Remote sensing image airport automatic detection method and system, terminal and computer medium

Abstract

The invention discloses a remote sensing image airport automatic detection method and system, a terminal and a computer medium, and the method comprises the steps: carrying out the region segmentation of an original remote sensing image, and obtaining a superpixel region; performing linear detection on the original remote sensing image to obtain a linear extraction result; calculating a saliency map of the remote sensing image according to a linear extraction result; and calculating an airport contour by combining the original remote sensing image and the saliency map thereof, and obtaining an airport detection result according to the contour. The method can provide support for the fields of aircraft integrated navigation, military reconnaissance, accurate strike and the like; the negative influence of background interference on the edge extraction result is suppressed through the salient region, and the calculation efficiency and accuracy of the detection method are considered.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and particularly relates to a method, a system, a terminal and a computer medium for automatically detecting airports in remote sensing images. Background Art

[0002] The detection and recognition of typical targets in remote sensing images is one of the hotspots in remote sensing image interpretation. As a strategic target for military and civilian use, quickly and accurately detecting airports from high-resolution remote sensing images has important practical value in fields such as aircraft integrated navigation, military reconnaissance, and precision strikes, and has received increasing attention from researchers. However, the background of high-resolution remote sensing images is complex, with a large number of easily confused ground objects such as roads, rivers, and artificial buildings, which increases the difficulty of detecting airports in remote sensing images.

[0003] Common methods for airport detection and recognition can be divided into methods based on edge extraction and methods based on significant region extraction. Methods based on edge extraction extract the edges of the airport runway from the image according to the parallel long straight line characteristics of the airport runway, and confirm the airport area through line detection methods such as Hough transformation combined with texture recognition. Methods based on significant region extraction determine the airport candidate regions through significant region extraction methods according to the difference between the airport and the surrounding textures, and detect the airport from the candidate regions.

[0004] The above two types of methods have their own advantages and disadvantages. Methods based on edge extraction have high computational efficiency, but due to background interference, the false alarm rate is higher than that of methods based on region segmentation. Methods based on significant region extraction do not rely on the results of edge extraction and can effectively suppress the negative impact of background interference on airport detection. However, this type of method requires pixel-by-pixel analysis, so it is slow and has high complexity. Summary of the Invention

[0005] To solve the above problems, the present invention proposes a method for automatically detecting airports in remote sensing images, including the following steps:

[0006] Perform region segmentation on the original remote sensing image to obtain superpixel regions;

[0007] Perform line detection on the original remote sensing image to obtain the line extraction result;

[0008] Calculate the saliency map of the remote sensing image according to the line extraction result;

[0009] Combine the original remote sensing image and the saliency map of the remote sensing image to calculate the airport contour, and obtain the airport detection result according to the contour.

[0010] Further, combining the original remote sensing image and the saliency map of the remote sensing image to calculate the airport contour, and obtaining the airport detection result according to the contour, specifically includes the following steps:

[0011] Calculate the energy function of the saliency map and the remote sensing image. The formula is: Where, represents the average brightness of the pixels within the target contour. The target is the airport. is a coefficient, set to 1; U represents the image to be extracted. represents the average value of all gray levels of the saliency map. H represents the Heaviside function. represents the level set function. represents a real number. Calculate the extraction result map I that minimizes the energy function. The formula is: Where, and represent the average brightness of the pixels inside and outside the target contour. The target is the airport. U represents the image to be extracted. represents the average value of all gray levels of the saliency map. H represents the Heaviside function. represents the level set function. , is the set of real numbers. and represent constants. represents the Gaussian mean error. Calculate the minimum bounding rectangle of the airport contour to obtain the final airport detection result.

[0012] Furthermore, use the simple linear iterative clustering algorithm to perform region segmentation on the original remote sensing image. The specific method of simple linear iterative clustering is as follows:

[0013] (10a) Assume that the number of superpixels to be segmented is k, and the total number of pixels is N. Divide the original remote sensing image evenly into k grids, each grid containing N / k pixels, and initialize the clustering center as Where represents the center of the k-th grid. represents the Lab gray level component at the center of the k-th grid.

[0014] (10b) Calculate the gradient magnitude of pixel points within the M×M neighborhood of each grid center in turn, and determine the pixel point with the minimum gradient in the neighborhood as the new clustering center; M is an odd number greater than or equal to 3.

[0015] (10c) Calculate the similarity between each pixel point and its nearest clustering center. The formula is:

[0016]

[0017]

[0018]

[0019] S represents the spacing between the clustering centers, and m represents the set compactness factor, which is used to measure the proportion of spectral information and spatial information in the similarity measurement; and and are the Lab gray-scale components at the center of the k-th grid respectively; and and are the Lab values of the i-th pixel point respectively;

[0020] (10d) In the 2S×2S local area, if is less than the minimum value among all previously calculated similarities , then it is considered that this pixel point is in the superpixel where the clustering center is located, making , otherwise the clustering center to which this pixel point belongs remains unchanged;

[0021] (10e) Recalculate the clustering centers and recluster, repeat the iteration, calculate the distance E between the clustering centers in the previous and current times. If it is less than the set threshold , the clustering ends; otherwise, repeat (10b)-(10d) until the threshold requirement is met; the final clustering result is the superpixel region segmentation result.

[0022] Furthermore, the original remote sensing image is detected for straight lines, which specifically includes the following steps:

[0023] (20a) Downsample the original remote sensing image with a resolution greater than the threshold Q using the Gaussian scale function;

[0024] (20b) Obtain the gradient magnitude and gradient direction map of the image, and rotate the gradient direction map by 90°, obtaining the local direction map of the remote sensing image;

[0025] (20c) Based on the local direction map of the remote sensing image, pseudo-sort the pixel points of the remote sensing image according to the gradient magnitude;

[0026] (20d) After the pseudo-sorting ends, set the initial state of all pixel points to unused, and set the corresponding pixel points with gradient values less than the threshold P to used;

[0027] (20e) Select the point with the largest gradient value among the points marked as unused as the seed point, search for the points within the set range of its gradient direction around it, and simultaneously generate a rectangle containing these points;

[0028] (20f) Calculate whether the density of same-sex points in the rectangle meets the condition. If not, truncate the rectangle into multiple ones until the density of same-sex points meets the condition. After meeting the condition, output the rectangle.

[0029] (20g) Repeat (20f) until the status of all points is marked as used.

[0030] Further, in step (20c), the pseudo-sorting of the pixel points of the remote sensing image according to the gradient magnitude is specifically as follows: evenly divide the range between the minimum gradient value and the maximum gradient value into several intervals, each interval corresponding to a value, and then place each pixel point into the corresponding interval according to its gradient magnitude.

[0031] Further, in the set range described in step (20e), the difference in the gradient directions of the points where the seed points are searched is less than the threshold T.

[0032] Further, calculating the saliency map of the remote sensing image based on the straight line extraction result includes the following steps:

[0033] (30a) For each superpixel , count all the straight lines falling within the area of this superpixel ;

[0034] (30b) Calculate the influence factor ;

[0035] where p represents the number of straight lines falling within the area of this superpixel, and len represents the length of the straight line.

[0036] Further, calculating the saliency map of the remote sensing image based on the straight line extraction result further includes the following steps:

[0037] (30c) Calculate the straight line density of the superpixel , and the formula is:

[0038]

[0039] where represents the number of pixels falling within the superpixel , represents the number of pixels included in the straight line segment falling within the superpixel , represents the influence factor.

[0040] Further, calculating the saliency map of the remote sensing image based on the straight line extraction result further includes the following steps:

[0041] (30d) Calculate the saliency of the superpixel , and the formula is:

[0042]

[0043] Among them, represents the line density of superpixels , and represents the Gaussian mean error.

[0044] A remote sensing image airport automatic detection system is also provided, including a region segmentation module, a line detection module, and a contour detection module. The three modules implement the automatic detection of airports in remote sensing images based on any of the above detection methods.

[0045] A terminal is also provided, which is characterized by including a memory and a processor, and a computer program instruction that is stored on the memory and loaded and executed by the processor for any of the above methods.

[0046] A computer-readable storage medium is also provided, storing a computer program that is loaded and executed by the processor for any of the above methods.

[0047] Compared with the prior art, the present invention has the following advantages:

[0048] The present invention can provide support for fields such as aircraft integrated navigation, military reconnaissance, and precision strike; by significantly suppressing the negative impact of background interference on the edge extraction result in the significant region, it takes into account both the computational efficiency and accuracy of the detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flowchart of the method according to an embodiment of the present invention.

[0050] Figure 2 is a remote sensing image to be detected according to an embodiment of the present invention.

[0051] Figure 3 is the result of line extraction of the remote sensing image to be detected according to an embodiment of the present invention.

[0052] Figure 4 is the saliency map of the remote sensing image to be detected according to an embodiment of the present invention

[0053] Figure 5 is the airport contour generated from the remote sensing image to be detected according to an embodiment of the present invention.

[0054] Figure 6 is the airport detection result of the remote sensing image according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The purpose of the present invention is to provide a method for automatically detecting airports in remote sensing images, which suppresses the negative impact of background interference on the edge extraction result through the significant region, takes into account both the efficiency and accuracy of airport detection, and provides support for fields such as aircraft integrated navigation, military reconnaissance, and precision strike.

[0056] To facilitate the understanding of the present application, the present application will be described more comprehensively below with reference to the relevant drawings. Embodiments of the present application are shown in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present application more thorough and comprehensive.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used in the specification of this application herein are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0058] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element.

[0059] Spatial relationship terms such as "under", "below", "beneath", "underneath", "above", "over", etc. can be used herein to describe the relationship of one element or feature shown in the drawings with other elements or features. It should be understood that in addition to the orientation shown in the drawings, spatial relationship terms also include different orientations of the device in use and operation. For example, if the device in the drawing is flipped, an element or feature described as "under other elements" or "beneath it" or "under it" will be oriented "above" other elements or features. Therefore, the exemplary terms "under" and "below" can include both the upper and lower orientations. In addition, the device can also include other orientations (such as rotating 90 degrees or other orientations), and the spatial descriptors used herein are accordingly interpreted.

[0060] It should be noted that when an element is considered to be "connected" to another element, it can be directly connected to the other element or connected to the other element through an intermediate element. In addition, "connection" in the following embodiments should be understood as "electrical connection", "communication connection", etc. if there is a transmission of electrical signals or data between the connected objects.

[0061] As used herein, the singular forms "a", "an" and "the" may also include the plural forms unless the context clearly dictates otherwise. It should also be understood that the terms "comprise / include" or "have" etc. specify the presence of the stated features, wholes, steps, operations, components, parts or combinations thereof, but do not preclude the presence or addition of one or more other features, wholes, steps, operations, components, parts or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the related listed items.

[0062] To make the objectives, technical solutions and advantages of this application clearer, the following further details this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0063] In order to balance the computational efficiency and accuracy of airport detection and solve the problem of considering false alarms and abnormal boundaries when synchronously extracting the airport contour during airport positioning, this embodiment provides an automatic airport detection method for remote sensing images, which reduces the negative impact of background interference on the edge extraction result through saliency region suppression. The specific steps are as follows:

[0064] (10) Use the Simple Linear Iterative Clustering (SLIC) algorithm to perform region segmentation on the input remote sensing image to obtain superpixel regions.

[0065] (20) For the input remote sensing image, use a line segment detector (LSD) to perform line detection on the image to obtain the line extraction result.

[0066] (30) Calculate the saliency map of the remote sensing image based on the line extraction result.

[0067] (40) Combine the remote sensing image and its saliency map to calculate the airport contour, and obtain the airport detection result based on this contour.

[0068] Combined with the attached Figure 1 , the specific steps of the above method are as follows:

[0069] (10) Use the Simple Linear Iterative Clustering (SLIC) algorithm to perform region segmentation on the input original remote sensing image to obtain superpixel regions.

[0070] (10a) Assume that the number of superpixels to be segmented is k, the total number of pixels is N, evenly divide the remote sensing image into k grids, each grid contains N / k pixels, and initialize the clustering center as where represents the center of the kth grid, and represents the Lab gray component at the center of the kth grid. represents the Lab gray component at the center of the kth grid.

[0071] (10b) Calculate the gradient magnitudes of the 9 pixel points within the 3×3 neighborhood of each grid center in turn, and determine the neighborhood pixel point with the minimum gradient as the new clustering center.

[0072] (10c) Calculate the similarity between each pixel and its nearest cluster center according to the following formula .

[0073]

[0074]

[0075]

[0076] S represents the spacing between cluster centers, and m represents the set compactness factor, which is used to measure the weights of spectral information and spatial information in similarity measurement; , , are the Lab gray components at the k-th grid center respectively; , , are the Lab values of the i-th pixel respectively.

[0077] (10d) In the 2S×2S local area, if is less than the minimum value among all previously calculated similarities , then it is considered that this pixel is in the superpixel where the cluster center is located, and the label is re-assigned so that , and the re-assigned label indicates that this pixel belongs to the new corresponding superpixel, otherwise the cluster center to which this pixel belongs remains unchanged.

[0078] (10e) Recalculate the cluster centers and re-cluster, repeat the iteration, calculate the distance E between the two cluster centers before and after. If it is less than the set threshold , the clustering ends; otherwise, repeat (10b)-(10d) until the threshold requirement is met. The clustering result is the superpixel segmentation result.

[0079] (20) For the input original remote sensing image (such as the remote sensing image shown in Figure 2 ), use a line segment detector (LSD) to detect lines in the remote sensing image, and obtain the line detection result as shown in Figure 3 .

[0080] (20a) Downsample the original remote sensing image with a large resolution (resolution greater than the threshold Q) using a Gaussian scale function, and do not perform downsampling on the original remote sensing image with a resolution less than or equal to the threshold Q.

[0081] (20b) Obtain the gradient magnitude and gradient direction map of the remote sensing image, and rotate the gradient direction map by 90° to obtain the local direction map of the remote sensing image.

[0082] (20c) Pseudo-sort the pixel points of the remote sensing image according to the gradient magnitude. That is, evenly divide the range between the minimum gradient value and the maximum gradient value into several intervals, each interval corresponding to a value, and then place each point into the corresponding interval according to its gradient magnitude.

[0083] (20d) After the sorting is completed, set the initial state of all points to unused, and then set the points corresponding to the smaller gradient values (less than the threshold P) to used.

[0084] (20e) Select the point with the largest gradient value from the points marked as unused as the seed point, search for points within a certain range of its gradient direction (the difference between the two gradient directions is less than the threshold T) around it, and generate a rectangle containing these points at the same time.

[0085] (20f) Then calculate whether the density of the same-sex points in the rectangle meets the condition. If not, truncate the rectangle into multiple ones until the density of the same-sex points meets the condition, and output the rectangle after meeting the condition.

[0086] (20g) Repeat (20f) until the status of all points is used.

[0087] (30) Calculate the saliency map of the remote sensing image according to the straight line extraction result, as Figure 4 shown.

[0088] (30a) For each superpixel , count all the straight lines falling within the area of this superpixel .

[0089] (30b) Calculate the influence factor according to the following formula

[0090] where, p represents the number of straight lines falling within the area of this superpixel, and len represents the length of the straight line

[0091] (30c) Calculate the straight line density of the superpixel according to the following formula

[0092]

[0093] where, represents the number of pixels falling within the superpixel , represents the number of pixels included in the straight line segment falling within the superpixel , represents the influence factor.

[0094] (30d) Calculate the saliency of the superpixel according to the following formula

[0095]

[0096] Among them, represents the line density of the superpixel, represents the Gaussian mean error

[0097] (40) Combine the original remote sensing image and its saliency map, and calculate the field contour, as Figure 5 shown.

[0098] (40a) Calculate the energy function of the saliency map and the original remote sensing image according to the following formula

[0099]

[0100] Among them, represents the average brightness of the pixels within the airport contour, is a coefficient, usually set to 1; U represents the image to be extracted, represents the average value of all gray levels of the saliency map, H represents the Heaviside function, represents the level set function, , is the set of real numbers.

[0101] (40b) Calculate the extraction result map I that minimizes the energy function according to the following formula

[0102]

[0103] Among them, and represent the average brightness of the pixels inside and outside the target contour, U represents the image to be extracted, represents the average value of all gray levels of the saliency map, H represents the Heaviside function, represents the level set function, , is the set of real numbers, and represent constants, represents the Gaussian mean error.

[0104] (40c) Calculate the minimum bounding rectangle of the computer field contour, which is the final airport detection result, as Figure 6 shown. The thresholds in this embodiment all need to be determined according to experience.

[0105] This embodiment also provides a remote sensing image airport automatic detection system, which includes a region segmentation module, a line detection module, and a contour detection module. The three modules implement the automatic detection of airports in remote sensing images according to the above detection method.

[0106] The present invention can provide support for fields such as aircraft integrated navigation, military reconnaissance, and precision strike; by significantly suppressing the negative impact of background interference on the edge extraction result in a significant area, it takes into account both the computational efficiency and accuracy of the detection method.

[0107] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A remote sensing image airport automatic detection method, characterized in that: The following steps are involved: Perform regional segmentation on the original remote sensing image to obtain super-pixel regions; Perform straight line detection on the original remote sensing image to obtain straight line extraction results; According to the straight line extraction results, the saliency map of the remote sensing image is calculated; The original remote sensing image and the saliency map of the remote sensing image are combined to calculate the airport outline, and the airport detection result is obtained based on the outline.

2. The automatic airport detection method of remote sensing images according to claim 1 is characterized in that: Combining the original remote sensing image and the saliency map of the remote sensing image, the airport outline is calculated, and the airport detection result is obtained according to the outline, which specifically includes the following steps: The energy function of the saliency map and remote sensing image is calculated as follows: in, represents the average brightness of pixels within the target outline, where the target is an airport. is a coefficient, set to 1; U represents the image to be extracted, represents the average value of all gray levels of the saliency map, H represents the Heaviside function, represents the level set function, represents a real number; Calculate the extraction result image I that minimizes the energy function. The formula is: in, and represents the average brightness of pixels inside and outside the target contour, the target is an airport, U represents the image to be extracted, represents the average value of all gray levels of the saliency map, H represents the Heaviside function, represents the level set function, , is the set of real numbers, and represents a constant, represents the Gaussian mean error; The minimum enclosing rectangle of the airport outline is calculated to obtain the final airport detection result.

3. The automatic airport detection method of remote sensing images according to claim 1 is characterized in that: The original remote sensing image is segmented using a simple linear iterative clustering algorithm. The specific method of simple linear iterative clustering is as follows: (10a) Assume that the superpixel to be segmented is The number of pixels is k, the total number of pixels is N, the original remote sensing image is evenly divided into k grids, each grid contains N / k pixels, and the initial cluster center is in represents the center of the kth grid, Represents the Lab grayscale component at the center of the kth grid; (10b) Calculate the M×M neighborhood of each grid center in turn The gradient size of each pixel is used to determine the neighboring pixel with the smallest gradient as the new cluster center; M is an odd number greater than or equal to 3; (10c) Calculate the similarity between each pixel and its nearest cluster center , the formula is: Among them, S represents the spacing between cluster centers, and m represents the set compact factor, which is used to measure the proportion of spectral information and spatial information in similarity measurement; , , are the Lab grayscale components at the center of the kth grid; , , are the Lab values ​​of the i-th pixel respectively; (10d) In the 2S×2S local region, if Smaller than the minimum of all previously calculated similarities , then the pixel is considered to be in the superpixel where the cluster center is located, so that , otherwise the cluster center to which the pixel belongs remains unchanged; (10e) Recalculate the cluster center and re-cluster, repeat the iteration, calculate the distance E between the two cluster centers, and if it is less than the set threshold , clustering ends; otherwise, repeat (10b)-(10d) until the threshold requirement is met; the final clustering result is the superpixel region segmentation result.

4. The automatic airport detection method of remote sensing images according to claim 1 is characterized in that: The straight line detection is performed on the original remote sensing image, which specifically includes the following steps: (20a) Downsampling the original remote sensing image with a resolution greater than a threshold Q using a Gaussian scaling function; (20b) Obtain the gradient magnitude and gradient direction map of the image, and rotate the gradient direction map by 90° to obtain the local direction map of the remote sensing image; (20c) Based on the local orientation map of the remote sensing image, the pixels of the remote sensing image are pseudo-sorted according to the gradient size; (20d) After the pseudo sorting is completed, the initial status of all pixels is set to unused, and the corresponding pixels whose gradient values ​​are less than the threshold P are set to used; (20e) Select the point with the largest gradient value from the points marked as unused as the seed point, search for points around it whose gradient direction is within the set range, and generate a rectangle containing these points; (20f) Calculate whether the density of the same-sex points in the rectangle meets the condition. If not, cut the rectangle into multiple pieces until the density of the same-sex points meets the condition. If the condition is met, output the rectangle. (20g) Repeat (20f) until all points are in the used state.

5. The automatic airport detection method of remote sensing images according to claim 4 is characterized in that: In step (20c), the pixel points of the remote sensing image are pseudo-sorted according to the gradient size as follows: the minimum gradient value and the maximum gradient value are evenly divided into several intervals, each interval corresponds to a value, and then each pixel point is placed in the corresponding interval according to its gradient size.

6. The automatic airport detection method of remote sensing images according to claim 4 is characterized in that: In step (20e), the gradient direction difference of the points searched for as seed points within the set range is less than the threshold value T.

7. The automatic airport detection method of remote sensing images according to claim 1 is characterized in that: According to the straight line extraction results, calculating the saliency map of the remote sensing image includes the following steps: (30a) For each superpixel , count all the straight lines falling into the superpixel area ; (30b) Calculation of impact factor ; Among them, p represents the number of lines falling into the superpixel area, and len represents the length of the line.

8. The automatic airport detection method of remote sensing images according to claim 7 is characterized in that: According to the straight line extraction results, calculating the saliency map of the remote sensing image also includes the following steps: (30c) Calculate superpixels The linear density is: in, Indicates that it falls into a superpixel The number of pixels in Indicates that it falls into a superpixel The number of pixels contained in the straight line segment in Represents the impact factor.

9. The automatic airport detection method of remote sensing images according to claim 8 is characterized in that: According to the straight line extraction results, calculating the saliency map of the remote sensing image also includes the following steps: (30d) Calculate superpixels The significance of is: in, Represents superpixel The linear density, represents the Gaussian mean error.

10. A remote sensing image airport automatic detection system, characterized in that: The invention comprises a region segmentation module, a straight line detection module and a contour detection module. The three modules realize automatic detection of airports in remote sensing images based on any detection method of claims 1-9.

11. A terminal, characterized in that: The invention comprises a memory and a processor, wherein the memory stores computer program instructions which are loaded by the processor and execute any detection method according to claims 1-9.

12. A computer-readable storage medium, characterized in that: A computer program is stored which is loaded by a processor and executes any one of the detection methods described in claims 1-9.

Images