Satellite night light image geometric correction method based on road network information control points

Abstract

The present application relates to a kind of satellite night light image geometric correction method based on road network information control point, belong to night light satellite remote sensing image processing technical field, after establishing road network vector database, obtain the high-resolution color night light satellite remote sensing image in city area range, obtain target night light image after pre-processing;Extract the road model in target night light image, extract feature points and filter invalid feature points, generate night light image road network intersection matrix An;Obtain open source road network vector intersection matrix Av by doing N times extension, use matrix An in matrix Av do sliding window calculation, and use the maximum value of window matching value Mw When corresponding each point position of the matrix Av of the matrix Av of the corresponding point position of the matrix An is updated, the coordinate value of the geometric correction of night light satellite remote sensing image is completed.The present application can effectively guarantee the precision and reliability of high-resolution night light satellite remote sensing image geometric correction, and simultaneously improve the processing speed of night light image geometric correction.

Description

Technical Field

[0001] This invention relates to the field of nighttime light satellite remote sensing image processing technology, and in particular to a geometric correction method for satellite nighttime light images based on road network information control points. Background Technology

[0002] In current geometric correction methods for nighttime light satellite remote sensing images, traditional feature point extraction algorithms are generally not applicable to nighttime ground features; the method of using regional network adjustment is also difficult to use after calculation; the method of extracting control points needs to be verified on existing daytime images, but since there are significant differences between nighttime light features and daytime light features, it is difficult to find an algorithm that is applicable to both and establishes a matching relationship.

[0003] Furthermore, previous algorithms for processing nighttime light satellite remote sensing images were based on low-resolution image processing methods (resolution above 5 meters). In recent years, with the development of satellite remote sensing technology, some satellites have emerged that can acquire high-resolution nighttime light remote sensing images (better than 1 meter resolution). At high resolution, there are higher requirements for the accuracy of geometric correction. At the same time, high resolution also provides the conditions for extracting road network models from nighttime urban data. Compared with high-resolution images, the previous low-resolution images could not extract relatively accurate road network models.

[0004] Therefore, due to the limitations of the low resolution and processing methods of previous satellite night light images, the accuracy and reliability of geometric correction for high-resolution satellite night light images cannot be guaranteed. At the same time, previous methods often use methods such as regional network adjustment that require iterative calculations, which are inefficient and unsuitable for commercial space remote sensing. Summary of the Invention

[0005] To address the limitations of previous geometric correction methods for nighttime light satellite remote sensing images, which are difficult to apply to high-resolution images, cannot guarantee accuracy and reliability, and suffer from low correction efficiency, this invention proposes a geometric correction method for satellite nighttime light images based on road network information control points.

[0006] To solve the above problems, the present invention adopts the following technical solution:

[0007] A method for geometric correction of satellite nighttime light images based on road network information control points, comprising the following steps:

[0008] Step 1: Establish a road network vector database;

[0009] Step 2: Acquire high-resolution color nighttime light satellite remote sensing images of the urban area;

[0010] Step 3: Preprocess the high-resolution color nighttime light satellite remote sensing image, including image enhancement, noise reduction, radiometric correction and preliminary geometric correction, to obtain the target nighttime light image;

[0011] Step 4: Use road detection methods to extract the road model from the target night light image, and use the intersections and bends of the main roads as feature points to extract feature points from the road model and filter out invalid feature points, finally obtaining the set of road network feature points;

[0012] Step 5: Generate a night light image road network intersection matrix An with the same resolution as the target night light image based on the road network feature point set. In the night light image road network intersection matrix An, the value of the position with intersection or bend is 1, and the value of the other positions is 0.

[0013] Step 6: Take the edge vector P0 of the target night light image, extend the edge vector P0 by N times to obtain vector P1, use vector P1 to retrieve all road network information inside vector P1 in the road network vector database, and generate an open source road network vector intersection matrix Av with an image pixel size N times that of the target night light image. The open source road network vector intersection matrix Av has a value of 1 at the position of road network intersection or bend point, and a value of 0 at the other positions.

[0014] Step 7: Perform a sliding window calculation on the open-source road network vector intersection matrix Av using the nighttime light image road network intersection matrix An. Move one pixel at a time, and record the sum of points at the same position in both the nighttime light image road network intersection matrix An and the open-source road network vector intersection matrix Av that are all 1. Record this sum as the window matching value Mw. After the nighttime light image road network intersection matrix An has finished sliding on the open-source road network vector intersection matrix Av, use the position of the nighttime light image road network intersection matrix An corresponding to the maximum value of the window matching value Mw as the true geographical location of the target nighttime light image. Update the coordinate values ​​of the corresponding points in the nighttime light image road network intersection matrix An with the coordinate values ​​of each point in the open-source road network vector intersection matrix Av corresponding to the maximum value of the window matching value Mw, thus completing the geometric correction of the nighttime light satellite remote sensing image.

[0015] This invention provides an effective method for geometric correction of high-resolution nighttime light satellite remote sensing images. It is primarily applicable to high-resolution color nighttime light images (resolution better than 1 meter). The method can clearly identify the road network in a city at night from the nighttime light image, extract the road model, and then determine the location of the control points by matching the feature points in the road model with existing high-precision road network control points, thereby achieving the purpose of geometric correction of the entire nighttime light image. The beneficial effects of this invention are:

[0016] (1) The present invention uses the method of matching the extracted set of road network feature points with the known and determined set of high-precision road network control points, which can effectively ensure the accuracy and reliability of geometric correction for high-resolution nighttime light satellite remote sensing images.

[0017] (2) When extracting road network feature points, the present invention does not require long-term calculation in actual processing, thus improving the efficiency of image geometric correction. At the same time, the present invention utilizes a pre-built road network vector database to extract control points from pre-downloaded open-source map vectors, realizing an automated and accurate feature point search method, which improves the processing speed of geometric correction of night light images. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a satellite nighttime light image geometric correction method based on road network information control points, as described in an embodiment of the present invention.

[0019] Figure 2 To extract target nighttime light images of road network feature points;

[0020] Figure 3 for Figure 2 The corresponding nighttime luminous image road network intersection matrix An;

[0021] Figure 4 These are open-source road network vectors retrieved from a road network vector database;

[0022] Figure 5 for Figure 4 The corresponding open-source road network vector intersection matrix Av. Detailed Implementation

[0023] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0024] This embodiment proposes a geometric correction method for nighttime light satellite remote sensing images based on control points combined with open-source road network data. This method is generally limited to processing high-resolution (better than 1 meter) color nighttime light satellite remote sensing images. Furthermore, this method is generally only applicable to urban remote sensing nighttime light images, which are typically acquired only within urban areas with lights. Moreover, within these urban areas, relatively sufficient and high-precision road network information can be found on the internet.

[0025] See Figure 1 The method proposed in this embodiment specifically includes the following steps:

[0026] Step 1: Establish a road network vector database.

[0027] Establishing a road network vector database specifically includes the following steps:

[0028] First, download open-source map vectors from the internet, such as Google Maps or OSM. Then, extract city and surrounding road network vectors from these vectors, typically using GeoJSON format, and store them in a database with spatial retrieval capabilities, such as PostGIS (a PostgreSQL extension). Filter the road network vectors, removing overly complex descriptions and secondary roads, retaining only major roads to ensure efficient later comparisons. The vector information stored in the database provides accurate geographic information, serving as the basis for correcting nighttime images.

[0029] Step 2: Acquire high-resolution color nighttime light satellite remote sensing images of the urban area, i.e., the images to be corrected. The resolution of the images is generally less than or equal to 1 meter.

[0030] Step 3: Preprocessing. Preprocessing operations are performed on the images acquired in Step 2, including image enhancement, noise reduction, radiometric correction, and preliminary geometric correction, to eliminate distortions caused by surface curvature and satellite side-slip.

[0031] Before performing geometric correction on the acquired high-resolution color nighttime light satellite remote sensing images, the images are preprocessed, including image enhancement, noise reduction, radiometric correction, and preliminary geometric correction. The specific steps and procedures are determined by the satellite model, imaging conditions, and mode. After preprocessing, the corresponding target nighttime light image is obtained.

[0032] Step 4: Extract road network feature points from the target nighttime light image.

[0033] Step 4.1: Extract the road model from the target night-light image using road detection methods. Many mature computer vision processing methods can be used to extract the road model from the target night-light image, such as color vector analysis, texture and shape analysis, edge detection, deep learning, and optical flow analysis. These existing image vision techniques and algorithms can effectively extract the road model from the target night-light image. In practice, deep learning methods combined with other auxiliary methods are generally used to extract the road model. Through deep learning, the road network feature points of the target night-light image are pre-processed through training, reducing the need for lengthy computations during actual processing and improving correction efficiency.

[0034] Furthermore, the process of extracting the road model from the target nighttime light image using road detection methods includes the following steps:

[0035] First, read the target night-light image and convert it into a grayscale image;

[0036] Then, a high-pass filter is applied to the grayscale image to enhance the edges in the image;

[0037] Next, the Canny edge detector is used to detect edges in the image;

[0038] Finally, morphological operations (closing operations) are used to connect the broken edges to extract the road model and obtain a continuous road network.

[0039] Step 4.2: Extract feature points from the road model. Feature points are generally selected from intersections of main roads or certain bends within a main road, and invalid feature points are filtered out, mainly those not at road intersections and those that do not meet the road calibration threshold. These can be dynamically changed according to the specific imaging parameters to finally obtain the set of road network feature points. The target nighttime image after feature point extraction is shown below. Figure 2 As shown.

[0040] Step 5: Generate the intersection matrix An of the road network in the nighttime light image.

[0041] A road network intersection matrix An is generated based on the set of road network feature points, with the same resolution as the target night light image. In the road network intersection matrix An, the value of the position with intersection or bend is 1, and the value of the other positions is 0. Figure 2 The corresponding nighttime light image road network intersection matrix An is as follows Figure 3 As shown in the figure, the value at the position marked by the red circle is 1, and the values ​​at the other positions are zero.

[0042] Step 6: Take the edge vector P0 of the target nighttime light image. Extend the edge vector P0 by N times to obtain vector P1, where N ranges from 1.2 to 1.5. The purpose of the extension is to ensure that the road network in P1, even with errors in the geometric coordinates of the nighttime light image, covers the road area in the retrieved road network region. Use vector P1 to retrieve all road network information within vector P1 from the road network vector database to obtain the open-source road network vector, such as... Figure 4 As shown, an open-source road network vector intersection matrix Av is generated based on the open-source road network vectors. Av has an image pixel size N times that of the target nighttime image. In the open-source road network vector intersection matrix Av, the values ​​at road network intersections or bends are 1, and the values ​​at other positions are 0. For example... Figure 5 As shown, Figure 5 Only a portion of the open-source road network vector intersection matrix Av is shown. Taking a target nighttime light image with a pixel size of 12000×5000 and N = 1.5 as an example, the size of the open-source road network vector intersection matrix Av is 18000×7500. Taking a target nighttime light image with a pixel size of 12000×5000 and N = 1.2 as an example, the size of the open-source road network vector intersection matrix Av is 14400×6000.

[0043] In this step, a road network matrix Av can also be generated by selecting a region in the open-source road network vector that is 1.2 times the size of the target night light image, based on the coordinates of the center point of the night light image. Here, N = 1.2 is the error range considering the inaccuracy of the original geometric positioning of the night light image. If the error is large, the value of N can be adjusted from 1.2 to 1.5 or 2 or other reasonable values.

[0044] The edge vector P0 of the target night light image is in the form of a polygon in GeoJson and can be generated using the four corner points of the image. There are two methods to extract the coordinates of the four corner points of the image: the first is to calculate the coordinates of the four corner points of the image from the image auxiliary data, and the second is to calculate the coordinates of the four corner points of the image through the center point, satellite position and imaging attitude.

[0045] Step 7: Match the road network feature points with the existing high-precision road network control points.

[0046] This step uses a sliding window calculation on the road network intersection matrix An of the nighttime light image within the open-source road network vector intersection matrix Av to find the optimal matching position of the road network intersection matrix An in the open-source road network vector intersection matrix Av.

[0047] Specifically, a sliding window calculation is performed on the road network intersection matrix An of the nighttime light image within the open-source road network vector intersection matrix Av. The road network intersection matrix An starts sliding from position [0, 0], moving one pixel at a time. Each movement records the sum or total number of points at the same position in both the nighttime light image road network intersection matrix An and the open-source road network vector intersection matrix Av that are all 1. This sum or total is recorded as the window matching value Mw. After the nighttime light image road network intersection matrix An has finished sliding on the open-source road network vector intersection matrix Av, the position of the nighttime light image road network intersection matrix An corresponding to the maximum value Max(Mw) of the window matching value Mw is the true geographical location of the target nighttime light image. The coordinates of each point in the open-source road network vector intersection matrix Av corresponding to the maximum value Max(Mw) of the window matching value Mw are used to update the coordinates of the corresponding points in the nighttime light image road network intersection matrix An, thereby completing the geometric correction of the nighttime light satellite remote sensing image.

[0048] The beneficial effects of this invention include:

[0049] (1) This invention provides an effective method for geometric correction of high-resolution nighttime light satellite remote sensing images. It is mainly applicable to high-resolution color nighttime light images (resolution better than 1 meter). It can clearly identify the road network in the city at night from the nighttime light image and extract the road model. Then, by matching the feature points in the road model with existing high-precision road network control points, the position of the control points is determined, thereby achieving the purpose of geometric correction of the entire nighttime light image.

[0050] (2) The present invention uses the method of matching the extracted set of road network feature points with the known and determined set of high-precision road network control points, which can effectively ensure the accuracy and reliability of geometric correction for high-resolution nighttime light satellite remote sensing images.

[0051] (3) When extracting road network feature points, the present invention does not require long-term calculations in actual processing, thus improving the efficiency of image geometric correction. At the same time, the present invention utilizes a pre-built road network vector database to extract control points from pre-downloaded open-source map vectors, realizing an automated and accurate feature point search method, which improves the processing speed of geometric correction of night light images.

[0052] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0053] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for geometric correction of satellite nighttime light images based on road network information control points, characterized in that, Includes the following steps: Step 1: Establish a road network vector database; Step 2: Acquire high-resolution color nighttime light satellite remote sensing images of the urban area; Step 3: Preprocess the high-resolution color nighttime light satellite remote sensing image, including image enhancement, noise reduction, radiometric correction and preliminary geometric correction, to obtain the target nighttime light image; Step 4: Use road detection methods to extract the road model from the target night light image, and use the intersections and bends of the main roads as feature points to extract feature points from the road model and filter out invalid feature points, finally obtaining the set of road network feature points; Step 5: Generate a night light image road network intersection matrix An with the same resolution as the target night light image based on the road network feature point set. In the night light image road network intersection matrix An, the value of the position with intersection or bend is 1, and the value of the other positions is 0. Step 6: Take the edge vector P0 of the target night light image, extend the edge vector P0 by N times to obtain vector P1, use vector P1 to retrieve all road network information inside vector P1 in the road network vector database, and generate an open source road network vector intersection matrix Av with an image pixel size N times that of the target night light image. The open source road network vector intersection matrix Av has a value of 1 at the position of road network intersection or bend point, and a value of 0 at the other positions. Step 7: Perform a sliding window calculation on the open-source road network vector intersection matrix Av using the nighttime light image road network intersection matrix An. Move one pixel at a time, and record the sum of points at the same position in both the nighttime light image road network intersection matrix An and the open-source road network vector intersection matrix Av that are all 1. Record this sum as the window matching value Mw. After the nighttime light image road network intersection matrix An has finished sliding on the open-source road network vector intersection matrix Av, use the position of the nighttime light image road network intersection matrix An corresponding to the maximum value of the window matching value Mw as the true geographical location of the target nighttime light image. Update the coordinate values ​​of the corresponding points in the nighttime light image road network intersection matrix An with the coordinate values ​​of each point in the open-source road network vector intersection matrix Av corresponding to the maximum value of the window matching value Mw, thus completing the geometric correction of the nighttime light satellite remote sensing image.

2. The method for geometric correction of satellite nighttime light images based on road network information control points according to claim 1, characterized in that, Step 1 includes the following steps: Download open-source vector maps from the internet; Extract city and surrounding road network vectors from open-source map vectors, and store the extracted road network vectors into a database with spatial retrieval capabilities; Filter the road network vectors stored in the database.

3. The method for geometric correction of satellite nighttime light images based on road network information control points according to claim 2, wherein the city and surrounding road network vectors are identified in GeoJSON format, and the database is a PostGIS database.

4. The method for geometric correction of satellite nighttime light images based on road network information control points according to claim 2, characterized in that, The open-source map vector is either Google Maps or OSM.

5. The method for geometric correction of satellite nighttime light images based on road network information control points according to claim 2, characterized in that, The database with spatial retrieval capabilities is the PostGIS database.

6. A method for geometric correction of satellite nighttime light images based on road network information control points according to any one of claims 1 to 5, characterized in that, The process of extracting the road model from a target nighttime light image using road detection methods includes the following steps: Convert the target night-light image to a grayscale image; A high-pass filter is applied to the grayscale image to enhance the edges in the image; Use the Canny edge detector to detect edges in an image; By connecting disconnected edges through morphological operations, a road model can be extracted to obtain a continuous road network.

7. A method for geometric correction of satellite nighttime light images based on road network information control points according to any one of claims 1 to 5, characterized in that, The road detection method can be any one of color vector analysis, texture and shape analysis, edge detection, deep learning, or optical flow analysis.

8. A method for geometric correction of satellite nighttime light images based on road network information control points according to any one of claims 1 to 5, characterized in that, In step 6, the edge vector P0 of the target night-light image is in the form of a GeoJson polygon, and the edge vector P0 is generated using the four corner points of the image.

9. A method for geometric correction of satellite nighttime light images based on road network information control points according to any one of claims 1 to 5, characterized in that, The value of N ranges from 1.2 to 2.

10. A method for geometric correction of satellite nighttime light images based on road network information control points according to any one of claims 1 to 5, characterized in that, The resolution of the high-resolution color nighttime light satellite remote sensing image is less than or equal to 1 meter.

Images