RESEARCH ON ACCURACY EVALUATION METHOD BASED ON THE RESULT OF IMAGE CORRECTION AT 10000-SCENES CLASS

RESULT OF IMAGE CORRECTION AT 10000-SCENES CLASS LI Miao1, CHEN HaiPeng1, QIU Bo1, XIE WenJun1, CHEN YuHeng1 1National Quality Inspection and Testing Center for Surveying and Mapping Products, 9105513@qq.com Commission VI, WG VI/4


INTRODUCTION
National basic geographic information is the focus of national spatial data infrastructure construction, The National Basic Geographic Information Database is updated annually for the entire country, we will collect and obtain domestic satellite remote sensing images such as ZY-3 covering the whole country, complete orthorectification of panchromaticchromatic images and multi-spectral images in the form of scenery, forming the results of national satellite image orthorectification processing.
As the basic image data of the 1:5 million terrain data update in the country, the quality level of image orthorectification processing results directly affects the terrain data update work and the final result quality [1] . Carrying out quality inspection of results is an important means to discover the quality of results and to ensure the quality of results. Due to the particularity of image ortho-correction processing results, targeted inspection techniques and methods are required to highlight the pertinence and effectiveness of quality inspection while ensuring comprehensive coverage of quality requirements.
Based on the actual inspection work, considering the operability and labor cost in scale production, this paper proposes a quality inspection and evaluation method based on the large-scale ortho-corrected image of 10000-Scenes class. Since the focus is on the accuracy of the mapping results when using the satellite as the basic mapping update, this paper focuses on the accuracy evaluation method of image correction results.

ACCURACY
When satellite images are used as the basic mapping update, it is necessary to further understand the relevant performance indicators of satellite images. In addition to considering the accuracy of the mapping results, it is also necessary to consider various factors such as operability and labor cost in scale production.
The ground positioning accuracy of satellite images refer to the accuracy index that can be achieved by determining the position of the ground point based on satellite images. It is related to the spatial resolution of the image, the geometric distortion of the image, and the image quality [2] .

Image Spatial Resolution
Spatial resolution, also known as ground resolution, is

Image Geometric Correction Accuracy
Geometric correction is to eliminate the geometric distortion of the image, that is, to transform the image   Table 3. Edge error

Image Texture Quality
The imaging quality of satellite images are mainly affected by factors such as ground illumination, cloud fog and flight attitude during photography. It can be evaluated and checked from the following aspects: (1)The image signal-to-noise ratio is small, the linearization in the dynamic range is better, the gray level is more, the image level is richer, and the resolution of the target in the shadow is improved.
(2) The image tilt angle should be small to reduce the deformation and blur of the back slope image; Cloud or fog coverage is generally less than 10%, and more than 10% need to try to supplement images.

STECHNICAL ROUTE DESIGN
The

Production Technology Route Analysis
In the production, the precision correction and analysis of the regional network adjustment orthorectification results based on high-precision control points in recent years are carried out. The RMSEs in the image plane position is less than 5m, and it has reliable geographical reference accuracy.
The previous results are used as a reference for correction image in the 1:50,000 terrain data update, the reference images are mosaiced in a million standard frame, and a total of 63 million frames are formed nationwide. The panchromatic image orthorectification is mainly based on the image of the million-frame reference image. The matching is corrected, and then the control point matching, editing and orthorectification processing corresponding to the multi-spectral image are completed.

Overall Technical Route
In the process of plane position accuracy detection, comprehensively consider the use of higher-precision geographical images, production-corrected reference images as reference data for accuracy evaluation,

Plane Position Accuracy Check Method
In the plane position accuracy detection process, through the use of internal inspection and reference data comparison method, the plane position accuracy is comprehensively analyzed in various ways, which has become a complete and scientific precision verification process. Comprehensively use higher precision image results, reference images, etc. as reference data for accuracy evaluation [3] , supplemented by panchromatic images and multi-spectral images for nesting inspection.
(1)The production images used in the production are

Edge Accuracy Evaluation Method
The result image has a overlapping edge and the entire overlap is not completely connected,partially incomplete connected,and the edge is not connected to the edge. Therefore, the RMSEs in the edge of GB/T18316-2008 [4] cannot be directly used for quality evaluation,but refer to GB/T24356-2009 [5] , and the quality assessment is carried out by the wrong-leakage The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15-17 November 2019, Guilin, Guangxi, China deduction method. That is, the four types of faults A, B, C, and D are determined according to the specific conditions of the joint edge problem, as shown in Table 4.  Table 4. Edge error classification table

Result Analysis
The panchromatic image of the single-view is   By using the technical methods of this paper to carry out the actual work, the results of the orthorectification correction processing in the 1:50,000 terrain data update are tested and evaluated, which proves that the route is feasible and the quality evaluation results are objective and reliable, and can correct the large-scale and 10000-Scenes class images.
The quality inspection and evaluation of the treatment results provide reference and method guidance.