Unmanned aerial vehicle runway end point detection method based on image texture

Abstract

The invention discloses an unmanned aerial vehicle runway end point detection method and device based on image texture, a medium and equipment, and the method comprises the steps: carrying out the amplitude suppression of pixel points of which the texture is in the vertical direction in a runway image based on a cosine function, and carrying out the amplitude enhancement of pixel points of which the texture direction is consistent based on an amplitude enhancement function; obtaining an amplitude correction first pixel point set; performing linear detection on the amplitude correction pixel point set by utilizing a Hough transform algorithm to obtain each detection line segment with a preset deviation angle, and correcting texture amplitudes of pixel points on each detection line segment through preset parameters to obtain an amplitude correction second pixel point set; and correcting the second pixel point set based on the amplitude to determine a voting weight matrix, calculating a voting graph of all voting points on the detection line segment by adopting a preset soft voting algorithm, and taking the maximum voting value in the voting graph as the end point of the unmanned aerial vehicle runway.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle navigation, and particularly relates to an unmanned aerial vehicle runway endpoint detection method and device based on image texture, a medium and equipment. BACKGROUND

[0002] In the prior art, the background area and the road surface area of a road image are estimated by using binocular vision technology, and then the texture response of each pixel point of the road image is estimated by using a Gabor filter set, amplitude correction is performed, a plurality of voting point selection strategies are set according to the background area to remove the influence of the background area, the search range of vanishing points is reduced by dynamically adjusting the candidate point range scheme, and the candidate point with the highest number of votes in the voting space is regarded as a vanishing point by combining an angle-priority voting strategy. Although the detection process of this method can improve the accuracy of vanishing point detection, the input image is required to be a binocular image, and common images are mostly images captured by a monocular camera, which are not suitable for this detection method. SUMMARY

[0003] The main purpose of the present application is to provide an unmanned aerial vehicle runway endpoint detection method and device based on image texture, which aims to solve the technical problem of the unmanned aerial vehicle runway endpoint detection method based on image texture.

[0004] To achieve the above-mentioned purpose, the present application provides an unmanned aerial vehicle runway endpoint detection method based on image texture, which is used for autonomous detection of the runway endpoint of a runway image collected by a monocular camera of an unmanned aerial vehicle under satellite denial conditions, and includes the following steps: performing amplitude suppression on pixel points with a vertical texture direction in the runway image based on a cosine function, performing amplitude enhancement on pixel points with a consistent texture direction based on an amplitude enhancement function, and obtaining a first set of amplitude-corrected pixel points; performing straight line detection on the first set of amplitude-corrected pixel points by using a Hough transform algorithm, obtaining each detected line segment with a preset offset angle, and correcting the texture amplitude of pixel points on each detected line segment by using a preset parameter to obtain a second set of amplitude-corrected pixel points; determining a voting weight matrix based on the second set of amplitude-corrected pixel points, calculating the voting map of all voting points on the detected line segment by using a preset soft voting algorithm, and taking the maximum voting value in the voting map as the endpoint of the unmanned aerial vehicle runway.

[0005] Optionally, the amplitude value of each pixel point in the amplitude value modified third pixel point set is corrected based on a cosine value of a texture direction of the pixel point, to obtain an amplitude value modified first pixel point set. Among the three pixel points below, find a pixel point with the smallest absolute value of an included angle difference with the pixel point . The amplitude value modified third pixel point set with an absolute value of the included angle difference less than a preset threshold is weighted to obtain the amplitude value modified first pixel point set.

[0006] Optionally, the process of extracting the texture graph of the runway image comprises: converting the runway image into a gray-scale image; and extracting the texture graph in the gray-scale image by using a Gabor filter, wherein the texture graph comprises amplitude values and texture directions of each pixel point.

[0007] Optionally, before the process of detecting the straight line by using the Hough transform algorithm, the method further comprises: converting the runway image into an HSV graph.

[0008] Optionally, the process of calculating the voting graph of all the voting points on the detected line segment by using the preset soft voting algorithm comprises: creating a voting score matrix with the same size as the runway image; and traversing all the voting points on the detected line segment, to construct a pentagonal voting area with each voting point as a center, wherein the pentagonal voting area comprises an isosceles triangle area and a rectangular area with the same side as the base of the isosceles triangle area, so that a ratio of a distance between a vertex P of the triangle and a midpoint K of a bottom I1J1 of the triangle to a height of the runway image is a first ratio, and a ratio of a distance between the vertex P of the isosceles triangle and a midpoint Q of a side I2J2 of the rectangular area away from the base of the triangle to the height of the runway image is a second ratio, and the point K belongs to the detected line segment PQ; taking each voting point in the pentagonal voting area as a candidate point, calculating a vote value of each voting point based on a preset soft voting formula and a voting weight matrix, and determining the voting graph based on the vote value.

[0009] Optionally, the preset soft voting formula has an expression as follows:

[0010]

[0011] wherein, V represents a voting score matrix, represents the number of votes of a candidate point , represents the difference between the abscissa of each voting point P on the detected line segment PQ and the abscissa of the candidate point , represents the difference between the abscissa of each voting point P on the detected line segment PQ and the abscissa of the candidate point , represents a preset proportion coefficient, represents the ratio of the distance between the voting point P and the candidate point L to the diagonal length of the runway image, I represents the voting weight of the voting point , and exp represents the exponential with base e.

[0012] In addition, to achieve the above object, the application also provides an unmanned aerial vehicle runway endpoint detection device based on image texture, which is used for autonomous detection of the runway endpoint of the runway image collected by the unmanned aerial vehicle monocular camera under satellite denial conditions, and comprises: a first amplitude correction module, which is used for amplitude suppression of pixel points with vertical texture in the runway image based on a cosine function, and amplitude enhancement of pixel points with consistent texture direction based on an amplitude enhancement function, to obtain a first set of amplitude-corrected pixel points; a second amplitude correction module, which is used for straight line detection of the first set of amplitude-corrected pixel points by using a Hough transform algorithm, to obtain each detected line segment with a preset offset angle, and correct the texture amplitude of pixel points on each detected line segment by using a preset parameter, to obtain a second set of amplitude-corrected pixel points; a runway endpoint calculation module, which is used for determining a voting weight matrix based on the second set of amplitude-corrected pixel points, and calculating the voting map of all voting points on the detected line segment by using a preset soft voting algorithm, and taking the maximum voting value in the voting map as the endpoint of the unmanned aerial vehicle runway.

[0013] To achieve the above object, the application further provides a computer readable storage medium, which comprises instructions, when the instructions are run on a computer, the computer executes the image texture based unmanned aerial vehicle runway endpoint detection method provided by the above embodiments.

[0014] To achieve the above object, the application further provides an electronic device, which comprises at least one processor, a memory and an input output unit; wherein the memory is used for storing a computer program, and the processor is used for calling the computer program stored in the memory to execute the image texture based unmanned aerial vehicle runway endpoint detection method provided by any of the preceding embodiments.

[0015] The embodiment of the present application provides a runway endpoint detection method, device, medium and equipment based on image texture. The vertical direction pixel points in a runway image are subjected to amplitude suppression based on a cosine function, and the pixel points with consistent texture directions are subjected to amplitude enhancement based on an amplitude enhancement function, so as to obtain an amplitude corrected first pixel point set. The Hough transform algorithm is used for straight line detection on the amplitude corrected pixel point set, so as to obtain each detected line segment with a preset offset angle, and the texture amplitude of the pixel points on each detected line segment is corrected through a preset parameter, so as to obtain an amplitude corrected second pixel point set. The voting weight matrix is determined based on the amplitude corrected second pixel point set, a preset soft voting algorithm is used for voting of all voting points on the detected line segment, the maximum voting value in the voting map is taken as the endpoint of the unmanned aerial vehicle runway, and the present application can more accurately detect the runway endpoint and has good robustness. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A runway endpoint detection algorithm flowchart is provided for the embodiment of the present application. Figure 2 A pixel point neighborhood graph is provided for the embodiment of the present application. Figure 3 A line in the Cartesian coordinate to a point in the Hough space is provided for the embodiment of the present application. Figure 4 A point in the Hough space to a line in the Cartesian coordinate is provided for the embodiment of the present application. Figure 5 A Hough transform detection runway straight line effect diagram is provided for the embodiment of the present application. Figure 6 A voting area schematic diagram is provided for the embodiment of the present application. Figure 7 An algorithm experiment result diagram is provided for the embodiment of the present application.

[0017] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0019] In a real modern combat environment, the satellite positioning function of the unmanned aerial vehicle is easily disturbed, which directly affects the safe recovery and landing process of the unmanned aerial vehicle on the airport runway. In order to solve this problem, the application provides an unmanned aerial vehicle runway endpoint detection method based on images, which can help the unmanned aerial vehicle to detect the runway endpoint based on the images taken by the camera under the condition of satellite rejection, and then provide the direction of the runway recovery for the unmanned aerial vehicle, and assist in guiding the unmanned aerial vehicle to realize safe landing.

[0020] In the unmanned aerial vehicle runway image, there are often lookout towers, houses, intersections and other objects, the texture distribution of these objects on the image is uniform and the amplitude is large, which will interfere with the detection result.

[0021] The unmanned aerial vehicle runway endpoint detection method provided by the application comprises two steps, the first step provides a weight updating method based on texture angle correlation and straight line information, and the second step uses a local texture soft voting method to detect the runway endpoint, and the algorithm flow is as shown in the accompanying Figure 1 The two steps are described as follows: With reference to Figure 1 The unmanned aerial vehicle runway endpoint detection method based on image texture provided by the first embodiment of the application can comprise: S10, the amplitude of the pixel point in the texture of the runway image is suppressed based on the cosine function, and the amplitude of the pixel point with consistent texture direction is enhanced based on the amplitude enhancement function, to obtain an amplitude modified first pixel point set; In an embodiment of the application, the specific execution process of step S10 can comprise the following: S101, extract the texture map of the runway image; In an embodiment of the application, the specific execution process of step S101 can comprise the following: S1011, convert the runway image into a gray scale image; S1012, use a Gabor filter to extract the texture map in the gray scale image, wherein the texture map contains the amplitude and texture direction of each pixel point.

[0022] S102, set an access marker matrix Flag, wherein the length and width of the access marker matrix Flag are the same as the size of the runway image, and the access marker matrix Flag is initialized to 0; S103, start from the lower left corner pixel point of the texture map, traverse each pixel point in the order from left to right and from bottom to top, set each pixel point in the access marker matrix Flag to 1, and correct the amplitude of each pixel point based on the cosine value of the texture direction of each pixel point, to obtain an amplitude corrected third pixel point set; Specifically, in order to eliminate the interference of vertical objects such as airport observation towers and buildings in the image on the detection results, a cosine function is used. Texture amplitude suppression is performed on pixels with a vertical direction to obtain the first pixel set.

[0023] S104, Pixels in the third pixel set of amplitude correction Among the three pixels below, find the pixel that matches the given pixel. The pixel with the smallest absolute value of the angle difference The third set of amplitude correction pixels with absolute values ​​of angle differences less than a preset threshold are weighted to obtain the first set of amplitude correction pixels; the voting weight matrix is ​​determined based on the second set of amplitude correction pixels.

[0024] Specifically, for a complete object in an image, the internal texture direction has a consistent inheritance. The texture amplitude is increased using formula (1), which enhances the understanding of the image at the semantic level.

[0025] (1) Among them, point For the pixels whose voting weights need to be updated, Its texture direction, It is considered to be the point located below it that is closest to its direction and whose angle difference is within a certain range. and Having the same object properties, and lie in The upper part, compared to It has a greater chance of becoming the finish line of the runway. Among the parameters... Set to 0.05.

[0026] For example, the specific steps for determining the voting weight matrix are as follows: Step 1: For pixels with a width of... COL The height is ROW The image, from the bottom left pixel position Starting from the left, traverse the pixels from left to right and from bottom to top. First, set an access marker matrix Flag, with the same width and height as the image size, and initialize it to 0; Step 2: Record the pixels that have been traversed as Let the access marker of that point Utilization points The texture direction value is used to correct its weight, resulting in the amplitude-corrected third pixel set. Proceed to Step 3; Step 3: As shown in the appendix Figure 2 As shown, the pixels in the texture map that are located at the voting weights to be updated The three pixel points below In the middle, find the point with the minimum absolute value of the angle difference, denoted as , the texture direction of which is , the absolute value of the angle difference between the texture directions of the two is , if , Step 5 is executed, otherwise Step 4 is executed, and the angle threshold value is set to 10° in the present application. Step 4: The weight of is subjected to the following weighting operation , to obtain the amplitude-corrected first pixel point set, and Step 5 is then executed. Step 5: The update of the point is ended, and the next pixel point in Step 1 is continued to be accessed to complete the update of the next point.

[0027] S20, using the Hough transform algorithm to detect straight lines from the amplitude-corrected pixel point set, to obtain each detected line segment with a preset offset angle, and to correct the texture amplitude of the pixel points on each detected line segment through a preset parameter, to obtain an amplitude-corrected second pixel point set; and determining a voting weight matrix based on the amplitude-corrected second pixel point set.

[0028] In an embodiment of the present application, before Step S20, the specific execution process of the image texture-based unmanned aerial vehicle runway endpoint detection method can include converting the runway image into an HSV image.

[0029] In an embodiment of the present application, the specific execution process of Step S20 can include the following: First, the principle of Hough transform is briefly introduced as follows: In Hough transform, for a determined straight line in the polar coordinate system, the perpendicular distance from the origin to the straight line is , and the included angle between the perpendicular and the axis is . This straight line can be represented by . (2) Taking as the coordinate axis, the Hough space can be obtained, and the straight line is a determined point in the Hough space; if there are a plurality of points in the polar coordinate system, each point can draw an infinite number of straight lines, and each straight line corresponds to a group of same , which is a curve in the Hough space, as shown in the accompanying drawings. Figure 3

[0030] ​​​To find a straight line in polar coordinates that contains as many points as possible, one can find the points where curves intersect most frequently in Hough space, thus determining the straight line, as shown in the appendix. Figure 4 As shown.

[0031] Since runway images typically contain clear boundary lines, center lines, and other straight lines, and these lines extend in the same direction as the runway endpoint, this application uses Hough transform to detect straight lines in the image and weights the voting points on these lines to increase the influence of these voting points.

[0032] The specific execution process of step S20 can be as follows: Step 1: Adjust pixel width ,high The image is processed using the Hough transform to detect straight lines. Pixels above the line are marked as 1, and the rest are marked as 0, forming a matrix L of the same size as the input image. Step 2: Construct a marker matrix with the same size as the image. The initial value is 0, and it traverses the pixels from left to right and from bottom to top, starting from the bottom left pixel position. Step 3: For each traversed pixel, denote it as , and its texture amplitude is . Find the value corresponding to that point in matrix L. If the value is 1, use the formula... The weights are adjusted to obtain the amplitude-corrected second pixel set; the voting weight matrix is ​​determined based on the amplitude-corrected second pixel set. Wherein, The value is 1.2. Continue this step until all pixels have been traversed.

[0033] Understandably, this application uses the aforementioned Hough transform line detection algorithm for updating voting point weights. First, the input image size is scaled to 256×256 pixels to improve computation speed. Second, the Hough transform is used to detect lines in the image, and the weights of the lines are updated accordingly. The line segments between (-70, 70) are preserved, and a maximum of 6 lines are retained per image.

[0034] The line detection results using the Hough transform line detection algorithm in this application are shown in the attached figure. Figure 5 As shown, from left to right, the images are the original image, the binary image of the detection result, and the detection effect image. It can be seen that the edges and centerline of the drone runway can be effectively detected.

[0035] S30. Use a preset soft voting algorithm to calculate the voting map of all voting points on the detected line segment, and take the maximum voting value in the voting map as the end point of the UAV runway.

[0036] In an embodiment of the present application, the specific implementation process of step S30 can include the following: A voting score matrix with the same size as the runway image is created. All the voting points on the detected line segment are traversed to construct a pentagonal voting region with each voting point as the origin, wherein the pentagonal region includes an isosceles triangle region and a rectangular region with the same side as the base of the isosceles triangle region, so that the ratio of the distance from the vertex P of the triangle to the midpoint K of the bottom I1J1 of the triangle to the height of the runway image is a first ratio, and the ratio of the distance from the vertex P of the isosceles triangle to the midpoint Q of the side I2J2 of the matrix region I1I2J2J1 away from the base of the triangle to the height of the runway image is a second ratio, and the point K belongs to the detected line segment PQ. Each voting point in the pentagonal voting region is taken as a candidate point, the vote value of each voting point is calculated based on a preset soft voting formula and a voting weight matrix, and the voting map is determined based on the vote value.

[0037] Exemplarily, a method based on local texture soft voting is given below to determine the voting map and the runway terminal point.

[0038] After the first point set with amplitude correction obtained by updating the texture weight based on the angle correlation and the second point set with amplitude correction obtained by updating the texture weight based on the straight line information, a texture weight matrix M with a dimension of 256x256 is obtained. The present application uses an improved local region soft voting mechanism to calculate the coordinates of the runway terminal point in the image using the texture weight matrix M . The specific steps are as follows: Step 1: Create a voting score matrix V with a size of 256x256.

[0039] Step 2: Traverse all the points in the matrix, and the points on the line segment are called voting points P , and create a region with the voting point P as the vertex. Specifically, as shown in the accompanying drawings Figure 6 , wherein , , wherein H is the height of the image, which is taken as 256 in the present application.

[0040] Step 3: All the voting points can be considered as candidate points, and all the candidate points in the voting region are traversed. For a candidate point , the expression of the preset soft voting formula is: (3) (4) wherein V represents the voting score matrix, represents the candidate point a vote value of the pixel point P, a difference between a horizontal coordinate of each voting point P on the detected line segment PQ and a horizontal coordinate of the candidate point P, a difference between a horizontal coordinate of each voting point P on the detected line segment PQ and a horizontal coordinate of the candidate point P, a difference between a horizontal coordinate of each voting point P on the detected line segment PQ and a horizontal coordinate of the candidate point P, a preset proportion coefficient, a distance between the voting point P and the candidate point P, a ratio of a diagonal length of the runway image, L a distance between the voting point P and the candidate point P, I a voting weight of the voting point P, and exp represents an exponential with base e.

[0041] Step 4: Obtain the voting map Vote, and select the point with the maximum vote value as the final runway terminal point. VP

[0042] Thus, the processor can output the detected unmanned aerial vehicle runway terminal coordinate value, and the terminal coordinate can assist in guiding the unmanned aerial vehicle to safely recover and land under the denial condition.

[0043] The second embodiment of the present application provides an unmanned aerial vehicle runway terminal detection device based on image texture, which is used for autonomous detection of a runway terminal of a runway image collected by a monocular camera of an unmanned aerial vehicle under a satellite denial condition, and includes a first amplitude correction module, a second amplitude correction module, and a runway terminal calculation module. The first amplitude correction module is used for amplitude suppression on pixel points with a vertical texture direction in the runway image based on a cosine function, and amplitude enhancement on pixel points with a consistent texture direction based on an amplitude enhancement function, to obtain a first set of amplitude-corrected pixel points. The second amplitude correction module is used for straight line detection on the first set of amplitude-corrected pixel points by using a Hough transform algorithm, to obtain each detected line segment with a preset offset angle, and to correct the texture amplitude of the pixel points on each detected line segment by using a preset parameter, to obtain a second set of amplitude-corrected pixel points. The runway terminal calculation module is used for determining a voting weight matrix based on the second set of amplitude-corrected pixel points, and calculating a voting map of all voting points on the detected line segment by using a preset soft voting algorithm, and taking the maximum vote value in the voting map as the terminal of the unmanned aerial vehicle runway.

[0044] The third embodiment of the present application provides a computer readable storage medium, which includes instructions, when the instructions are run on a computer, cause the computer to execute the unmanned aerial vehicle runway terminal detection method based on image texture provided by the above-mentioned embodiments.

[0045] The fourth embodiment of the present application provides an electronic device, which includes at least one processor, a memory, and an input and output unit. The memory is used for storing a computer program, and the processor is used for calling the computer program stored in the memory to execute the unmanned aerial vehicle runway terminal detection method based on image texture provided by any of the preceding embodiments. ​​

[0046] The above merely provides the preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structure or equivalent flowchart transformation based on the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An image texture based unmanned aerial vehicle runway end detection method, characterized in that, The application relates to runway endpoint autonomous detection of runway images collected by a UAV monocular camera under satellite denial conditions, and comprises the following steps: Based on a cosine function, the amplitude of pixel points with vertical texture in the runway image is suppressed, and based on an amplitude enhancement function, the amplitude of pixel points with consistent texture direction is enhanced to obtain an amplitude-modified first pixel point set; A Hough transform algorithm is used to detect straight lines of the amplitude-modified pixel point set to obtain each detected line segment with a preset offset angle, and the texture amplitude of pixel points on each detected line segment is corrected through a preset parameter to obtain an amplitude-modified second pixel point set; and a voting weight matrix is determined based on the amplitude-modified second pixel point set; A preset soft voting algorithm is used to calculate the voting graph of all voting points on the detected line segment, and the maximum voting value in the voting graph is taken as the endpoint of the UAV runway; The preset soft voting algorithm includes the following steps: A voting score matrix with the same size as the runway image is created; All voting points on the detected line segment are traversed, and a pentagonal voting area is constructed with each voting point as the origin, wherein the pentagonal area comprises an isosceles triangle area and a rectangular area with the same side as the base of the isosceles triangle area, so that the ratio of the distance from the top vertex P of the triangle to the midpoint K of the bottom I1J1 of the triangle to the height of the runway image is a first ratio, and the ratio of the distance from the top vertex P of the isosceles triangle to the midpoint Q of the side I2J2 of the matrix area I1I2J2J1 away from the base of the triangle to the height of the runway image is a second ratio, and the point K belongs to the detected line segment PQ; Each voting point in the pentagonal voting area is taken as a candidate point, the number of votes of each voting point is calculated based on a preset soft voting formula and a voting weight matrix, and the voting graph is determined based on the number of votes. 2.The image texture based UAV runway end detection method of claim 1, wherein, The amplitude of the pixel points with consistent texture direction is enhanced based on the amplitude enhancement function to obtain the amplitude-modified first pixel point set, which comprises the following steps: A texture map of the runway image is extracted; An access marker matrix Flag is set, wherein the length and width of the access marker matrix Flag are the same as the size of the runway image, and the access marker matrix Flag is initialized as 0; Starting from the lower left corner pixel point of the texture map, each pixel point is traversed in the order of from left to right and from bottom to top, each pixel point in the access marker matrix Flag is set as 1, and the amplitude of each pixel point is corrected based on the cosine value of the texture direction of each pixel point to obtain an amplitude-modified third pixel point set; Amplitude-correcting the pixel points in the third pixel point set Among the three pixel points below, find the pixel point with the minimum absolute value of the included angle difference with the pixel point Weighting the third pixel point set with the absolute value of the included angle difference less than the preset threshold, to obtain the first amplitude-corrected pixel point set.​ 3.The image texture based UAV runway end detection method of claim 1, wherein, The process of extracting the texture map of the runway image comprises the following steps: The runway image is converted into a grayscale image; A Gabor filter is used to extract the texture map in the grayscale image, wherein the texture map contains the amplitude and texture direction of each pixel point. 4.The image texture based UAV runway end detection method of claim 1, wherein, Before the Hough transform algorithm is used to detect straight lines of the amplitude-modified pixel point set, the method further comprises the following steps: The runway image is converted into an HSV image.

5. The image texture based UAV runway end detection method of claim 1, wherein, The expression of the preset soft voting formula is as follows: wherein, V represents a voting score matrix, represents a vote value of a candidate point , represents a difference between a horizontal coordinate of each voting point P which has the same longitudinal coordinate point as the candidate point on the detection line segment PQ and a horizontal coordinate of the candidate point , represents a preset proportional coefficient, represents a ratio of a distance between the voting point P and the candidate point to a diagonal length of the racetrack image L , I represents a voting weight of the voting point , and exp represents an exponential with base e.

6. An image texture based unmanned aerial vehicle runway end detection apparatus, comprising: The application relates to runway endpoint autonomous detection of runway images collected by a UAV monocular camera under satellite denial conditions, and comprises the following steps: The first amplitude correction module is configured to perform amplitude suppression on pixel points with a vertical texture direction in the runway image based on a cosine function, and perform amplitude enhancement on pixel points with a consistent texture direction based on an amplitude enhancement function, to obtain a first set of amplitude-corrected pixel points; The second amplitude correction module is configured to perform straight line detection on the first set of amplitude-corrected pixel points by using a Hough transform algorithm, to obtain each detected line segment with a preset offset angle, and correct the texture amplitude of pixel points on each detected line segment by using a preset parameter, to obtain a second set of amplitude-corrected pixel points; and the second set of amplitude-corrected pixel points determine a voting weight matrix. The runway end calculation module is configured to calculate a voting map of all voting points on the detected line segment by using a preset soft voting algorithm, and take a maximum voting value in the voting map as the end of the runway for the unmanned aerial vehicle.

7. A computer readable storage medium characterized in that, The computer program product comprises instructions which, when executed on a computer, cause the computer to perform the image texture-based unmanned aerial vehicle runway end detection method of any one of claims 1-5.

8. An electronic device, comprising: The electronic device comprises: at least one processor, a memory, and an input-output unit; The memory is configured to store a computer program, and the processor is configured to invoke the computer program stored in the memory to execute the image texture-based unmanned aerial vehicle runway end detection method of any one of claims 1-5.

Images