is widely used. The dissertation gives it a through discussion and uses narrow-sense GCPs in image rough rectification. Then the dissertation proposes to use linear features asgeneralized GCPs to do image precise rectification, aiming at overcoming weakness of narrow- sense GCPs method.The extraction of homogeneous linear features from small-scale topographic map and low resolution image is implemented by man-machine combined way. The extracted skeleton lines are further digitally shaped to become gray gradient images that characterize the global geometric relationship in topographic map and low resolution image. Through two gray gradient images matching, low resolution image is wholly registered with small-scale topographic map.The complicated geometric characterization of RS image caused by various influence especially by terrain relief is taken into consideration. The measure of image sub-division and border lines evaluated with big weights in the corresponding sub-images Least Square Matching is taken.III. Multi-resolution images positioning level by levelA group of ortho-rectified RS images of the same area, with different resolution values, are used as experiment data. Any two of them form an image pair. The experiments of image pairs matching show that image similarity decreases with their image resolution differences increase. Therefore, given the condition that low resolution image is already ortho-rectified, to achieve geo-positioning high resolution RS images with high precision, it is necessary to perform multi-resolution images matching level by level.Current image matching metric and matching method is discussed. On the basis of current usable RS image types, a multi-resolution image construction scheme is proposed. Through the constructed multi-resolution image, land objects in high resolution RS image can be identified and positioned quickly. And also the multi-resolution image can be used to interpret environments of interested objects level by level, so that the mission of reconnoiter lands is thoroughly supported.IV. UAV's high resolution RS image positioning.The advantages and shortcomings of UAV high-resolution RS system are analyzed, along with the difficulties in the absolute orientation of UAV high-resolution images. The technical scheme utilizes rectified low resolution RS image as positioning basis, then rectifies UAV images and mosaic them to form a big high resolution orhto-image is raised. The scheme is verified applicable with real image data captured by low attitude UAV.In short, this dissertation proposes a new scheme to produce ortho-rectified RS products. Other from, conventional surveying & mapping products producing specification and procedure, this scheme is abnormal, but suitable for applications in many aspects with many purposes. The scheme's principle, technical procedure is thoroughly explained. The usability of such products is also illustrated.