Optimized Feature Points Extraction And Matching For Oblique Images Of Wide Baseline

The hopes of people want to improve visualization of the real world make thedevelopment of photogrammetry. Now photogrammetry has entered the era of digitalphotogrammetry and its most distinctive features from other stages are processing of the rawdata based on digital images, which means human measurements replaced by computer vision.The advent of digital aerial camera led to the development of aerial photogrammetrytechniques. For processing digital aerial imagery data which obtained by aircraft equippedwith a digital camera and be perpendicular to the ground photography, it has been formed andperfected a variety of advanced and mature the data processing system. The digital aerialimagery data processing system is capable of automatic aerial triangulation, automaticacquisition of digital elevation models and rapid production of orthophotos and otherapplications on vertical images.To reflect the intuitive world which compliance with human visual, oblique aerialphotogrammetry as a high-tech development is very rapid, which innovated the traditionalway of imaging. So that the system is not limited to vertical photography, but to use multiplecameras alternative the single camera photography for photographs from multiple directions,and obtain high-resolution digital aerial imagery from vertical and inclined multiple angles.Present digital photogrammetric method and system do not have the ability to effectively dealwith oblique aerial images, such as traditional aerial triangulation block adjustment do notsupport large dip in the mathematical relationship. So to achieve high-precision mappingusing oblique images, it is critical to employ both vertical and oblique images to blockadjustment simultaneously. The most difficult is how to automatically extract and matchbetween vertical and oblique images. The purpose of this paper is to study correspondingimage points extraction and optimization method which applicable to oblique images，andultimately achieve automatic matching between the vertical image and oblique images. Majorworks of this paper as follows:1. Based on oblique images. Firstly, a comprehensive discussion of oblique aerialphotogrammetry technology is put forward. Including the system components, the imagingmode, the basic principle and the key processing technique of oblique photographicphotographic camera systems, and importantly analyzed features of oblique image. All thestudy of the text in this article is based on features of oblique images.2. For the feature about geometric deformation, resolution changes, large angle imagerotation, terrain shade and big difference for feature information among oblique images, thispaper study the feature point extraction method based on invariant properties. First, the simplefeature point extraction operator is presented, and the oblique images were employed to theoperation to indicate the reason of invariant properties points’extraction.3. The study of the shadow detection algorithm aimed to optimize the extraction of feature points, filter the feature point located in the shadow contour edges or corners ofbuildings, to avoid matching difficulties due to the inconsistency of shadow area andorientation on oblique images.4. Research on feature point matching algorithm applied to oblique images, which alsobased on the inherent characteristics of the oblique images. For wide baseline matching ideas,combined with the pyramid structure, achieving invariant feature points matching between thevertical image and oblique images.