Digital Surface Model(DSM) is an essential data for applications of GIS and intelligent land information. DSM is usually generated from aerial image and satellite image by manually.
An image matching is very important for automatic DSM generation and...
Digital Surface Model(DSM) is an essential data for applications of GIS and intelligent land information. DSM is usually generated from aerial image and satellite image by manually.
An image matching is very important for automatic DSM generation and has been studied as a main topic in computer vision and digital photogrammetry. However the image matching processing has a difficulty caused by fundamental problems of photogrammetry which are central projection and occlusion.
The large-format aerial digital camera ADS40, which was developed in 2000, directly acquire digital form image with linear CCD(Charge-Coupled Device) sensor. The ADS40 acquires 100% overlapped multiple images with several linear array CCD sensors using pushbroom scanning method. Therefore, the geometric and radiometric characteristics of nADS40 image are quite different from those of traditional frame aerial camera image. Moreover ADS40 equipped with GPS/INS on board provides direct exterior orientation for digital image so that DSM generation could be much more automated.
This study investigates geometric and radiometric characteristics of the ADS40 image. Moreover this study suggests the image matching strategy for extracting 3D spatial information with the ADS40 image. Since the ADS40 Level 1 image satisfies quasi-epipolar line geometry, the search space for image matching could be reduced in great deal. To speed up image matching as well as cut down errors of image matching, coarse-to-fine strategy was implemented.
Among multiple image combination for image matching, GRNN image which is one of nadir images in ADS40 plays a primary role in image matching. Experimental results presented that image matching between GRNN image as master and REDF, GRNF, PANB images as target, respectively didn't show such a big difference.
As ADS40 captures multiple images from multiple linear CCDs, it is possible to perform multiple image matching. Based on experiments, multiple images more than two images doesn't enhance the image matching outcomes. In other words, thers are no considerable advantage as well as benefit in image matching with more than two images. From experimental results, image matching with two images showed that the result of success rate is 92% in open area, 81% in building area, and 59% in dense apartments area.