A method and device for selecting edge area of ​​remote sensing image

Abstract

The present invention discloses a method and device for selecting a blade edge region of a remote sensing image. The method comprises: extracting a region of interest ROI image containing a blade edge region of a ground object in a single-band image of a remote sensing image based on a priori knowledge base of the blade edge region; extracting an edge binary image corresponding to the ROI image; restoring the fragmented edges of the edge binary image based on a connected restoration algorithm to obtain a restored edge image; extracting a blade edge region of the restored edge image based on the priori knowledge base of the blade edge region and an edge straight line; determining the quality of the blade edge region based on the priori knowledge base of the blade edge region; and iteratively extracting the target blade edge region in the remote sensing image based on the quality. The present invention can improve the accuracy of blade edge detection and reduce the complexity of manual participation.

Description

Technical Field

[0001] The invention relates to the technical field of aerospace remote sensing image processing, in particular to a method and a device for selecting a remote sensing image edge region. Background Art

[0002] Modulation Transfer Function (MTF) is a reflection of image resolution. The change of remote sensing image MTF is an important indicator for evaluating remote sensing image quality, remote sensing satellite camera on-orbit operation and performance change. The knife edge method is the best method to calculate remote sensing image MTF, and the accuracy of using this method to calculate image MTF lies in the selection of the image knife edge area.

[0003] At present, black and white target areas are mainly selected as the edge areas of remote sensing images, such as Baotou target, Songshan target, Yuxi target, etc., but the number of remote sensing images covering these black and white target fields is small, and it is impossible to calculate and evaluate the MTF curve of the satellite camera in orbit in real time. And for some remote sensing satellites with medium and low spatial resolution, such as the GF-4 satellite with a resolution of 50m, it is impossible to clearly identify black and white targets. Secondly, there are now dozens of domestic satellites in orbit, and more than hundreds of images are acquired every day. In the absence of black and white targets, the method of manually selecting the edge area of ​​remote sensing images is time-consuming and laborious, and is not suitable for real-time monitoring of the MTF of a large number of remote sensing images. In addition, the edge edges extracted directly in large-scale medium and low spatial resolution remote sensing images containing complex information using edge detection and other algorithms are too fragmented, and the edge area used for MTF calculation cannot be well determined. Summary of the invention

[0004] The technical problem solved by the present invention is to overcome the deficiencies of the prior art and provide a method and device for selecting a remote sensing image edge region.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for selecting a knife edge area of ​​a remote sensing image, comprising:

[0006] Based on the prior knowledge base of edge region, the ROI image of the region of interest containing the edge region of the ground object in the single-band image of the remote sensing image is extracted;

[0007] Extracting an edge binary image corresponding to the ROI image;

[0008] Restoring the fragmented edges of the edge binary image based on a connected restoration algorithm to obtain a restored edge image;

[0009] Extracting the edge region of the restored edge image based on the edge region prior knowledge base and the edge straight line;

[0010] Determining the quality of the edge region according to the edge region prior knowledge base;

[0011] Based on the quality level, a target edge region in the remote sensing image is extracted iteratively.

[0012] Optionally, before extracting the region of interest ROI image containing the edge region of the ground object in the single-band image of the remote sensing image based on the edge region prior knowledge base, the method further includes:

[0013] According to the features of the knife edge region required by the modulation transfer function curve corresponding to the remote sensing image obtained by calculation, a priori knowledge base of the knife edge region is established.

[0014] Optionally, the knife edge region prior knowledge base includes:

[0015] The inclination angle of the blade edge in the vertical or horizontal direction is between 2.5° and 6°;

[0016] The number of pixels on the edge of the blade is greater than 10;

[0017] The grayscale contrast of the area on both sides of the blade is greater than the grayscale threshold and is evenly distributed, and the noise is lower than the noise threshold;

[0018] The width of the area on both sides of the blade is greater than or equal to 10 pixels.

[0019] Optionally, extracting an edge binary image corresponding to the ROI image includes:

[0020] Based on the Sobel operator, obtaining the first-order gradient of the ROI image;

[0021] The OSTU algorithm is used to obtain the edge of the first-order gradient of the ROI image according to a set threshold value to obtain the edge binary image.

[0022] Optionally, the restoring the fragmented edges of the edge binary image based on a connected restoration algorithm to obtain a restored edge image includes:

[0023] Using an 8-connected region labeling method, determining each connected edge in the edge binary image;

[0024] Calculate the number of pixels of each connected edge respectively, and regard the connected edges with less than 15 pixels as suspected fragmented connected edges;

[0025] Calculate the distance from each suspected fragmented edge to other edges respectively, and take the suspected fragmented edge with the minimum distance less than 10 pixels as the fragmented edge;

[0026] Obtain the connected edge closest to each fragmented edge, calculate the 4 pixel points closest to the two, obtain the fitting line using the least squares method, and obtain the area requiring edge restoration based on the fitting line;

[0027] The first-order gradient image in the area to be restored is extracted based on the Sobel operator, and the segmentation threshold is automatically determined using the OSTU algorithm to extract the edge of the area to be restored.

[0028] The fragmented edges are connected to the edges of the area to be restored to obtain the restored edge image.

[0029] Optionally, determining the quality of the edge region according to the edge region prior knowledge base includes:

[0030] The extracted edge regions are evaluated according to the edge region prior knowledge base, and the average grayscale values ​​of pixels on both sides of the edge and the standard deviation of the grayscale values ​​of pixels on both sides of the edge of each edge region are calculated respectively.

[0031] Optionally, the iteratively extracting the target edge region in the remote sensing image based on the quality degree includes:

[0032] Determine the average value and standard deviation of each edge area to obtain a determination result;

[0033] extracting the target blade edge region in the remote sensing image according to the determination result;

[0034] Iteratively execute the step of evaluating the extracted edge area according to the edge area prior knowledge base, respectively calculating the average grayscale value of pixels on both sides of the edge of each edge area, and the standard deviation of the grayscale value of pixels on both sides of the edge, until the step of extracting the target edge area in the remote sensing image according to the judgment result, until the number of pixels of the edge straight line is less than 10.

[0035] In order to solve the above technical problems, an embodiment of the present invention provides a device for selecting a knife edge area of ​​a remote sensing image, comprising:

[0036] ROI image extraction module, used to extract the ROI image of the region of interest containing the edge of the ground object in the single-band image of the remote sensing image based on the prior knowledge base of the edge area;

[0037] An edge binary image extraction module is used to extract an edge binary image corresponding to the ROI image;

[0038] A restored edge image acquisition module is used to restore the fragmented edges of the edge binary image based on a connected restoration algorithm to obtain a restored edge image;

[0039] A knife edge region extraction module, used for extracting the knife edge region of the restored edge image based on the knife edge region prior knowledge base and the edge straight line;

[0040] A quality determination module, used for determining the quality of the edge region according to the edge region prior knowledge base;

[0041] The target edge region extraction module is used to iteratively extract the target edge region in the remote sensing image based on the quality level.

[0042] Optionally, the device further comprises:

[0043] The prior knowledge base building module is used to build the prior knowledge base of the edge region according to the features of the edge region required by the modulation transfer function curve corresponding to the remote sensing image obtained by calculation.

[0044] Optionally, the knife edge region prior knowledge base includes:

[0045] The inclination angle of the blade edge in the vertical or horizontal direction is between 2.5° and 6°;

[0046] The number of pixels on the edge of the blade is greater than 10;

[0047] The grayscale contrast of the area on both sides of the blade is greater than the grayscale threshold and is evenly distributed, and the noise is lower than the noise threshold;

[0048] The width of the area on both sides of the blade is greater than or equal to 10 pixels.

[0049] Optionally, the edge binary image extraction module includes:

[0050] An image first-order gradient acquisition unit, used for acquiring the first-order gradient of the ROI image based on a Sobel operator;

[0051] The edge binary image acquisition unit is used to acquire the edge of the first-order gradient of the ROI image according to a set threshold value by using the OSTU algorithm to obtain the edge binary image.

[0052] Optionally, the restored edge image acquisition module includes:

[0053] A connected edge determination unit, used to determine each connected edge in the edge binary image by using an 8-connected region labeling method;

[0054] A fragmented connected edge acquisition unit is used to calculate the number of pixels of each connected edge, and regard the connected edges with the number of pixels less than 15 as suspected fragmented connected edges;

[0055] A fragmentation edge acquisition unit, used to calculate the distance between each suspected fragmentation edge and other edges, and to take the suspected fragmentation edge with a minimum distance less than 10 pixels as a fragmentation edge;

[0056] A restoration area acquisition unit is used to obtain the connected edge closest to each fragmented edge, calculate the 4 pixel points closest to the two, obtain a fitting straight line using the least squares method, and obtain the area requiring edge restoration based on the fitting straight line;

[0057] The edge extraction unit of the area to be restored is used to extract the first-order gradient image in the area to be restored based on the Sobel operator, and use the OSTU algorithm to automatically determine the segmentation threshold to extract the edge of the area to be restored;

[0058] The restored edge image acquisition unit is used to connect the fragmented edges with the edges of the area to be restored to obtain a restored edge image.

[0059] Optionally, the quality determination module includes:

[0060] The mean value and standard deviation calculation unit is used to evaluate the extracted edge area according to the edge area prior knowledge base, and respectively calculate the mean value of the pixel grayscale values ​​on both sides of the edge of each edge area and the standard deviation of the pixel grayscale values ​​on both sides of the edge.

[0061] Optionally, the target edge region extraction module includes:

[0062] A determination result acquisition unit, used for determining the average value and standard deviation of each edge region to obtain a determination result;

[0063] A target edge region extraction unit, used for extracting the target edge region in the remote sensing image according to the determination result;

[0064] The iterative execution unit is used to iteratively execute the mean value standard deviation calculation unit, the determination result acquisition unit and the target edge region extraction unit until the number of pixels of the edge line is less than 10.

[0065] The advantages of the present invention compared with the prior art are:

[0066] 1) The established prior knowledge base of the knife edge area is applicable to a variety of low and medium spatial resolution remote sensing images, which can improve the accuracy of knife edge detection;

[0067] 2) The proposed connectivity restoration algorithm can restore the fragmented edges extracted from large remote sensing images, and on this basis, a more complete edge area can be obtained;

[0068] 3) The algorithm process can quickly obtain multiple high-quality edge areas in remote sensing images, which is suitable for the MTF curve calculation of a large number of medium and low resolution remote sensing images, and can obtain the continuous change of satellite camera MTF in a short time;

[0069] 4) The algorithm for iteratively extracting edge regions can remove high-noise regions to the greatest extent and retain high-quality edge regions;

[0070] 5) The algorithm has a high degree of automation, which can reduce the complexity of human participation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A flowchart of a method for selecting a knife edge region of a remote sensing image provided by an embodiment of the present invention;

[0072] Figure 2 A schematic diagram of a knife edge area selection process provided by an embodiment of the present invention;

[0073] Figure 3 A schematic diagram of a ROI region extraction result provided by an embodiment of the present invention;

[0074] Figure 4 A schematic diagram of edge extraction and edge restoration results provided by an embodiment of the present invention;

[0075] Figure 5 A schematic diagram of a high-quality edge region extraction result provided by an embodiment of the present invention;

[0076] Figure 6 A schematic structural diagram of a device for selecting a knife edge region of a remote sensing image provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0077] Embodiment 1

[0078] Reference Figure 1 , which shows a flowchart of the steps of a method for selecting a remote sensing image edge region provided by an embodiment of the present invention, such as Figure 1 As shown, the method may include the following steps:

[0079] Step 101: extracting a region of interest ROI image containing the edge region of the ground object in a single-band image of a remote sensing image based on a priori knowledge base of the edge region.

[0080] The technical solution of the embodiment of the present invention can be combined with Figure 2 to Figure 5 The following description is given.

[0081] In the embodiment of the present invention, the knife edge region prior knowledge base refers to a pre-established knowledge base for extracting knife edge regions in satellite remote sensing images.

[0082] The process of establishing the prior knowledge base of the edge region can be described in detail in combination with the following specific implementation method.

[0083] In a specific implementation of the present invention, before the above step 101, the following steps may also be included:

[0084] Step A1: Establishing a priori knowledge base of the knife edge region according to the features of the knife edge region required by the modulation transfer function curve corresponding to the remote sensing image obtained by calculation.

[0085] In this embodiment, the following prior knowledge base of the knife edge region may be established according to the characteristics of the knife edge region required for calculating the MTF curve of the remote sensing image.

[0086] Knowledge (1): The inclination angle of the blade edge in the vertical or horizontal direction should be between 2.5° and 6°;

[0087] Knowledge (2): The number of pixels on the edge of the blade is more than 10;

[0088] Knowledge (3): The grayscale contrast of the areas on both sides of the blade is large and the distribution is more uniform, with less noise;

[0089] Knowledge (4): The width of the area on both sides of the blade must not be less than 10 pixels.

[0090] After establishing a priori knowledge base of blade edge regions, the priori knowledge base of blade edge regions can be used to extract the ROI image of the region of interest containing the blade edge region of the object in the single-band image of the remote sensing image. Specifically, according to knowledge (3), it can be concluded that the objects with relatively uniform grayscale distribution in the remote sensing image include rivers, bare soil, farmland, buildings, etc. Among them, rivers and bare soil on the banks, farmland growing different crops, building roofs and shadows, etc. can all form areas with large grayscale contrast. Therefore, in order to reduce the influence of other object edges on the extraction of the edges of these four objects, a classification model composed of MNDWI and NDVI decision trees is established to segment the multispectral remote sensing image. The segmented area is binarized and the buffer zone is calculated. The radius of the segmented area buffer zone is different for images with different spatial resolutions. According to knowledge (4), the radius of the buffer zone must be at least 10 times greater than the spatial resolution. In order to extract as many blade edge areas as possible, this patent sets the radius of the buffer zone to 15 times the spatial resolution. Using the buffer zone, the ROI image containing the blade edge area of ​​the object is extracted from the single-band image of the remote sensing image (the ROI extraction result can be as shown in the figure). Figure 3 shown).

[0091] After extracting the region of interest ROI image containing the edge region of the ground object in the single-band image of the remote sensing image based on the edge region prior knowledge base, step 102 is executed.

[0092] Step 102: extracting the edge binary image corresponding to the ROI image.

[0093] After extracting the region of interest ROI image containing the edge region of the ground object in the single-band image of the remote sensing image based on the edge region prior knowledge base, the edge binary image corresponding to the ROI image can be extracted. Specifically, it can be described in detail in combination with the following specific implementation method.

[0094] In a specific implementation of the present invention, the above step 102 may include:

[0095] Sub-step B1: obtaining the first-order gradient of the ROI image based on the Sobel operator;

[0096] Sub-step B2: using the OSTU algorithm to obtain the edge of the first-order gradient of the ROI image according to a set threshold value to obtain the edge binary image.

[0097] In this embodiment, the Sobel operator can be used to obtain the first-order gradient of the ROI image, and the OSTU algorithm can be used to automatically set the threshold to obtain the edge of the ROI area of ​​the remote sensing image.

[0098] After the edge binary image corresponding to the ROI image is extracted, step 103 is executed.

[0099] Step 103: Restoring the fragmented edges of the edge binary image based on a connected restoration algorithm to obtain a restored edge image.

[0100] After the edge binary image corresponding to the ROI image is extracted, the fragmented edges of the edge binary image can be restored based on the connected restoration algorithm to obtain a restored edge image. This process can be described in detail in conjunction with the following specific implementation method.

[0101] In a specific implementation of the present invention, the above step 103 may include:

[0102] Sub-step C1: using the 8-connected region labeling method, determining each connected edge in the edge binary image;

[0103] Sub-step C2: Calculate the number of pixels of each connected edge respectively, and regard the connected edges with less than 15 pixels as suspected fragmented connected edges;

[0104] Sub-step C3: Calculate the distance between each suspected fragmented edge and other edges respectively, and take the suspected fragmented edge with the minimum distance less than 10 pixels as the fragmented edge;

[0105] Sub-step C4: obtaining the connected edge closest to each fragmented edge, calculating the 4 pixel points closest to the two, obtaining a fitting straight line using the least squares method, and obtaining the area requiring edge restoration based on the fitting straight line;

[0106] Sub-step C5: extracting the first-order gradient image in the area to be restored based on the Sobel operator, and automatically determining the segmentation threshold using the OSTU algorithm to extract the edge of the area to be restored;

[0107] Sub-step C6: Connect the fragmented edges with the edges of the area to be restored to obtain a restored edge image.

[0108] In this embodiment, a connectivity restoration algorithm is developed to restore the fragmented edges extracted by the Sobel operator. The specific algorithm steps are as follows:

[0109] 1) Use the 8-connected region labeling method to determine each connected edge in the ROI, remove the connected edges with less than 2 pixels, and label the remaining connected edges as 1, 2, 3..., n;

[0110] 2) Calculate the number of pixels of each connected edge. According to knowledge (2), if the number of pixels is less than 15, it is considered as a suspected fragmented connected edge.

[0111] 3) Re-label the suspected fragmented connected edges as i1, i2, i3, ..., in. Calculate the distance from each suspected fragmented edge to other edges respectively. If the minimum distance is less than 10 pixels, the edge is considered a fragmented edge, otherwise it is considered a noise edge and removed;

[0112] 4) Re-label the fragmented connected edges as j1, j2, j3, ..., jn. Obtain the connected edge closest to each fragmented edge, calculate the 4 pixels closest to the two, use the least squares method to obtain the fitting line, take the direction perpendicular to the fitting line as the expansion direction, and start from both ends of the line to expand 20 pixels to both sides to obtain the area that needs edge restoration;

[0113] 5) Use the Sobel operator to extract the first-order gradient image in the area to be restored, and use the OSTU algorithm to automatically determine the segmentation threshold to extract the edges X1, X2, X3..., Xn of the area to be restored;

[0114] 6) Connect the fragmented edges j1, j2, j3, ..., jn with the edges of the area to be restored X1, X2, X3 ..., Xn, respectively, to obtain the restored edges Y1, Y2, Y3 ..., Yn (the edge extraction and edge restoration results can be as shown in Figure 4 shown).

[0115] After the fragmented edges of the edge binary image are restored based on the connected restoration algorithm to obtain the restored edge image, step 104 is executed.

[0116] Step 104: extracting the edge region of the restored edge image based on the edge region prior knowledge base and the edge straight line.

[0117] After the restored edge image is obtained, the edge region of the restored edge image can be extracted based on the prior knowledge base of the edge region and the edge line. Specifically, the restored edge image is subjected to Hough transform to obtain the edge line, and the edge line is marked with 8 connected regions, labeled 1, 2, 3..., n. Each edge line is judged: the number of pixels of each connected edge is calculated, and whether it is suitable as a knife edge is determined according to knowledge (2); a linear line is fitted according to the pixels of each connected edge line, and the inclination angle of the edge with the along-track direction and the perpendicular-track direction is determined, and whether the connected edge is suitable as a knife edge is determined according to knowledge (1). The edge must satisfy both knowledge (1) and knowledge (2), otherwise the edge will be eliminated.

[0118] After determining the blade edge, obtain the two endpoints of the edge respectively. Take the two endpoints as the starting point. If the edge is a horizontally inclined edge, the extension direction is along the track. If the edge is a vertically inclined edge, the extension direction is perpendicular to the track. Based on knowledge (4), each is extended by 10 pixels as the blade edge area.

[0119] After the edge region of the edge image is restored based on the edge region prior knowledge base and the edge straight line extraction, step 105 is executed.

[0120] Step 105: Determine the quality of the edge region based on the edge region prior knowledge base.

[0121] After the edge region of the restored edge image is extracted, the quality of the edge region can be determined based on the prior knowledge base of the edge region. Specifically, after the edge region is extracted, the extracted edge region is evaluated based on knowledge (3). The average values ​​V1 and V2 of the grayscale values ​​of the pixels on both sides of the edge of each edge region are calculated respectively; the standard deviations S1 and S2 of the grayscale values ​​of the pixels on both sides of the edge are calculated.

[0122] Step 106: Iteratively extract the target edge region in the remote sensing image based on the quality score.

[0123] The target edge area is the high-quality edge area, that is.

[0124] Determine the values of V1, V2, S1, and S2 for each edge region. If V1 >> V2 or V1 << V2 and S1 < a, S2 < a are satisfied, it is determined that the quality of the edge region is good. Through experiments, the value of a is set to 8. If the above conditions are not met, remove the two end points, continue to expand with the new two end points, and continue the determination in step 6. If the conditions are met, stop the iteration; if the number of pixels of the edge line is less than 10, stop the iteration, and finally obtain the high-quality edge region (the extraction result of the high-quality edge region can be as shown in Figure 5 ).

[0125] Embodiment 2

[0126] Refer to Figure 6 , which shows the structural schematic diagram of a device for selecting an edge region of a remote sensing image provided by an embodiment of the present invention. As shown in Figure 6 , the device may include the following steps:

[0127] The ROI image extraction module 610 is used to extract the region of interest ROI image containing the ground object edge region in the single-band image of the remote sensing image based on the edge region prior knowledge base;

[0128] The edge binary image extraction module 620 is used to extract the edge binary image corresponding to the ROI image;

[0129] The restored edge image acquisition module 630 is used to perform restoration processing on the fragmented edges of the edge binary image based on the connected restoration algorithm to obtain a restored edge image;

[0130] The edge region extraction module 640 is used to extract the edge region of the restored edge image based on the edge region prior knowledge base and the edge line;

[0131] The quality determination module 650 is used to determine the quality of the edge region according to the edge region prior knowledge base;

[0132] The target edge region extraction module 660 is used to iteratively extract the target edge region in the remote sensing image based on the quality;

[0133] Optionally, the device further includes:

[0134] The prior knowledge base establishment module is used to establish the edge region prior knowledge base according to the characteristics of the edge region required by the modulation transfer function curve corresponding to the calculated remote sensing image.

[0135] Optionally, the edge region prior knowledge base includes:

[0136] The inclination angle of the edge in the vertical or horizontal direction is between 2.5° and 6°;

[0137] The number of pixels on the edge of the blade is greater than 10;

[0138] The grayscale contrast of the area on both sides of the blade is greater than the grayscale threshold and is evenly distributed, and the noise is lower than the noise threshold;

[0139] The width of the area on both sides of the blade is greater than or equal to 10 pixels.

[0140] Optionally, the edge binary image extraction module includes:

[0141] An image first-order gradient acquisition unit, used for acquiring the first-order gradient of the ROI image based on a Sobel operator;

[0142] The edge binary image acquisition unit is used to acquire the edge of the first-order gradient of the ROI image according to a set threshold value by using the OSTU algorithm to obtain the edge binary image.

[0143] Optionally, the restored edge image acquisition module includes:

[0144] A connected edge determination unit, used to determine each connected edge in the edge binary image by using an 8-connected region labeling method;

[0145] A fragmented connected edge acquisition unit is used to calculate the number of pixels of each connected edge, and regard the connected edges with the number of pixels less than 15 as suspected fragmented connected edges;

[0146] A fragmentation edge acquisition unit, used to calculate the distance between each suspected fragmentation edge and other edges, and to take the suspected fragmentation edge with a minimum distance less than 10 pixels as a fragmentation edge;

[0147] A restoration area acquisition unit is used to obtain the connected edge closest to each fragmented edge, calculate the 4 pixel points closest to the two, obtain a fitting straight line using the least squares method, and obtain the area requiring edge restoration based on the fitting straight line;

[0148] The edge extraction unit of the area to be restored is used to extract the first-order gradient image in the area to be restored based on the Sobel operator, and use the OSTU algorithm to automatically determine the segmentation threshold to extract the edge of the area to be restored;

[0149] The restored edge image acquisition unit is used to connect the fragmented edges with the edges of the area to be restored to obtain a restored edge image.

[0150] Optionally, the quality determination module includes:

[0151] The mean value and standard deviation calculation unit is used to evaluate the extracted edge area according to the edge area prior knowledge base, and respectively calculate the mean value of the pixel grayscale values ​​on both sides of the edge of each edge area and the standard deviation of the pixel grayscale values ​​on both sides of the edge.

[0152] Optionally, the target edge region extraction module includes:

[0153] A determination result acquisition unit, used for determining the average value and standard deviation of each edge region to obtain a determination result;

[0154] A target edge region extraction unit, used for extracting the target edge region in the remote sensing image according to the determination result;

[0155] The iterative execution unit is used to iteratively execute the mean value standard deviation calculation unit, the determination result acquisition unit and the target edge region extraction unit until the number of pixels of the edge line is less than 10.

[0156] The specific implementation methods described in this application can enable those skilled in the art to understand this application more comprehensively, but do not limit this application in any way. Therefore, those skilled in the art should understand that this application can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and technical essence of this application should be included in the protection scope of the patent of this application.

[0157] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for selecting edge areas of remote sensing images, characterized in that: include: Based on the prior knowledge base of edge region, the ROI image of the region of interest containing the edge region of the ground object in the single-band image of the remote sensing image is extracted; Extracting an edge binary image corresponding to the ROI image; The fragmented edges of the edge binary image are restored based on the connected restoration algorithm to obtain a restored edge image; including: using the 8-connected region labeling method to determine each connected edge in the edge binary image; respectively calculating the number of pixels of each connected edge, and taking the connected edges with a number of pixels less than 15 as suspected fragmented connected edges; respectively calculating the distance from each suspected fragmented edge to other edges, and taking the suspected fragmented edges with a minimum distance less than 10 pixels as fragmented edges; obtaining the connected edge closest to each fragmented edge, calculating the 4 pixel points closest to the two, obtaining a fitting straight line using the least squares method, and obtaining the area to be restored based on the fitting straight line; extracting the first-order gradient image in the area to be restored based on the Sobel operator, and automatically determining the segmentation threshold using the OSTU algorithm to extract the edge of the area to be restored; connecting the fragmented edges with the edges of the area to be restored to obtain a restored edge image; Extracting the edge region of the restored edge image based on the edge region prior knowledge base and the edge straight line; Determining the quality of the edge region according to the edge region prior knowledge base; including: evaluating the extracted edge region according to the edge region prior knowledge base, and calculating the average grayscale value of pixels on both sides of the edge of each edge region, and the standard deviation of the grayscale value of pixels on both sides of the edge; Based on the quality, iteratively extract the target edge area in the remote sensing image; including: judging the average value and standard deviation of each edge area to obtain a judgment result; extracting the target edge area in the remote sensing image according to the judgment result; iteratively executing the step of evaluating the extracted edge area according to the edge area prior knowledge base, respectively calculating the average value of the pixel grayscale value on both sides of the edge of each edge area, and the standard deviation of the pixel grayscale value on both sides of the edge, until the step of extracting the target edge area in the remote sensing image according to the judgment result, until the number of pixels of the edge straight line is less than 10.

2. The method according to claim 1, characterized in that Before extracting the region of interest ROI image containing the edge region of the ground object in the single-band image of the remote sensing image based on the prior knowledge base of the edge region, the method further includes: According to the features of the knife edge region required by the modulation transfer function curve corresponding to the remote sensing image obtained by calculation, a priori knowledge base of the knife edge region is established.

3. The method according to claim 2, characterized in that The blade edge region prior knowledge base includes: The inclination angle of the blade edge in the vertical or horizontal direction is between 2.5° and 6°; The number of pixels on the edge of the blade is greater than 10; The grayscale contrast of the area on both sides of the blade is greater than the grayscale threshold and is evenly distributed, and the noise is lower than the noise threshold; The width of the area on both sides of the blade is greater than or equal to 10 pixels.

4. The method according to claim 1, characterized in that The step of extracting the edge binary image corresponding to the ROI image comprises: Based on the Sobel operator, obtaining the first-order gradient of the ROI image; The OSTU algorithm is used to obtain the edge of the first-order gradient of the ROI image according to a set threshold value to obtain the edge binary image.

5. A device for selecting edge areas of remote sensing images, characterized in that: include: ROI image extraction module, used to extract the ROI image of the region of interest containing the edge of the ground object in the single-band image of the remote sensing image based on the prior knowledge base of the edge region; An edge binary image extraction module is used to extract an edge binary image corresponding to the ROI image; The restored edge image acquisition module is used to restore the fragmented edges of the edge binary image based on the connected restoration algorithm to obtain a restored edge image; it includes: A connected edge determination unit, used to determine each connected edge in the edge binary image by using an 8-connected region labeling method; A fragmented connected edge acquisition unit is used to calculate the number of pixels of each connected edge, and regard the connected edges with the number of pixels less than 15 as suspected fragmented connected edges; A fragmentation edge acquisition unit, used to calculate the distance between each suspected fragmentation edge and other edges, and to take the suspected fragmentation edge with a minimum distance less than 10 pixels as a fragmentation edge; A restoration area acquisition unit is used to obtain the connected edge closest to each fragmented edge, calculate the 4 pixel points closest to the two, obtain a fitting straight line using the least squares method, and obtain the area requiring edge restoration based on the fitting straight line; The edge extraction unit of the area to be restored is used to extract the first-order gradient image in the area to be restored based on the Sobel operator, and use the OSTU algorithm to automatically determine the segmentation threshold to extract the edge of the area to be restored; A restored edge image acquisition unit is used to connect the fragmented edges with the edges of the area to be restored to obtain a restored edge image; A knife edge region extraction module, used for extracting the knife edge region of the restored edge image based on the knife edge region prior knowledge base and the edge straight line; A quality determination module is used to determine the quality of the edge region according to the edge region prior knowledge base; comprising: A mean value standard deviation calculation unit is used to evaluate the extracted edge region according to the edge region prior knowledge base, and respectively calculate the mean value of the pixel grayscale values ​​on both sides of the edge of each edge region, and the standard deviation of the pixel grayscale values ​​on both sides of the edge; A target edge region extraction module is used to iteratively extract the target edge region in the remote sensing image based on the quality level; comprising: A determination result acquisition unit, used for determining the average value and standard deviation of each edge region to obtain a determination result; A target edge region extraction unit, used for extracting the target edge region in the remote sensing image according to the determination result; The iterative execution unit is used to iteratively execute the mean value standard deviation calculation unit, the determination result acquisition unit and the target edge region extraction unit until the number of pixels of the edge line is less than 10.

6. The device according to claim 5, characterized in that The device also includes: The prior knowledge base building module is used to build the prior knowledge base of the edge region according to the features of the edge region required by the modulation transfer function curve corresponding to the remote sensing image obtained by calculation.

7. The device according to claim 6, characterized in that The blade edge region prior knowledge base includes: The inclination angle of the blade edge in the vertical or horizontal direction is between 2.5° and 6°; The number of pixels on the edge of the blade is greater than 10; The grayscale contrast of the area on both sides of the blade is greater than the grayscale threshold and is evenly distributed, and the noise is lower than the noise threshold; The width of the area on both sides of the blade is greater than or equal to 10 pixels.

8. The device according to claim 5, characterized in that The edge binary image extraction module comprises: An image first-order gradient acquisition unit, used for acquiring the first-order gradient of the ROI image based on a Sobel operator; The edge binary image acquisition unit is used to acquire the edge of the first-order gradient of the ROI image according to a set threshold value by using the OSTU algorithm to obtain the edge binary image.

Images