Discussion on the establishment and application method of image control point library

(China Land and Resources Aerogeophysical Remote Sensing Center, Beijing, 100083)

This paper discusses the principle and method of establishing the control point database of SPOT5 image, and introduces how to establish the control point database suitable for SPOT5 correction, extract the control points and match the image to be corrected automatically or manually, and find the points with the same name on the orthographic image to establish the control point pair, so as to realize the geometric correction of the image.

Keywords: SPOT5 image; Control point database; Matching; Geometric correction

1 Introduction

Orthographic correction of SPOT5 image is a basic process of image processing in land dynamic monitoring project. Both physical model equation method and polynomial correction method need to use ground control points to solve transformation matrix to realize geometric correction. At present, the land dynamic monitoring project has begun to establish many image control points, including aerial or high-resolution remote sensing digital image control points after orthorectification, graphic control points on large-scale digital topographic map after scanning correction and newly measured field GPS image control points. But in order to effectively use these control points, the key is to effectively manage them. These control points should be saved after use and can be called and updated when used again. Therefore, it is very necessary to establish a database of control points and realize the effective utilization of control points. Because of the particularity of image control points, this paper expounds the establishment and use of image control point database.

Image control points are images containing geographical features, and their storage format is a grid with geographical information. Image control points use the matching between images to find points with the same name on orthographic images, which avoids the difficulty that traditional control point markers are difficult to identify on orthographic images. With the support of computer software and hardware and pattern recognition technology, using image control points instead of traditional control points for geometric correction can greatly reduce labor intensity and improve work efficiency and correction accuracy.

2. The establishment of image control point database.

The basic purpose of image control point database is to manage control points effectively, extract control points conveniently and realize geometric correction of images. This requires that the design of database should pay attention to the use of control points. In the geometric correction process of SPOT5, the approximate spatial range of the image to be corrected is known according to the track parameters of SPOT5, and the control points are extracted according to this range. The basic function of control point database is to query control points based on spatial range. At the same time, each image control point contains two kinds of data, image data and attribute data. The connection of these two kinds of data is an important basis for the application of control point database. Remote sensing software with automatic position prediction function should be used to apply the control point database. The design of image control point database is based on these three basic principles.

2. 1 image control point source

Image control points are files based on images and supplemented by vectors.

(1) Cut a small image of a typical area from aerial or remote sensing digital images and use it as an image control point after orthorectification.

(2) According to the coordinates of the field GPS control point, mark the position of the control point on the original image, and attach the original data description file and field digital photos.

(3) Taking the small-area digital image of typical terrain area obtained by scanning the corrected digital topographic map as the image control point.

(4) The reference coordinate system of the image control point should be consistent with the coordinate system of the desired result image.

2.2 the content of the basic control point library

The control point library manages image control points, including image data and attribute data. No matter which method is used to collect image control points, there are both kinds of data. Image data and attribute data are stored in different libraries. The storage format of image data is raster, and the storage format of attribute data is vector. Image data and attribute data must be connected in different libraries.

2.2. 1 image data

Image control points are images containing obvious features stored in grid form. In the database, because of the particularity of raster image, it cannot be stored as a record like attribute data. Each image is stored as a raster file in a directory and managed according to the directory. The size of an image is generally between 100× 100 pixels and 200×200 pixels, whichever can contain obvious features. Image control points are different from traditional control points in that they have image data. The obvious feature in the image means that it can be distinguished from other features around it in a certain range, which can be the intersection of a road, the bend of a river or even an island. Its characteristics make it possible to select points for geometric correction in areas where traditional control points cannot be determined.

attribute data

Attribute data is used to describe the geographical location and other relationships of control points. To use a group of images as control points of geometric correction, they must have correct mutual position relationship in a certain projection space. The geographic location of the image control point is described by its attribute data. In order to describe the geographical relationship correctly, the attribute data of each control point should have a consistent ID identification number associated with the image data, so as to realize the correct connection between the image data and the attribute data. The attribute data format of all control points is the same, so the attribute database is a relational database, and the attributes of each image control point are recorded in a specified format. Attribute data records include: the source of image control points; Control point coordinates; Data description; Reference ellipsoid; The proportion of the image; Hyperlink digital photos of field GPS control points.

reference frame

All image control points are projected into a reference coordinate system. In order to improve the performance of the database and correctly express the geographical relationship of control points in the whole database, the control points should adopt a unified coordinate system to facilitate the correct query, extraction and use of control points.

2.3 the structure of the control point library

The number of control points in different areas is different, and the data of control points in a large area is very large. It is very inconvenient to query and use control points for a single database in a large area, which will reduce the performance of the database. Therefore, in order to query and extract control points quickly and conveniently, it is necessary to establish a database layer by layer by index to form a tree-structured control point library. Because the control points are distributed according to the geographical location, it is reasonable to divide a large area into several small areas according to the geographical location range, and the small areas can be further subdivided according to the actual situation, thus establishing a top-down index database step by step.

The top-level database is a global database, which describes the information of partitioned databases within the whole database construction scope, and it is also a relational database. The contents described in the record include: the name of the sub-database, the range information contained in the sub-database, etc. According to the actual situation, the sub-database can describe the information of the lower-level database or the information of the control points.

Figure 1 tree database structure

In this tree database structure, the leaf database is in the most basic position, which describes the information of image control points. When you want to extract control points from the database, you can query down from the top-level database layer by layer until the leaf database queries the contents of the basic control point library, as shown in figure 1.

As can be seen from the structure diagram, images and attribute data are stored in a directory in the form of files, and the management of image database is actually the management of file directory. Only by organizing the file directory reasonably can the connection between image data and attribute data be realized, which requires the directory structure to be consistent with the naming and attribute database.

3 Application of Control Point Library

The purpose of image control point library is to organize and manage control points effectively, and to extract control points in a certain image range conveniently for geometric correction. Due to the introduction of the original data orbit parameters, the SPOT5 image to be corrected can get an image with geographical information. According to this geographic information range, starting from the top control point database, find the sub-database of this image range, and then enter the next database to make the same judgment until the bottom database, and then extract the image control points located in this image range for geometric correction.

Once the control points are extracted, their geographic location data can be obtained. According to its location and geographic information of remote sensing image to be corrected, the approximate location range of control points on the image is automatically matched, and the search within this range can greatly shorten the search process of matching points with the same name and improve the speed and accuracy of matching.

When using a control point, whether it is a GPS control point or an image control point, a simple method is to identify the approximate range of the control point that needs to be corrected on the image, which can be obtained only according to the geographical coordinates of the control point and the range of the image, and then click on the computer screen with the mouse to get the image pixel coordinates of the control point with the same name. In order to realize the automation of geometric correction, it is necessary to use image matching technology to automatically match the control points of the same name points on the image to be corrected. According to different types of control points, different matching techniques are adopted. The application flow of control point database is shown in Figure 2.

Fig. 2 Application process of control value database

3. 1 remote sensing digital image control point registration method

By using the registration method based on regional features and point features, a small piece of typical area is intercepted from the aerial or remote sensing digital image corrected by orthography as the image control point.

3. 1. 1 Region-based registration method

The region-based registration method is to statistically compare a region in the image to be registered with a region of the same size in the reference image, and its similarity evaluation criterion is to take the maximum from the standardized cross-correlation coefficients of the two regions. You can also transform the image from time domain to frequency domain by FFT, and then register it. For images with large displacement, the rotation of the images can be corrected first, and then the mapping relationship between the two images can be established. However, if there is large noise and gray difference in the image, this cross-correlation metric becomes unreliable.

3. 1.2 Registration method based on point features

The registration method based on point features has high performance. It has two processes: feature extraction and feature registration. A series of image segmentation techniques are used in feature extraction and boundary detection. Such as Canny operator, Laplacian Gaussian operator and region growing operator. The extracted spatial features include closed boundaries, open boundaries, intersecting lines and other features. The algorithms of feature matching include: cross-correlation, distance transformation, dynamic programming, structure matching and so on.

3.2 GPS control point registration method

For the field GPS control points, the position of the control points marked on the original image and the small-area digital image of the typical feature area obtained on the scanned and corrected digital topographic map are used as control points, and the method of artificial matching of the same name points is adopted.

Because the control points of topographic map only provide a kind of structural information of ground objects, similar to the texture of images. The spectral information of ground objects is not reflected, which is inconsistent with the content in the image to be corrected; It cannot directly match the image by using the data in the control points. Therefore, we can only use these structural information to manually match the points with the same name, and find the points with the same name of the control points according to the texture characteristics of the image within the range where the approximate position of the control points can be predicted automatically.

Combining the above methods, after matching a sufficient number of control points to find points with the same name, we can solve the transformation matrix according to these control points to realize geometric correction.

The establishment of image control point database is a basic work, and a large amount of data should be input into the database. Once the database is established, it can be updated with the latest remote sensing data. When it is necessary to use control points for geometric correction of new remote sensing data, control points can be extracted conveniently and quickly, which can improve work efficiency and provide technical support for land survey.

refer to

Barbara Chitova and Jane Lu Se. "Overview of Image Registration Methods." Imaging and Visual Computing, Vol.21,pp.977 ~1000, 2003.

Zhang Zuxun, Zhang Jianqing. Digital photogrammetry. Wuhan: Wuhan Technical University of Surveying and Mapping Press, 1996.

Zhang Zuxun, Zhang Jianqing. Small-panel differential correction of remote sensing images in mountainous areas. Proceedings of the Third Cross-Strait Symposium on Surveying and Mapping Development, 2000, 12.

Chen Yuefeng, the boy is beautiful. Image database system model based on content query [J]. Journal of China Image Graphics, 1997.