Line Feature Extraction And Matching Based On Digital Close-Range Image

Along with the advancement in acquiring airborne or space-borne images,building models in cities and large scale city models can be produced such as thoseprovided by Google Earth and Virtual Earth. However, three dimensional (3D)models with detailed geometry and textural information are still not available inmost of the existing systems. This requires more close-range data, such as groundimages, terrestrial laser scanning data, or oblique images from helicopter. Detailed3D modeling is a hot topic nowadays either for image-based modeling or LiDARpoint cloud-based modeling. Image based modeling remains the most economical,flexible, convenient, and popular method for3D building modeling. One key issuein image-based modeling is image matching, which uses the computer to simulatehuman eyes to automatically determine corresponding primitives by stereoobserving. Image matching is always the common academic frontiers and difficultproblem in the photogrammetry and computer vision communities.The author’s research group has been engaged in the research of3D laserscanning technology, and has made some achievements in the ancient heritagepreservation. We discovered that the scanner cannot be able to get the edge ofbuildings completely, which it requires the help of image matching to obtain theedge information. Now the edge information in the scene is still not able tocorrectly matched by image matching, especially in the image of building，not onlythe lack of texture but also the problem that the surface is non-continuous. In orderto solve this problem and combined with the existing conditions, I presentedmethod of line matching based on the constraints in object which guided by thepoint clouds. This method can reduce the search scope and the reliability of the linematching. For the problem of linear feature extraction, I found the method of “lineinstead of the curve”. It can be expressed the edge with a straight segment. Meanwhile, in order to improve linear similarity measure can distinguish betweenthe nature of the calculation, I use a moving window adaptive linear correlationcoefficient that the surface texture of the lack of non-contiguous areas and regionscan be matched reliably and improve the success rate of a straight line and thecorrect matching rate.I used the c#and matlab to achieve the algorithms in this paper thatline-feature matching of Point cloud-based constraints to guide the physical andrelated experiments. The experiments show that the method in this paper can getthe reliable line features of the same name in the left and right images. It laid agood foundation to get the three-dimensional edge information of buildings andother artificial surface features.