Aerial No Control Points Based On Linear Geometric Correction Feature

This paper mainly made a research on geometric correction technology of history aviation image which lack of aerial parameters. As an important part of remote sensing images, aerial images have important applications in national defense, investigation of land change, disaster assessment and so on. And history aviation image reflects landscape changes over time, which has important historical and practical significance. Geometric correction of historical aerial image is the premise of applications list above.However, the lack of control point of historical aerial image made it difficult for its geometric correction. Compared with DOM (Digital Orthophoto Map), historical aerial image is quite different because of the lapse of time, so it is difficult to match each other only use they gray value. But in aspects of line features such as rivers and roads it maintains high consistency. With comprehensive analysis of existing image feature matching algorithm, we suggest it is better for us to extract the line features in this two images, with the help of line match algorithm and the knowledge of photogrammetry to get the control points to achieve the aerial imagery orientation. This paper mainly research on problems list below:1) In consideration of the spatial resolution difference between historical aerial images and the DOM images now available, this article built images pyramid and smoothed images with Gaussian Blur, then extracted image edge features by using Canny operator and extracted feature lines with Probabilistic Hough Transform with controlling geometric information of the lines.2) To describe the lines, they were partitioned into pixel support regions (PSR). The corresponding sub-regions were described to build mean-standard deviation line descriptor (MSLD). Optimal descriptor sub-dimension and matching criteria obtained by experiment were used to match feature lines between historical aerial images and the DOM images. By matching the historical aerial images and the DOM images, this experiment achieved7pairs of matching line features and showed a better feature lines’matching result by using MSLD operator.3) Finally, this experiment designed an algorithm of automatic interior orientation and exterior orientation elements calculating based on line features. Fiducial marks of the historical aerial images were extracted accurately by using OPTA algorithm, Canny operator and Freeman chain code to achieve automatic interior orientation of the images. And the calculating of exterior orientation elements was based on the theory of single image space resection in photogrammetry. The result showed that:the sampling points had an error of6.7m in the longitude direction and an error of7.1m the latitude direction.