A terrain comparison navigation system for unmanned aerial vehicles

Abstract

The present invention belongs to the field of unmanned aerial vehicle navigation, and discloses a terrain comparison navigation system for unmanned aerial vehicles, including a shooting device, a terrain comparison device and a navigation device; the shooting device is used to obtain terrain images during the flight of the unmanned aerial vehicle; the terrain comparison device is used to obtain an ordered set of zoom ratios, and the terrain image is matched with a terrain contour database based on the ordered set of zoom ratios to determine the position of the unmanned aerial vehicle; the navigation device is used to navigate the unmanned aerial vehicle according to the position of the unmanned aerial vehicle. The process of obtaining the ordered set of zoom ratios is to first calculate the ranking value of the zoom ratio, and then sort the zoom ratios in descending order based on the ranking value, so as to obtain the ordered set of zoom ratios. The present invention effectively increases the probability of occurrence of an event that only requires the zoom images corresponding to the front part of the zoom ratios in the ordered set to be identified in order to determine the position of the unmanned aerial vehicle.

Description

Technical Field

[0001] The invention relates to the field of unmanned aerial vehicle navigation, and in particular to a terrain comparison navigation system for unmanned aerial vehicles. Background Art

[0002] Terrain matching navigation is a technology that uses terrain contour information for navigation. Terrain matching navigation relies on detailed terrain databases that contain ground height, terrain contours and other geographic information. In terrain matching navigation, radar or other sensors are usually used to measure the terrain contour on the actual flight path of the aircraft. The actual terrain data collected by the sensor is then compared with the terrain data pre-stored in the database. Based on the comparison results, the deviation between the current position of the aircraft and the expected path is calculated, and corresponding navigation corrections are made. Terrain matching navigation allows aircraft to navigate autonomously without external navigation signals.

[0003] For routes between two locations with fixed flight paths, in addition to using GPS for navigation, a database of terrain contours beneath the route can also be generated, allowing the drone to have a backup navigation method.

[0004] In terrain matching navigation, altitude data is a very important data. Through altitude comparison, the drone can determine whether it deviates from the predetermined flight altitude, so that it can make timely adjustments when the altitude is too high or too low to ensure flight safety. However, drones usually determine their own altitude based on GPS or barometer. When the barometer and GPS are unable to work, the drone can usually only estimate the navigation altitude by comparing the terrain contour in the captured terrain image with the terrain contour in the terrain database based on the scaling ratio of the terrain image. When the altitude of the drone cannot be accurately determined, in the process of comparing the contour of the terrain image with the contour of the terrain database, it is usually necessary to scale the terrain image at multiple ratios to obtain a terrain contour with a high enough degree of matching with the terrain contour in the terrain database from the different images obtained by scaling. However, the prior art usually scales directly in the order of scaling from small to large or from large to small, which results in that when the scaling range is set to be relatively large, it may be necessary to compare more images with different scaling ratios to obtain the terrain matching result. Obviously, this will take a long time to obtain the positioning result, so that the drone can only fly at a lower flight speed. Summary of the invention

[0005] The purpose of the present invention is to disclose a terrain matching navigation system for unmanned aerial vehicles, so as to solve the problem of how to increase the probability of the event of successfully estimating the flight altitude and speed based on a small number of images with different zoom ratios when the unmanned aerial vehicle uses terrain matching navigation, so as to obtain the positioning result in a shorter time and improve the flight speed of the unmanned aerial vehicle.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The present invention provides a terrain comparison and navigation system for a UAV, comprising a shooting device, a terrain comparison device and a navigation device;

[0008] The photographing device is used to obtain terrain images during the flight of the UAV;

[0009] The terrain comparison device is used to obtain an ordered set of scaling ratios, match the terrain image with a terrain profile database based on the ordered set of scaling ratios, and determine the position of the UAV;

[0010] The navigation device is used to navigate the UAV according to the position of the UAV;

[0011] Among them, obtaining the ordered set of scaling ratios includes:

[0012] use represents the kth terrain image captured during the flight, and U represents the set of all zoom ratios;

[0013] Then for the scaling factor z in U, During the processing, the ranking value is calculated as follows:

[0014] ;

[0015] represents the sorted value of z, and denote the first weight and the second weight respectively, Indicates the judgment value of the scaling ratio z. If and ,but is 1; if and ,but is 1; if and ,but is 1, in other cases, is 0; Indicates the maximum value of the scaling factor in U. and They represent the final scaling ratios of the k-1th and k-2th terrain images respectively;

[0016] The scaling ratios in U are sorted in descending order of the sorting values ​​to obtain an ordered set of scaling ratios.

[0017] Preferably, matching the terrain image with a terrain contour database based on an ordered set of scaling ratios to determine the position of the drone comprises:

[0018] S1, for the i-th scaling ratio in the ordered set of scaling ratios, scaling the terrain image using the i-th scaling ratio to obtain a scaled image, , N represents the total number of scaling ratios in the ordered set of scaling ratios;

[0019] S2, obtaining feature data of terrain contour from the zoomed image;

[0020] S3, matching the feature data with the terrain profile database to obtain the similarity of the feature data;

[0021] S4, judging whether the similarity is higher than the set similarity threshold, if so, then ending the calculation, estimating the height of the UAV according to the i-th scaling ratio; and using the longitude and latitude corresponding to the feature data in the terrain profile database as the longitude and latitude of the UAV; if not, then entering S5;

[0022] S5, add 1 to the value of i and enter S1.

[0023] Preferably, acquiring characteristic data of terrain contour from the zoomed image comprises:

[0024] filtering the scaled image to obtain a filtered image;

[0025] Feature extraction is performed on the filtered image to obtain feature data of the terrain contour.

[0026] Preferably, filtering the scaled image to obtain a filtered image includes:

[0027] Grayscale the scaled image to obtain a grayscale image;

[0028] Use the Sobel algorithm to calculate each pixel in the grayscale image, obtain the gradient value of each pixel, and obtain the set H of pixels whose gradient values ​​are greater than the adaptive gradient value threshold;

[0029] The grayscale image is filtered based on the set H to obtain a filtered image.

[0030] Preferably, filtering the grayscale image based on the set H to obtain a filtered image includes:

[0031] Based on the set H, the filtering influence value of each pixel in the grayscale image is calculated respectively;

[0032] Each pixel in the grayscale image is filtered based on the filtering influence value to obtain a filtered image.

[0033] Preferably, the calculation formula of the filtering influence value is:

[0034] ;

[0035] Represents the filtering influence value of pixel v, It represents the number of pixels in set H among all pixels whose distance from v is less than D. is an adaptive value. If v belongs to the set H, then is 1, otherwise is 0, represents the set of all pixels whose distance from v is less than D, express The gray value of pixel u in express The total number of pixels in express The maximum gray value of the pixel in , and Represent the first filter coefficient, the second filter coefficient and the third filter coefficient respectively.

[0036] Preferably, filtering each pixel in the grayscale image based on the filtering influence value to obtain a filtered image includes:

[0037] Use and Indicates the minimum and maximum values ​​of the filtering influence on the pixels in the grayscale image;

[0038] Calculate the filter impact value segment interval:

[0039] ;

[0040] Indicates the segment interval of the filtering influence value, and M represents a set positive integer;

[0041] based on Divide the pixels in the grayscale image into M filter sets;

[0042] The pixels in each filter set are filtered in descending order of the maximum value of the filter influence value of the pixels in the filter set to obtain a filtered image.

[0043] Preferably, based on The pixels in the grayscale image are divided into M sets, including:

[0044] The range of the filter influence value corresponding to the first filter set is ;

[0045] The value range of the filter influence value corresponding to the mth filter set is , m∈[2,M];

[0046] The pixel point is stored in the filter set corresponding to the value range within which the filter influence value of the pixel point falls.

[0047] Preferably, filtering the pixels in each filter set in descending order of the maximum values ​​of the filter influence values ​​of the pixels in the filter set to obtain a filtered image includes:

[0048] S10, sorting the filter set in descending order according to the maximum values ​​of the filter influence values ​​of the pixels in the filter set to obtain a filter ordered set;

[0049] S20, initialize the value of b to 1, and obtain the first filter set from the filter ordered set as the current set;

[0050] S30, in the grayscale image, filtering the pixels in the current set to obtain a relay image ;

[0051] S40, adding 1 to the value of b, obtaining the filter set ranked at position b from the filter ordered set as the current set, and proceeding to S50;

[0052] S50, in Filter the pixels in the current set to obtain the relay image ;

[0053] S60, determine whether the value of b is greater than M, if so, proceed to S70, if not, proceed to S40;

[0054] S70, As the final filtered image.

[0055] Preferably, Filter the pixels in the current set to obtain the relay image ,include:

[0056] The wavelet transform filtering algorithm is used in Filter the pixels in the current set to obtain the relay image .

[0057] Compared with the prior art, the present invention sorts the zoom ratios by calculating the sorting value during the process of terrain matching navigation of the drone, thereby realizing that the zoom ratio with the higher zoom ratio finally determined is arranged at the front position in the ordered set of zoom ratios. The probability of occurrence of the event that the position of the drone can be determined by only identifying the zoom images corresponding to the front part of the zoom ratio in the ordered set is effectively increased, and the probability of occurrence of the time when the positioning result of the drone is obtained in a shorter time is increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 The present invention is a schematic diagram of a terrain comparison navigation system for an unmanned aerial vehicle.

[0060] Figure 2 A schematic diagram of the process of determining the position of a drone according to the present invention. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0062] like Figure 1 In one embodiment shown, the present invention provides a terrain comparison and navigation system for a UAV, comprising a photographing device, a terrain comparison device and a navigation device;

[0063] The photographing device is used to obtain terrain images during the flight of the UAV;

[0064] The terrain comparison device is used to obtain an ordered set of scaling ratios, match the terrain image with a terrain profile database based on the ordered set of scaling ratios, and determine the position of the UAV;

[0065] The navigation device is used to navigate the UAV according to the position of the UAV;

[0066] Among them, obtaining the ordered set of scaling ratios includes:

[0067] use represents the kth terrain image captured during the flight, and U represents the set of all zoom ratios;

[0068] Then for the scaling factor z in U, During the processing, the ranking value is calculated as follows:

[0069] ;

[0070] represents the sorted value of z, and denote the first weight and the second weight respectively, Indicates the judgment value of the scaling ratio z. If and ,but is 1; if and ,but is 1; if and ,but is 1, in other cases, is 0; Indicates the maximum value of the scaling factor in U. and They represent the final scaling ratios of the k-1th and k-2th terrain images respectively;

[0071] The ranking value calculation formula of the present invention can make the ranking value higher when the zoom ratio is closer to the height change trend and the zoom ratio determined by the previous terrain image; and can make the ranking value lower when the zoom ratio is farther away from the height change trend and the zoom ratio determined by the previous terrain image. In this way, the possibility that the ranking value belongs to the correct ranking value can be expressed from two different directions, thereby improving the effectiveness of the ranking value;

[0072] The scaling ratios in U are sorted in descending order of the sorting values ​​to obtain an ordered set of scaling ratios.

[0073] In the process of terrain matching navigation used by the unmanned aerial vehicle, the present invention sorts the zoom ratios by calculating the sorting values, thereby realizing that the zoom ratios with higher zoom ratios that are finally determined are sorted to the front positions in the ordered set of zoom ratios. The probability of occurrence of an event that only requires identification of the zoom images corresponding to the front part of the zoom ratios in the ordered set to determine the position of the unmanned aerial vehicle is effectively increased, and the probability of occurrence of the time when the positioning result of the unmanned aerial vehicle is obtained in a shorter time is increased.

[0074] Preferably, the first weight and the second weight are and .

[0075] Preferably, the scaling ratio of the present invention refers to the scaling ratio between the terrain image and the image based on which the terrain contour database is generated.

[0076] When generating a terrain contour database, it is necessary to take multiple images at different locations on the route and identify the terrain features in these images, such as edges, corners, line segments, etc. These features usually correspond to terrain contours. Then various algorithms are used to estimate the depth information of each point in the image (this can be achieved using machine learning models, image matching and other technologies). The depth information is then converted into a DEM to obtain a grid model representing the ground height. The DEM data is used to draw the contour of the terrain.

[0077] Preferably, if Figure 2 As shown, the terrain image is matched with the terrain contour database based on the ordered set of zoom ratios to determine the position of the UAV, including:

[0078] S1, for the i-th scaling ratio in the ordered set of scaling ratios, scaling the terrain image using the i-th scaling ratio to obtain a scaled image, , N represents the total number of scaling ratios in the ordered set of scaling ratios;

[0079] S2, obtaining feature data of terrain contour from the zoomed image;

[0080] S3, matching the feature data with the terrain profile database to obtain the similarity of the feature data;

[0081] S4, judging whether the similarity is higher than the set similarity threshold, if so, then ending the calculation, estimating the height of the UAV according to the i-th scaling ratio; and using the longitude and latitude corresponding to the feature data in the terrain profile database as the longitude and latitude of the UAV; if not, then entering S5;

[0082] S5, add 1 to the value of i and enter S1.

[0083] The present invention obtains feature data of terrain contours in zoomed images corresponding to different zoom ratios, and then matches these feature data with feature data corresponding to terrain contours at different locations in a terrain contour database. The probability that the terrain image belongs to a certain longitude and latitude is represented by the similarity between the feature data, thereby realizing the estimation of the longitude and latitude of the UAV.

[0084] In S4, when the calculation is finished, the i-th scaling ratio is the finally determined scaling ratio corresponding to the terrain image.

[0085] Preferably, the scaling ratio can be in the range of . You can do this every fixed value range, for example, every tenth. Take a value as the scaling factor.

[0086] Preferably, the process of obtaining the longitude and latitude corresponding to the feature data in the terrain profile database is as follows:

[0087] Obtaining a set A of pixel points of contour lines corresponding to feature data having a similarity with feature data higher than a set similarity threshold in a terrain contour database;

[0088] The longitude and latitude of the midpoint of the pixel points in set A are used as the longitude and latitude corresponding to the feature data in the terrain profile database.

[0089] Preferably, the similarity threshold is 0.9.

[0090] Preferably, estimating the height of the drone according to the i-th scaling ratio comprises:

[0091] H1 represents the shooting height of the terrain image obtained when generating the terrain profile database;

[0092] Let zi represent the i-th scaling ratio, then the estimation formula for the height of the drone is:

[0093] ;

[0094] H2 represents the estimated altitude of the drone.

[0095] Specifically, if the scaling ratio is greater than 1, it means that the height of the drone is greater than the shooting height of the terrain image obtained when the terrain profile database is generated; if the scaling ratio is less than 1, it means that the height of the drone is less than the shooting height of the terrain image obtained when the terrain profile database is generated. Therefore, the height of the drone can be estimated by the scaling ratio and the shooting height of the terrain image obtained when the terrain profile database is generated.

[0096] When generating a terrain profile database, the shooting heights of terrain images acquired at different locations are all the same.

[0097] Preferably, acquiring characteristic data of terrain contour from the zoomed image comprises:

[0098] filtering the scaled image to obtain a filtered image;

[0099] Feature extraction is performed on the filtered image to obtain feature data of the terrain contour.

[0100] Filtering can effectively reduce or eliminate random noise in images and enhance specific features in images, such as edges, textures, or shapes, thereby improving image quality and facilitating the subsequent extraction of more accurate feature data.

[0101] Preferably, filtering the scaled image to obtain a filtered image includes:

[0102] Grayscale the scaled image to obtain a grayscale image;

[0103] Use the Sobel algorithm to calculate each pixel in the grayscale image, obtain the gradient value of each pixel, and obtain the set H of pixels whose gradient values ​​are greater than the adaptive gradient value threshold;

[0104] The grayscale image is filtered based on the set H to obtain a filtered image.

[0105] The traditional filtering method generally only filters each pixel in the original grayscale image. However, this filtering method does not take into account that the pixels in the neighborhood used as references during filtering are likely to be noise pixels. Obviously, this filtering method does not provide a high enough accuracy for the filtering results, as the accuracy of the reference data is not high enough.

[0106] The present invention improves this method by acquiring the set H to control the filtering order of different pixel points to improve the accuracy of the reference data, thereby improving the accuracy of the filtering.

[0107] Preferably, the adaptive gradient value threshold is four fifths of the maximum value of the gradient value in the grayscale image.

[0108] Preferably, filtering the grayscale image based on the set H to obtain a filtered image includes:

[0109] Based on the set H, the filtering influence value of each pixel in the grayscale image is calculated respectively;

[0110] Each pixel in the grayscale image is filtered based on the filtering influence value to obtain a filtered image.

[0111] In the present invention, the larger the filtering influence value is, the more important it is and the higher its reference value to the surrounding pixels. Therefore, the present invention performs filtering based on the filtering influence value, which can improve the accuracy of the grayscale values ​​of the pixels in the neighborhood referenced in the subsequent filtering process and obtain better filtering results.

[0112] Preferably, the calculation formula of the filtering influence value is:

[0113] ;

[0114] Represents the filtering influence value of pixel v, It represents the number of pixels in set H among all pixels whose distance from v is less than D. is an adaptive value. If v belongs to the set H, then is 1, otherwise is 0, represents the set of all pixels whose distance from v is less than D, express The gray value of pixel u in express The total number of pixels in express The maximum gray value of the pixel in , and Represent the first filter coefficient, the second filter coefficient and the third filter coefficient respectively.

[0115] In the present invention, when the number of pixel points belonging to set H around the pixel point increases, the probability that the pixel point belongs to set H is higher, and the grayscale value change of other pixel points around the pixel point is more drastic, the filtering influence value of the pixel point is greater. In this way, the pixel points with a greater degree of influence on the filtering results of the surrounding pixels can be filtered earlier, and a more accurate filtering result can be obtained.

[0116] Preferably, the first filter coefficient, the second filter coefficient and the third filter coefficient are respectively , and .

[0117] Preferably, the value of D is 5.

[0118] Preferably, filtering each pixel in the grayscale image based on the filtering influence value to obtain a filtered image includes:

[0119] Use and Indicates the minimum and maximum values ​​of the filtering influence on the pixels in the grayscale image;

[0120] Calculate the filter impact value segment interval:

[0121] ;

[0122] Indicates the segment interval of the filtering influence value, and M represents a set positive integer;

[0123] based on Divide the pixels in the grayscale image into M filter sets;

[0124] The pixels in each filter set are filtered in descending order of the maximum value of the filter influence value of the pixels in the filter set to obtain a filtered image.

[0125] The present invention does not directly determine the filtering order of different pixels based on the filtering influence value, because this will result in too many images of the intermediate filtering process, affecting the filtering efficiency. Therefore, the present invention segments the filtering influence value, and each only filters the pixels corresponding to the filtering influence value belonging to the same segment interval, thereby achieving the improvement of filtering efficiency while providing more effective reference data for the subsequent filtering process.

[0126] Preferably, the set positive integer is 10.

[0127] Preferably, based on The pixels in the grayscale image are divided into M sets, including:

[0128] The range of the filter influence value corresponding to the first filter set is ;

[0129] The value range of the filter influence value corresponding to the mth filter set is , m∈[2,M];

[0130] The pixel point is stored in the filter set corresponding to the value range within which the filter influence value of the pixel point falls.

[0131] Preferably, filtering the pixels in each filter set in descending order of the maximum values ​​of the filter influence values ​​of the pixels in the filter set to obtain a filtered image includes:

[0132] S10, sorting the filter set in descending order according to the maximum values ​​of the filter influence values ​​of the pixels in the filter set to obtain a filter ordered set;

[0133] S20, initialize the value of b to 1, and obtain the first filter set from the filter ordered set as the current set;

[0134] S30, in the grayscale image, filtering the pixels in the current set to obtain a relay image ;

[0135] S40, adding 1 to the value of b, obtaining the filter set ranked at position b from the filter ordered set as the current set, and proceeding to S50;

[0136] S50, in Filter the pixels in the current set to obtain the relay image ;

[0137] S60, determine whether the value of b is greater than M, if so, proceed to S70, if not, proceed to S40;

[0138] S70, As the final filtered image.

[0139] After obtaining the filter set, the present invention sorts the filtering order of the filter set based on the maximum value of the filtering influence value of the pixel points in the filter set. When filtering the pixel points in the latter filter set, it is based on the previously accumulated filtering results, so that the later the filtering process is, the more accurate the data that can be referenced is, which effectively improves the quality of the final filtered image.

[0140] Preferably, filtering the pixels in the current set includes:

[0141] The pixels in the current set are filtered in descending order of filtering impact value.

[0142] Preferably, filtering the pixels in the current set includes:

[0143] Filter each pixel in the current set in a random order.

[0144] Preferably, Filter the pixels in the current set to obtain the relay image ,include:

[0145] The wavelet transform filtering algorithm is used in Filter the pixels in the current set to obtain the relay image .

[0146] Wavelet transform filtering algorithm is a filtering technology based on wavelet analysis. It mainly decomposes the frequency components of the signal through wavelet transform and removes noise using threshold processing method.

[0147] Preferably, navigating the drone according to the position of the drone includes:

[0148] Get the next waypoint of the drone;

[0149] Plan a flight path based on the current location and the target location of the next waypoint. The flight path may include a straight path or a path obtained after obstacle avoidance optimization based on terrain, obstacles and other conditions.

[0150] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A terrain comparison navigation system for unmanned aerial vehicles, characterized in that: It includes a photographing device, a terrain comparison device and a navigation device; The photographing device is used to obtain terrain images during the flight of the UAV; The terrain comparison device is used to obtain an ordered set of scaling ratios, match the terrain image with a terrain profile database based on the ordered set of scaling ratios, and determine the position of the UAV; The navigation device is used to navigate the UAV according to the position of the UAV; Among them, obtaining the ordered set of scaling ratios includes: use represents the kth terrain image captured during the flight, and U represents the set of all zoom ratios; Then for the scaling factor z in U, During the processing, the ranking value is calculated as follows: ; represents the sorted value of z, and denote the first weight and the second weight respectively, Indicates the judgment value of the scaling ratio z. If and ,but is 1; if and ,but is 1; if and ,but is 1, in other cases, is 0; Indicates the maximum value of the scaling factor in U. and They represent the final scaling ratios of the k-1th and k-2th terrain images respectively; The scaling ratios in U are sorted in descending order of the sorting values ​​to obtain an ordered set of scaling ratios.

2. A terrain comparison navigation system for unmanned aerial vehicles according to claim 1, characterized in that: Determine the position of the drone by matching the terrain image to a database of terrain contours based on an ordered set of scales, including: S1, for the i-th scaling ratio in the ordered set of scaling ratios, scaling the terrain image using the i-th scaling ratio to obtain a scaled image, , N represents the total number of scaling ratios in the ordered set of scaling ratios; S2, obtaining feature data of terrain contour from the zoomed image; S3, matching the feature data with the terrain profile database to obtain the similarity of the feature data; S4, judging whether the similarity is higher than the set similarity threshold, if so, then ending the calculation, estimating the height of the UAV according to the i-th scaling ratio; and using the longitude and latitude corresponding to the feature data in the terrain profile database as the longitude and latitude of the UAV; if not, then entering S5; S5, add 1 to the value of i and enter S1.

3. A terrain comparison navigation system for unmanned aerial vehicles according to claim 1, characterized in that: Get terrain contour feature data from the zoomed image, including: filtering the scaled image to obtain a filtered image; Feature extraction is performed on the filtered image to obtain feature data of the terrain contour.

4. A terrain comparison navigation system for unmanned aerial vehicles according to claim 3, characterized in that: Filtering the scaled image to obtain a filtered image includes: Grayscale the scaled image to obtain a grayscale image; Use the Sobel algorithm to calculate each pixel in the grayscale image, obtain the gradient value of each pixel, and obtain the set H of pixels whose gradient values ​​are greater than the adaptive gradient value threshold; The grayscale image is filtered based on the set H to obtain a filtered image.

5. A terrain comparison navigation system for unmanned aerial vehicles according to claim 4, characterized in that: The grayscale image is filtered based on the set H to obtain a filtered image, including: Based on the set H, the filtering influence value of each pixel in the grayscale image is calculated respectively; Each pixel in the grayscale image is filtered based on the filtering influence value to obtain a filtered image.

6. A terrain comparison navigation system for unmanned aerial vehicles according to claim 5, characterized in that: The calculation formula of the filtering influence value is: ; Represents the filtering influence value of pixel v, It represents the number of pixels in set H among all pixels whose distance from v is less than D. is an adaptive value. If v belongs to the set H, then is 1, otherwise is 0, represents the set of all pixels whose distance from v is less than D, express The gray value of pixel u in express The total number of pixels in express The maximum gray value of the pixel in , and Represent the first filter coefficient, the second filter coefficient and the third filter coefficient respectively.

7. The terrain comparison navigation system for unmanned aerial vehicles according to claim 5, characterized in that: Filtering each pixel in the grayscale image based on the filtering influence value to obtain a filtered image, including: Use and Indicates the minimum and maximum values ​​of the filtering influence on the pixels in the grayscale image; Calculate the filter impact value segment interval: ； Indicates the segment interval of the filtering influence value, and M represents a set positive integer; based on Divide the pixels in the grayscale image into M filter sets; The pixels in each filter set are filtered in descending order of the maximum value of the filter influence value of the pixels in the filter set to obtain a filtered image.

8. A terrain comparison navigation system for unmanned aerial vehicles according to claim 7, characterized in that: based on The pixels in the grayscale image are divided into M sets, including: The range of the filter influence value corresponding to the first filter set is ; The value range of the filter influence value corresponding to the mth filter set is ,m∈[2,M]; The pixel point is stored in the filter set corresponding to the value range within which the filter influence value of the pixel point falls.

9. The terrain comparison navigation system for unmanned aerial vehicles according to claim 7, characterized in that: According to the maximum value of the filter influence value of the pixel points in the filter set from large to small, the pixel points in each filter set are filtered respectively to obtain a filtered image, including: S10, sorting the filter set in descending order according to the maximum values ​​of the filter influence values ​​of the pixels in the filter set to obtain a filter ordered set; S20, initialize the value of b to 1, and obtain the first filter set from the filter ordered set as the current set; S30, in the grayscale image, filtering the pixels in the current set to obtain a relay image ; S40, adding 1 to the value of b, obtaining the filter set ranked at position b from the filter ordered set as the current set, and proceeding to S50; S50, in Filter the pixels in the current set to obtain the relay image ; S60, determine whether the value of b is greater than M, if so, proceed to S70, if not, proceed to S40; S70, As the final filtered image.

10. A terrain comparison navigation system for unmanned aerial vehicles according to claim 9, characterized in that: exist Filter the pixels in the current set to obtain the relay image ,include: The wavelet transform filtering algorithm is used in Filter the pixels in the current set to obtain the relay image .

Images