A New Approach of Digital Orthorectification Map Generation for GF3 Satellite Data

摘要

The successful launch of GF3 satellite has achieved long-term, sustained and stable acquisition of DOM product all over the China, however, it is a difficult project to deploy angle transmitter or select the same image points between SAR image and optical image manually, which only suitable for SAR image in some small areas, and cannot meet the needs of production and engineering application in large areas. In order to reduce the cost and improve the efficiency of surveying and mapping, a new approach of DOM product generation in large areas was proposed in this paper, which can automatically achieve the process of heterogeneous data matching of SAR data based on the optical DOM reference map and the process of dodging and mosaic line editing. 100 scenes of SLC image of GF3 satellite with the FSII imaging mode covering a large area and Beijing-Tianjin-Hebei region were selected as experimental image, by the heterogeneous data matching, high-precise control points and link points were automatically acquired, and the block adjustment of GF3 data in the experimental area was completed. The RMSE of all control points and link points in azimuth direction was better than 0.41 pixels, the RMSE in range direction was better than 0.85 pixels, and the maximum error was better than 3 pixels. According to the requirement of quality examination, based on the ZY3-DOM data, 5 checkpoints were well-distributed selected for each scene of image. The experimental results showed that the horizontal accuracy of 66% of image orthorectification result were better than 1 pixel (10 m), and 34% of image orthorectification result were better than 2 pixels (20 m), in which the accuracy between 10 and 20 meters were mountain or high-mountain terrain. And the GF3 orthorectification result can meet the horizontal accuracy requirement of 1:50000 scale surveying and mapping in China. At the same time, for the orthorectification result of 100 scenes of GF3 data, dodging and mosaic line were automatically generated and processed, and appropriate dodging points were added to adjust the grey value of one or both sides of image. It can be seen from this article that the hue transition of whole experimental area was more natural, there was no obvious hue difference, and the amount of manual editing was not big. Therefore, the degree of automation of this approach proposed in this paper was very high, which can meet the needs of dodging process in large areas of GF3 data.