Airport runway foreign matter identification method based on triangle barycenter feature point matching

Abstract

The application relates to the technical field of airport runway foreign matter identification, in particular to an airport runway foreign matter identification method based on triangle barycenter feature point matching, which realizes high-precision feature point calculation by using a SIFT feature point matching algorithm, and then combines pixel values in a triangle region formed by the triangle feature points to judge whether the precision of corresponding matching points between a reference image and a to-be-matched image meets the requirements, so that the matching points meeting the precision requirements are reserved and the matching points not meeting the requirements are eliminated, the image feature point matching is more accurate, the matching speed is faster, and the speed and accuracy of foreign matter identification in the airport runway shooting image are increased.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of airport runway foreign object debris identification, and particularly relates to an airport runway foreign object debris identification method based on triangle barycenter feature point matching. BACKGROUND

[0002] The airport runway foreign object debris (FOD) refers to any foreign object that does not belong to the airport but appears in the airport area and can damage the aviation instruments or aviation systems, and the FOD is mainly composed of various scattered parts of the aircraft, metal tools, stones, wood blocks and plastic products.

[0003] The artificial detection is the most common FOD detection method at present, which has the advantages of detecting millimeter-level FOD, but the comprehensive quality of the detection personnel is high, and the reliability is not high in general due to the influence of the external conditions such as personal emotion, responsibility, working environment, labor intensity, light intensity and the like.

[0004] At present, three FOD detection systems (Tarsier, FODetect and FODFinder) in the FAA-certified FOD system in the world are based on millimeter wave radar technology. Although the FOD detection system based on the millimeter wave radar is not affected by the day, night and visibility, it still has the following limitations: ① the precision of detecting and identifying foreign objects is still poor, and only foreign objects of 5 cm or more can be identified, and millimeter-level foreign objects (such as lead bullets) cannot be identified; ② video detection is needed as an auxiliary detection means; ③ the equipment cost is high, and the operation and maintenance cost is also high. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an airport runway foreign object debris identification method based on triangle barycenter feature point matching, which improves the accuracy of image feature point matching to solve the problem of low identification precision of airport runway foreign object debris in the prior art.

[0006] The technical solution provided by the present application is as follows:

[0007] An airport runway foreign object debris identification method based on triangle barycenter feature point matching, characterized in that it comprises the following steps:

[0008] S1, photographing the same position of the airport runway from different angles to obtain multiple angle images of the same position of the airport runway;

[0009] S2, selecting one image obtained in step S1 as a reference image, and selecting the remaining images obtained in step S1 as to-be-matched images;

[0010] S3, performing feature point matching on the reference image and the image to be matched using a SIFT algorithm to obtain corresponding matching point pairs;

[0011] S4, removing false matching point pairs in the matching point pairs based on a triangle barycenter principle;

[0012] S5, performing image fusion on the reference image and the image to be matched according to the matching point pairs after removal to complete foreign matter recognition.

[0013] Further limitation, the step S4 includes removing one-to-many false matching point pairs, and the step of removing one-to-many false matching point pairs is specifically:

[0014] In the reference image, any feature point B is selected as a reference feature point, the nearest feature point A and the nearest feature point C with the shortest Euclidean distance from the feature point B are obtained by using Euclidean distance calculation, and a reference triangle is obtained through the nearest feature point A, the reference feature point B and the nearest feature point C;

[0015] According to the nearest feature point A, the reference feature point B and the nearest feature point C, corresponding feature points a, b and c in the matching image are found, and a to-be-detected triangle is obtained through the feature points a, b and c;

[0016] The gray values contained in the reference triangle and the gray values contained in the to-be-detected triangle are respectively determined by vector representation method;

[0017] The barycenter H of the reference triangle is obtained according to the gray values in the reference triangle, and the barycenter h of the to-be-detected triangle is obtained according to the gray values in the to-be-detected triangle;

[0018] According to the matching error of the relative positions between the reference feature point B and the barycenter H and the relative positions between the feature point b and the barycenter h, one-to-many false matching point pairs are removed.

[0019] Further limitation, the calculation method of the barycenter H and the barycenter h is:

[0020]

[0021]

[0022] Wherein, ABC is the reference triangle, abc is the to-be-detected triangle, (x Ai , y Ai ) is the coordinate of any pixel point in the reference triangle region, P Ai is the gray value of the pixel at the corresponding pixel point coordinate, X c is the horizontal coordinate of the barycenter H of the reference triangle, Y c is the vertical coordinate of the barycenter H of the reference triangle; (x aj , yaj ) is any pixel point coordinate in the triangle region to be detected, P aj is the gray value of the pixel at the corresponding pixel point coordinate, M c is the horizontal coordinate of the center of gravity h of the detected triangle, N c is the vertical coordinate of the center of gravity h of the detected triangle.

[0023] Further limitation, the step S4 further includes eliminating the cross-line error matching point pair before eliminating a pair of multiple error matching point pairs, and the step of eliminating the cross-line error matching point pair is specifically:

[0024] A two-dimensional linear model is established for the matching point pair obtained in step S3:

[0025]

[0026] Wherein, (x i , y i ) is the coordinate of any feature point on the reference image in the reference image, (x j , y j ) is the coordinate of the corresponding feature point on the to-be-matched image in the to-be-matched image, len α is the horizontal vector of the reference image relative to the to-be-matched image, and Grad is the corresponding two-dimensional linear value of any matching point pair.

[0027] The corresponding two-dimensional linear values calculated from all matching point pairs between the reference image and the to-be-matched image are counted, and the number S i of corresponding two-dimensional linear values in different statistical scale ranges is calculated in combination with the statistical scale.

[0028] Wherein, Min(Grad i ) is the minimum value of two-dimensional linearity, Max(Grad i ) is the maximum value of two-dimensional linearity, k is the total number of statistical scales, and i is the i-th statistical scale.

[0029] According to S i <Max(S i )*n, the two-dimensional linear values located in the statistical scale range are eliminated, and the corresponding matching point pairs are eliminated according to the eliminated two-dimensional linear values, wherein n is the elimination factor.

[0030] Further limitation, the calculation of the statistical scale Scale i is:

[0031]

[0032] Wherein, Min(Grad) is the minimum value of two-dimensional linearity, Max(Grad) is the maximum value of two-dimensional linearity, k is the total number of statistical scales, and i is the i-th statistical scale.

[0033] Further limited, the step S5 specifically includes the following steps:

[0034] An affine transformation matrix model is established according to the matching point pair obtained according to the step S4, and the to-be-identified image is converted into a to-be-recognized image unified with the field of view of the reference image according to the affine transformation matrix model;

[0035] The reference image and the to-be-identified image are image fused and the foreign matter is displayed through image brightness mean value calculation;

[0036] The foreign matter is recognized and marked through a canny edge detection operator, and the foreign matter recognition is completed.

[0037] Further limited, the image brightness mean value MoI is calculated as:

[0038]

[0039] Wherein, and are the gray scales of the to-be-identified image, is the gray scale of the reference image.

[0040] The present application has the advantages of:

[0041] The airport runway foreign matter recognition method based on the triangular barycenter feature point matching provided by the present application realizes feature point calculation with high matching precision by using the SIFT feature point matching algorithm, and then judges whether the precision of the corresponding matching points between the reference image and the to-be-matched image meets the requirements or not by combining the pixel values in the triangular region formed by the triangular feature points, so as to realize the retention of the matching points meeting the precision requirements and the elimination of the matching points not meeting the requirements, so that the image feature point matching is more accurate, the matching speed is faster, and the speed and accuracy of the foreign matter recognition in the airport runway photographed image are increased. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 It is the method step diagram of the embodiment 1 of the present application;

[0043] Figure 2 It is the airport runway image photographed in the embodiment 1 of the present application;

[0044] Figure 3 It is the number statistical diagram of the two-dimensional linear values in different value ranges in the embodiment 2 of the present application;

[0045] Figure 4 It is the reference triangular schematic diagram and the to-be-detected triangular schematic diagram in the embodiment 2 of the present application, wherein, Figure 4 a is a reference triangular schematic diagram in the reference image, Figure 4b is a triangle diagram to be detected in the image to be matched;

[0046] Figure 5 the diagram for the pixel point P at any position in any triangle in embodiment 2 of the present application is calculated;

[0047] Figure 6 the fusion image obtained in embodiment 3 of the present application is calculated;

[0048] Figure 7 the foreign matter marking diagram after the runway image is fused in embodiment 3 of the present application. DETAILED DESCRIPTION

[0049] Embodiment 1

[0050] Reference Figure 1 The embodiment provides an airport runway foreign matter identification method based on triangle barycenter feature point matching, and comprises the following steps:

[0051] S1, photographing the same position of the airport runway from different angles to obtain multiple angle images of the same position of the airport runway;

[0052] Specifically, in step S1, photographing can be performed while the trolley moves on the airport runway. Since the photograph is a two-dimensional planar figure, if there is foreign matter, it usually needs to be photographed from not less than 2 angles. The foreign matter image from multiple angles can distinguish the imaging difference of the three-dimensional object in the airport runway at different angles to identify the foreign matter. In order to photograph the airport runway from multiple angles, multiple cameras can be installed on the trolley to take pictures at the same time. For example, 4 cameras are installed, the lenses of the 4 cameras are inclined downward, and the same position of the airport runway is photographed. The shooting pictures are basically the same, the obtained images are different in angle, the center of the photographed picture belongs to the same position of the airport runway, and the different cameras are arranged at equal intervals in a circle due to different setting directions. Reference Figure 2 Multiple angle images of the same position of the airport runway are obtained by the 4 cameras. The trolley will photograph multiple groups of images of the same position during the running and moving process, and it is necessary to analyze and judge whether there is foreign matter in the airport runway at each position.

[0053] S2, selecting one image obtained in step S1 as a reference image and selecting the remaining images obtained in step S1 as images to be matched;

[0054] Specifically, in step S2, one of the obtained 4 images is selected as a reference image, and the remaining three images are selected as images to be matched. For example, the image in the upper left corner of Figure 2 is selected as a reference image, and the other three images are selected as images to be matched.

[0055] S3, performing feature point matching on the reference image and the to-be-matched images using the SIFT algorithm to obtain corresponding matching point pairs;

[0056] Specifically, in step S3, the reference image and the three to-be-matched images need to be matched in terms of feature points, and the SIFT algorithm is used to calculate the corresponding matching point pairs between the reference image and the three to-be-matched images, that is, there are matching point pairs between the reference image and each to-be-matched image.

[0057] S4, removing the erroneous matching point pairs in the matching point pairs based on the principle of triangle barycenter;

[0058] Since the matching point pairs obtained by using the SIFT algorithm contain erroneous matching point pairs, the erroneous matching point pairs are different from the feature points of the reference image and the feature points of the to-be-matched images but are regarded as a matching point pair, which leads to different pixel gray scales and / or positions of the matching point pairs of the two images. The existence of the erroneous matching point pairs is caused by whether the image contains foreign objects. The erroneous matching point pairs can be located at the position of the foreign object image of the airport runway or at the position of the airport runway plane, which leads to an inaccurate image radiation conversion model established by using the matching point pairs, and further leads to an inaccurate image after radiation conversion, and finally leads to a failure or low accuracy of foreign object recognition. Therefore, the erroneous matching point pairs are removed in step S4, so as to ensure the accuracy of foreign object recognition and reduce the calculation time of foreign object recognition and improve the recognition efficiency.

[0059] S5, performing image fusion on the reference image and the to-be-matched images according to the removed matching point pairs to complete foreign object recognition.

[0060] Specifically, after the erroneous matching points are removed in step S4, the corresponding affine transformation matrix model is established by using the matching point pairs of the reference image and each to-be-matched image. The to-be-matched image is affine transformed into a to-be-recognized image unified with the field of view of the reference image by using the corresponding affine transformation matrix model. Then, the reference image and the three to-be-recognized images are fused to obtain a fused image. If there is a foreign object on the airport runway, the foreign object is identified and marked by using the gray scale difference and the canny edge detection operator, and the foreign object recognition is completed.

[0061] Embodiment 2

[0062] Further, step S4 specifically includes the following steps:

[0063] S42, rejecting a pair of multi-error matching point pairs, in order to reduce the calculation time of the error matching point pairs in step S42, it is further preferred that step S4 further comprises step S41: rejecting the cross-linear error matching point pairs, that is, before performing the step of rejecting a pair of multi-error matching point pairs, the step of rejecting the cross-linear error matching point pairs is performed first; wherein the cross-linear error matching point pairs are mainly the matching point pairs whose line slopes are greatly different from the line slopes between most matching single pairs, and the matching point pairs are considered as error matching point pairs.

[0064] The step of rejecting the cross-linear error matching point pairs is specifically:

[0065] S411, establishing a two-dimensional linear model for the matching point pairs obtained in step S3:

[0066]

[0067] Wherein, (x i , y i ) is the coordinates of any feature point on the reference image in the reference image, (x j , y j ) is the coordinates of the corresponding feature point on the to-be-matched image in the to-be-matched image, len α is the horizontal dimension of the reference image relative to the to-be-matched image, and Grad is the corresponding two-dimensional linear value of any matching point pair.

[0068] S412, calculating the corresponding two-dimensional linear values of all matching point pairs between the reference image and the to-be-matched image, and combining the statistical scale to calculate the number S i of the corresponding two-dimensional linear values in different statistical scale ranges. i The calculation of the statistical scale Scale

[0069]

[0070] Wherein, Min(Grad i ) is the minimum two-dimensional linear value, Max(Grad i ) is the maximum two-dimensional linear value, k is the total number of statistical scales, i is the i-th statistical scale, and k is a positive integer, preferably k is greater than 2.

[0071] S413, according to S i <Max(S i )*n, the two-dimensional linear values located in the statistical scale range are rejected, and the corresponding matching point pairs are rejected according to the rejected two-dimensional linear values, wherein n is a rejection factor.

[0072] Specifically, by calculating the slope of the connecting line between the feature points of the reference image and the to-be-matched image, obviously incorrect matching point pairs can be removed, for example, the connecting line slope of the feature point pairs between the reference image and the to-be-matched image is mostly 1, and the matching point pairs corresponding to the slope of 2 or -0.4, which deviates greatly from 1, can be removed.

[0073] The slope of the connecting line of the matching point pairs, i.e., the two-dimensional linear value corresponding to the matching point pairs, can be obtained through the established two-dimensional linear model. The matching point pairs of the reference image and three to-be-matched images are calculated through the two-dimensional linear model respectively, and three groups of data are obtained, each group of data containing a two-dimensional linear value corresponding to all matching point pairs. The two-dimensional linear values in each group of data are counted to obtain the minimum value Min(Grad) and the maximum value Max(Grad) in all two-dimensional linear values, and the number corresponding to each two-dimensional linear value can also be counted.

[0074] In order to facilitate the calculation of the number corresponding to different two-dimensional linear values, the number of two-dimensional linear values in different value ranges is calculated using statistical scales, for example, Figure 3 As described above, the horizontal coordinate in the graph is the two-dimensional linear value, and the vertical coordinate is the number. As can be seen from the graph, the number of two-dimensional linear values in the range of -0.5 to -0.25 is the largest, which is 200. It can be considered that when the two-dimensional linear value corresponding to the matching point pair is in the range of -0.5 to -0.25, the matching point pair is basically accurate. The matching point pairs that may be incorrect can be removed. In order to avoid too many matching point pairs being removed or too many correct matching point pairs being removed, the matching point pairs that are basically incorrect are removed by using a removal factor to improve the accuracy of removal. The removal factor is selected as 0.2, for example. When the number of two-dimensional linear values in any statistical scale is less than 40, it is determined that the matching point pair corresponding to the two-dimensional linear value is an incorrect matching point pair and can be removed, thereby completing the preliminary removal and screening of the matching point pairs.

[0075] The steps of removing the one-to-many incorrect matching point pairs are as follows:

[0076] S421, selecting any feature point B in the reference image as a reference feature point, calculating the nearest feature point A and the nearest feature point C with the shortest Euclidean distance from the feature point B using the Euclidean distance, and obtaining a reference triangle through the nearest feature point A, the reference feature point B, and the nearest feature point C;

[0077] S422, finding the corresponding feature point a, feature point b, and feature point c in the matching image according to the nearest feature point A, the reference feature point B, and the nearest feature point C, and obtaining a to-be-detected triangle through the feature point a, the feature point b, and the feature point c;

[0078] S423, determining the gray values contained in the reference triangle and the gray values contained in the to-be-detected triangle respectively through vector representation;

[0079] S424, obtaining the center of gravity H of the reference triangle according to the gray value in the reference triangle, and obtaining the center of gravity h of the to-be-detected triangle according to the gray value in the to-be-detected triangle, and the calculation method of the center of gravity H and the center of gravity h is as follows:

[0080]

[0081]

[0082] Wherein, ABC is the reference triangle, abc is the to-be-detected triangle, (x Ai , y Ai ) is the coordinate of any pixel point in the reference triangle region, P Ai is the gray value of the pixel at the corresponding pixel point coordinate, X c is the horizontal coordinate of the center of gravity H of the reference triangle, Y c is the vertical coordinate of the center of gravity H of the reference triangle; (x aj , y aj ) is the coordinate of any pixel point in the to-be-detected triangle region, P aj is the gray value of the pixel at the corresponding pixel point coordinate, M c is the horizontal coordinate of the center of gravity h of the to-be-detected triangle, and N c is the vertical coordinate of the center of gravity h of the to-be-detected triangle;

[0083] S425, according to the matching error of the relative positions between the reference feature point B and the center of gravity H and the relative positions between the feature point b and the center of gravity h, the one-to-many error matching point pair is removed.

[0084] Specifically, after the removal of the cross line shape error matching point pair, the removal of the one-to-many error matching point pair is started, and the one-to-many error matching point pair represents that one feature point on the reference image and multiple feature points on one to-be-matched image form a matching point pair, at this time, it is considered that the one-to-many matching point pair is also error, and should be discarded.

[0085] Reference Figure 4 a, first, select any feature point B in the reference image as a reference feature point, then calculate the two feature points A and C with the shortest distance from the reference feature point by using the Euclidean distance, take the feature point A as a similar feature point A, and take the feature point C as a similar feature point C, so as to construct a reference triangle ABC according to the three feature points on the reference image, and correspondingly, reference Figure 4 b, according to the similar feature point A, the reference feature point B and the similar feature point C, find the corresponding feature points a, b and c in the to-be-matched image, and through the feature points a, b and c, a to-be-detected triangle abc can be constructed on the to-be-matched image.

[0086] refer to Figure 5 For example, a pixel P at any position in any triangle △XYZ can be represented as:

[0087] P = X + u(ZX) + v*(YX)

[0088] Where X, Y, Z, and P represent the pixel coordinates of corresponding points in △XYZ, and u and v are scalar coefficients, when P satisfies:

[0089] u≥0, v≥0, u+v≤1

[0090] Right now By dot product of vectors AC and AB on both sides of the equation and rearranging, we get:

[0091]

[0092]

[0093] This allows us to obtain all the pixels contained within the three feature regions. By analyzing these pixels, we can obtain the grayscale values ​​of the corresponding pixels in the reference triangle and the corresponding pixels in the triangle to be detected.

[0094] Starting from the centroids obtained from the reference triangle and the triangle to be detected, the matching error E is calculated based on the relative positions between reference feature point B and centroid H, and between feature point b and centroid h. Figure 4 Calculate the errors e1 between ∠α and ∠δ and between ∠β and ∠δ, respectively. Error e2 between:

[0095]

[0096]

[0097] E = e1 + e1

[0098] When the matching error E is greater than 0.02, the matching point pair consisting of the corresponding reference feature point B and feature point b is considered inaccurate. That is, when E≤0.02, the corresponding matching single point pair is retained, and when E>0.02, the corresponding matching point pair is removed. This completes the removal of erroneous matching point pairs and improves the accuracy of foreign object identification.

[0099] Referring to Table 1, different image feature point matching algorithms were used to simultaneously perform feature point matching on different images, and the resulting matching point pairs were analyzed and statistically analyzed.

[0100] Table 1 shows the results of feature point matching tests on different images using different image feature point matching algorithms.

[0101]

[0102] It can be directly obtained from Table 1 that the airport runway foreign matter recognition method based on the triangular barycenter feature point matching provided by the application has less time for image feature point matching, higher number of correctly matched points, and no error matched point pairs appear in the tested images, and high accuracy.

[0103] Embodiment 3

[0104] Based on Embodiment 2, step S5 specifically includes the following steps:

[0105] S51, an affine transformation matrix model is established according to the matched point pairs obtained in step S4, and the to-be-matched image is converted into a to-be-recognized image unified with the field of view of the reference image according to the affine transformation matrix model;

[0106] S52, the reference image and the to-be-recognized image are image fused and the foreign matter is displayed through image brightness mean value calculation, and the image brightness mean value Mol is calculated as:

[0107]

[0108] wherein, and are the gray scales of the to-be-recognized image, is the gray scale of the reference image;

[0109] S53, the foreign matter is recognized and marked through a canny edge detection operator, and the foreign matter recognition is completed.

[0110] Specifically, an affine transformation model is established, and the reference image and each to-be-matched image correspond to a different affine transformation model. After step S4, the matched point pairs with high accuracy are obtained, then the final eliminated matched point pairs are substituted into the corresponding affine transformation model to obtain the result of the perspective transformation matrix, and finally the corresponding to-be-recognized image is converted from the corresponding to-be-matched image through the determined affine transformation model, wherein the affine transformation model is:

[0111]

[0112] wherein, is the position of any pixel point in the reference image f 4 (x,y) is the position of the corresponding pixel point in the to-be-matched image is the position of the corresponding pixel point in the to-be-recognized image after affine transformation, is the position of the corresponding pixel point in the to-be-recognized image is the perspective transformation matrix.

[0113] Through the above operation, the three images to be matched are converted into corresponding three images to be identified through three corresponding affine transformation models, the three images to be identified are unified with the reference image in a field of view, and then the three images to be identified and the reference image are simultaneously input, a fusion image is obtained through image brightness mean value calculation processing, and reference Figure 6 When there is a foreign object in the fusion image, because the observation angles are different, the shape information of the foreign object saved by each image is different, so that the shape information of the foreign object after final fusion cannot be coincided to generate a gray deviation. In order to mark the foreign object in order to count the position, quantity and shape and other information, the external contour and the internal contour of the gray deviation can be extracted by using the canny edge detection operator through the gradient transformation degree between pixels, and the foreign object area is divided into two areas, an internal area m1 and an external area m2 according to the external contour and the internal contour, and whether the foreign object exists at this position is judged through the average gray difference of the two areas:

[0114]

[0115]

[0116] area(m1)-area(m2)≤ε

[0117] Wherein, N represents the number of internal area pixel points, M represents the number of external area pixel points, f(x, y) represents the gray value of the corresponding pixel point, and ε represents a threshold value, which can be selected as 115, and reference Figure 7 When area(m1)-area(m2)≤ε, the corresponding foreign object is marked, and when area(m1)-area(m2)>ε, it is considered that no foreign object is contained, so that the recognition of the foreign object on the airport runway is completed.

Claims

1. An airport runway foreign object identification method based on triangle barycenter feature point matching, characterized in that, The method comprises the following steps: S1, taking pictures of the same position of the airport runway from different angles to obtain multiple angle images of the same position of the airport runway; S2, selecting one image obtained in step S1 as a reference image and selecting the remaining images obtained in step S1 as to-be-matched images; S3, using a SIFT algorithm to perform feature point matching on the reference image and the to-be-matched images to obtain corresponding matching point pairs; S4, removing false matching point pairs in the matching point pairs based on a triangle barycenter principle; The step S4 comprises removing one-to-many false matching point pairs, and the step of removing the one-to-many false matching point pairs is specifically as follows: In the reference image, any feature point B is selected as a reference feature point, the nearest feature point A and the nearest feature point C with the shortest Euclidean distance from the feature point B are calculated by using the Euclidean distance, and a reference triangle is obtained by the nearest feature point A, the reference feature point B and the nearest feature point C; According to the nearest feature point A, the reference feature point B and the nearest feature point C, corresponding feature points a, b and c in the to-be-detected triangle are found in the to-be-matched image, and a to-be-detected triangle is obtained by the feature points a, b and c; The gray values contained in the reference triangle and the gray values contained in the to-be-detected triangle are determined by vector representation method respectively; The barycenter H of the reference triangle is obtained according to the gray values in the reference triangle, and the barycenter h of the to-be-detected triangle is obtained according to the gray values in the to-be-detected triangle; The one-to-many false matching point pairs are removed according to the matching error of the relative positions between the reference feature point B and the barycenter H and the relative positions between the feature point b and the barycenter h; S5, performing image fusion on the reference image and the to-be-matched image according to the matching point pairs after removal to complete foreign matter recognition; The step S5 specifically comprises the following steps: An affine transformation matrix model is established according to the matching point pairs obtained in step S4, and the to-be-matched image is converted into a to-be-recognized image unified with the field of view of the reference image according to the affine transformation matrix model; The reference image and the to-be-recognized image are fused by image brightness mean value calculation, and the foreign matter is displayed; The foreign matter is recognized and marked by a canny edge detection operator, and the foreign matter recognition is completed. 2.The airport runway foreign object identification method based on triangle barycenter feature point matching according to claim 1, characterized in that, The calculation method of the barycenter H and the barycenter h is as follows: Wherein, ABC is the reference triangle, abc is the triangle to be detected, (x Ai , y Ai ) is any pixel point coordinate in the reference triangle region, P Ai is the gray value of the pixel at the corresponding pixel point coordinate, X c is the horizontal coordinate of the reference triangle barycenter H form, Y c is the vertical coordinate of the reference triangle barycenter H; (x aj , y aj ) is any pixel point coordinate in the triangle to be detected region, P aj is the gray value of the pixel at the corresponding pixel point coordinate, M c is the horizontal coordinate of the detection triangle barycenter h, N c is the vertical coordinate of the detection triangle barycenter h. 3.The airport runway foreign object identification method based on triangle barycenter feature point matching of claim 1, characterized in that, Before removing the one-to-many false matching point pairs, the step S4 further comprises removing a cross-line false matching point pair, and the step of removing the cross-line false matching point pair is specifically as follows: A two-dimensional linear model is established for the matching point pairs obtained in step S3: wherein (x i , y i ) is the coordinate of any feature point in the reference image, (x j , y j ) is the coordinate of the corresponding feature point in the image to be matched, len α is the horizontal dimension of the reference image relative to the image to be matched, and Grad is the two-dimensional linear value corresponding to any matching point pair; The corresponding two-dimensional linear values calculated from all matching point pairs between the reference image and the image to be matched are counted, and the number S of the corresponding two-dimensional linear values in different statistical scale ranges is calculated in combination with statistical scales i ; According to S i <Max(S i )*n eliminates the two-dimensional linear values within the statistical scale range, and the corresponding matching point pairs are eliminated according to the eliminated two-dimensional linear values, where n is an elimination factor. 4.The method of claim 3, wherein, The statistical scale Scale i The calculation is: Wherein, Min(Grad) is the two-dimensional linear minimum value, Max(Grad) is the two-dimensional linear maximum value, k is the total number of statistical scales, and i is the i th statistical scale. 5.The airport runway foreign object identification method based on triangle barycenter feature point matching of claim 1, characterized in that, The image brightness mean value MoI is calculated as follows: wherein, and are gray scales of the image to be recognized, are gray scales of the reference image.

Images