Research on the Construction of Control Point Database of High Resolution Satellite Image GPS Image

(Zhengzhou 4500 16, Information Center of Henan Provincial Department of Land and Resources)

Abstract: Based on the analysis of various control data used in the data correction of SPOT 5_2.5 m high resolution satellite image, the necessity of establishing a database of GPS image control points is expounded. By discussing the selection of image control points, field survey, accuracy evaluation and the construction of GPS image control point database, it is proposed that selecting GPS control points as the control data of SPOT5_2.5 m high-resolution satellite image data correction can ensure the accuracy of image correction, save time and reduce investment.

Keywords: satellite remote sensing control point image correction database

0 quotation

With the rapid development of information technology, satellite remote sensing technology has made a breakthrough, especially on May 4, 2002, the French SPOT 5 earth remote sensing satellite entered the scheduled orbit, which greatly promoted the scientific and technological progress and management level of various application industries. High-resolution satellite remote sensing has made remarkable achievements in land resources investigation and evaluation, land use dynamic monitoring, land renewal investigation and large and medium-scale topographic mapping.

For SPOT 5_2.5 m high-resolution satellite image data, geometric correction mainly adopts two models: two-dimensional polynomial and three-dimensional digital differential correction, and the correction control data mainly includes 1 ∶ 1 10,000 or larger scale digital raster topographic map (DRG) and land use digital raster map (LUDRG). Through relevant research, the author thinks that it is more appropriate to choose image control points as the correction control data of high-resolution satellite image data. In view of this idea, the project team conducted a series of discussions and studies, and established a database of GPS image control points in some areas of Henan Province based on MapGIS platform, which laid the foundation for future related work.

1 status

At present, satellite image correction with various resolutions basically refers to topographic map, digital raster topographic map or digital raster map of land use. Aiming at the production of SPOT 5_2.5 m digital orthophoto map, the Cadastral Department of the Ministry of Land and Resources has specially formulated the Technical Specification for the Production of SPOT 5_2.5 m digital orthophoto map. It is clearly stipulated that the digital orthophoto map of SPOT 5_2.5 m should take 1 ∶ 1 10,000 (or larger scale) digital raster topographic map, land use digital raster map or high-precision field control points as control data. Through the relevant work in recent years, the author thinks that the current correction control data, especially in Henan Province, have the following problems.

1. 1 The existing topographic map of 1 ∶ 1 10,000 in Henan Province has not been fully covered, and the accuracy of topographic maps is different, and the present situation is poor.

There are 6565 topographic maps in Henan Province, covering1:110,000. At present, only 5,600 topographic maps have been mapped, and about 15% has not been mapped. Most of the existing topographic maps were measured by surveying and mapping departments, geological and mineral surveying and mapping units and coalfield geological surveying and mapping units in the 1960s and 1980s respectively, with different surveying and mapping accuracy. Due to the deformation of paper maps, the actual deviation between the line connecting the km grid and the intersection point of the km grid on some topographic maps is 1 ~ 3 mm, and some of them are more than 3 mm. If these topographic maps are used as control data to carry out SPOT5_2.5 m high-resolution satellite images, secondly, the existing topographic maps are thirty or forty years ago, and some surface elements have long been unrecognizable, so it is difficult to find feature points with the same name. Even the updated topographic map only updates the main features such as the main roads and residential areas above the towns, and most other features and contours follow the original map.

1.2 land use status map (database) is difficult to meet the accuracy requirements.

The investigation on the present situation of land use in Henan Province began in the late 1980s and ended in the mid-1990s. The survey method basically adopts1:1000 aerial image plan or1:35000 color infrared aerial photograph and1:1000 topographic map for field mapping, and then carries out indoor mapping, area measurement and leveling. All the processes are manual operations, which are interfered by various factors, so the quality of drawing varies greatly. If the current land use map (database) is used as the control data to correct the SPOT 5_2.5 m high-resolution satellite image data, its correction accuracy can not meet the requirements of the specification.

2 image control point selection

This experimental study involves eight SPOT 5 satellite images in Pingdingshan, Xuchang and Luohe, Henan Province. The DEM scale covering the experimental area is1∶ 50,000, and * * * selected 152 image correction control points.

The selection principle of image control points is that the distribution of points is relatively uniform, the characteristics are obvious, the traffic is convenient and the number is sufficient. Image control points should be selected on the panchromatic spectrum as far as possible, avoiding high-voltage lines and large areas of water.

In order to improve the efficiency and accuracy of field survey, after selecting the image control points, the selected image control points are made into the Field Survey Results Table of Image Control Points which is easy to carry and save, and the number, points and enlarged schematic diagram of image control points, WGS84, 1954 Beijing, 1980 An three sets of coordinates and point descriptions are recorded respectively, which are used to establish the graphic image database of GPS image control points.

3 Field investigation of image control points

The field survey of image control points adopts the attached route method, and the average distance between each image control point is about 13 km. The point with "P" in front of the serial number indicates the image control point of this measurement, and the point with "C" in front is the C-level GPS control network, which together with the C-level point * * constitutes the GPS control network (Figure 1).

Figure 1 Schematic diagram of the distribution of image control points and reference C-class GPS control points.

In this GPS control survey, three groups of data (WGS 84, Beijing 1954, Xi 'an 1980 coordinates) of the C-class GPS control network of Henan Province are used as initial data, and according to the global positioning system (GPS) survey specification, three sets of GPS receivers are used as a group to carry out synchronous static observation. The observation period is 45 minutes, the satellite altitude angle is ≥15, and the total number of effective observation satellites is ≥ 4. Operators fill in the field survey record form, and describe the image control points in detail through digital photography and point recording. South surveying and mapping software is used to calculate the baseline and adjust the measured data to fit the elevation. The fitting heights of three groups of data and correction control points based on three groups of coordinate systems are calculated respectively. 152 the median error of the weakest point in the plane position of the image control point is 6.8 cm, and the accuracy in elevation fitting is 0.32 1 m, and the accuracy of the results meets the requirements of the specification.

4 image data processing and DOM production

Image data processing mainly includes image registration, fusion, orthorectification, mosaic and the production of 1∶ 1 million orthographic image (DOM). Because the SPOT 5_2.5 m satellite image is received synchronously by single scene multi-spectral data and panchromatic data, its graphics have good geometric correlation, and the registration of multi-spectral data and panchromatic data is difficult and the accuracy is not high, so the relative registration method is adopted. In the process of image data fusion, considering the differences between scenes caused by different data receiving time, cloud interference in the air and uneven light on the ground, the data are preprocessed by linear stretching and texture enhancement before data fusion, so that the whole scene image has moderate brightness, clear texture and prominent details, thus improving the visual interpretation accuracy. Image fusion mainly adopts the most basic product merging algorithm to directly fuse remote sensing data with two spatial resolutions. After fusion, through histogram adjustment, USM sharpening, color balance, chroma saturation adjustment and contrast enhancement, the whole scene image is truly uniform in color, moderate in brightness and clear, and thematic information, especially texture information, is enhanced.

Using the LPS orthorectification module in ERDAS, a professional remote sensing image processing software, the remote sensing image is orthorectified. This correction adopts SPOT 5 physical model, and the control points are evenly distributed in the whole scene image, with 25 control points in each scene, and there are more than two common control points in the overlapping area of adjacent scene images. Orthophoto correction is based on the measured GPS control points and 1∶50000 DEM, and the fusion data of SPOT 5_2.5 m is corrected in the unit of scene, and the sampling interval is 2.5 m. ..

The batch processing module of LPS orthogonal projection module in ERDAS is adopted for image mosaic, and more than two common control points are collected for two adjacent images, thus ensuring the accuracy of image mosaic.

DOM is made by Image Info tool and cut according to the national standard1:10000, covering a complete county-level administrative area. The map is decorated according to the technical requirements of high-resolution image data processing and database construction. MapGIS software is used, and the projection parameters are according to Gauss-Kruger projection, Beijing coordinate system 1954 and national elevation datum 1985.

5 DOM accuracy evaluation

DOM accuracy evaluation is based on field checkpoints, and the evaluation method is checkpoint selection method: by selecting feature points with the same name between DOM images and field checkpoints, the correction error and median error are calculated.

5. 1 checkpoint selection and field measurement

Checkpoint selection: a scene image is randomly selected as an evaluation unit, and 30 relatively evenly distributed checkpoints different from the calibration control points are selected. The selection principle of inspection points is consistent with that of image control points, and the influencing factors such as high-voltage lines and large water areas should be avoided as far as possible when selecting points.

Checkpoint measurement: The field measurement of the checkpoint is consistent with the measurement method of the image control point, that is, a complete GPS control network is formed by using the attached route method, and the measurement is carried out synchronously with the same accuracy in a static way.

5.2 Calculation of correction accuracy

The accuracy evaluation formula is as follows:

Comprehensive study on large-scale and efficient processing technology and database construction of remote sensing images in Henan Province

Where: RMS- point median error;

N- the number of checkpoints;

Ui-x and y coordinates of checkpoints on DOM images;

VI —— X and Y coordinates of GPS field checkpoint.

According to the production accuracy index of 1 ∶ 1 10,000 DOM in Technical Specification for Making Digital Orthophoto Map of SPOT5_2. 5m； ",plain and hilly areas point error is not more than 5 m; The mountain area is not more than 7.5m；; The mountain area shall not exceed 10m. The selected areas for this accuracy evaluation are mainly plain areas, and some are hilly areas. After calculation, the median error of the selected points is 2.62 m, which fully meets the production accuracy requirements of 1 ∶ 1 10,000 DOM. See table 1 for the calculation table of calibration accuracy evaluation.

Table 1 revised accuracy evaluation calculation table

sequential

Establish 6 databases of GPS image control points

In order to achieve high efficiency and high precision of geometric control and result check in accurate geocoding, a database of GPS image control points is established to meet the requirements of image correction and registration.

The establishment of the database of GPS image control points, taking the geographical map of Henan Province1∶ 500,000 as the working base map, input the spatial coordinate files of control points, collect the attributes and graphic files, and establish a unified image control point file with mathematical foundation.

The collected image information of image control points includes not only the geographic coordinate information of general image control points, but also the texture information, resolution information and scale information of characteristic ground objects related to the image to be corrected.

Collect attribute information of control points. Collect control point attributes, record resolution, scale, range, ellipsoid information, projection information, coordinate system information (Beijing 1954 coordinate, Xi 'an 1980 coordinate, WGS84 coordinate), database production unit, production date, etc.

Fig. 2 Schematic diagram of graphic image database of image control points

7 concluding remarks

Land renewal investigation, dynamic monitoring of land use by remote sensing and law enforcement inspection of land illegal cases should not only consider the correction accuracy of remote sensing images, but also consider their present situation, image processing time and input cost. The establishment of the database of GPS image control points not only meets the accuracy requirements of SPOT 5_2.5 m satellite image correction, but also lays a foundation for similar work in the same area in the future, greatly reducing the input cost and saving the image processing time, which plays a role of "once and for all".

Take the exam and contribute.

Don Rong 'an et al., 2003. Remote sensing image processing method of ERDAS imaging [M]. Beijing: Tsinghua University Publishing House.

Wang. 1990. principle of photogrammetry [M]. Wuhan: Wuhan Technical University of Surveying and Mapping Press.

You Shuzhu, Liu Shunxi. 2002. Application of GPS in land change survey [J]. Bulletin of Surveying and Mapping (5): 1 ~ 3

Zhang Jixian et al. 2000. Database of graphic image control points and its application [J]. Bulletin of Surveying and Mapping (1)

(Originally published in Land and Resources Informatization, No.3, 2007)