A method, apparatus, equipment and medium for identifying slot areas in aircraft parts

Abstract

This application discloses a method for identifying groove areas in aircraft parts, relating to the field of machining, and aims to solve the technical problem of low efficiency in existing methods for identifying groove areas in aircraft parts. The method includes the following steps: performing a first identification on a three-dimensional model image of the aircraft part to obtain a first identification result, wherein the first identification involves identifying groove areas in the three-dimensional model image of the aircraft part. The first identification includes a first coarse identification based on mathematical morphological pattern difference evaluation, a second coarse identification based on the feature patch attributes of the groove region's neighborhood, and a third coarse identification based on deep neural network learning; obtaining groove area identification result data based on the first identification result; and interactively judging the groove area identification result data with the three-dimensional model image of the aircraft part to obtain the groove areas in the three-dimensional model image of the aircraft part. This application can achieve automated identification of groove areas in aircraft parts, improving the efficiency of groove area identification.

Description

Technical Field

[0001] This application relates to the field of machining, and in particular to a method, apparatus, equipment and medium for identifying groove areas in aerospace parts. Background Technology

[0002] In the aerospace field, the main feature surfaces of common frame and beam structural components include webs, sidewalls, fillets, bottom corners, and rib tops. The groove area composed of feature facets is a higher-order three-dimensional geometric element and one of the smallest machining units in CNC machining. In the existing technology, taking groove machining as an example, toolpath generation software requires manual selection of the driving and guiding surfaces composed of the feature facets corresponding to the groove. Manually selecting the feature facets corresponding to the groove is inefficient and cannot meet the needs of automation, high efficiency, and low cost. Summary of the Invention

[0003] The main objective of this application is to provide a method for identifying slot areas in aircraft parts, aiming to solve the technical problem of low efficiency in existing methods for identifying slot areas in aircraft parts.

[0004] To address the aforementioned technical problems, this application proposes a method for identifying slot areas in aircraft parts, comprising the following steps:

[0005] The first recognition is performed on the three-dimensional model image of the aircraft component to obtain the first recognition result. The first recognition is to identify the groove area of ​​the three-dimensional model image of the aircraft component. The first recognition includes a first coarse recognition based on mathematical morphological pattern difference evaluation, a second coarse recognition based on the feature patch attributes of the groove region feature neighborhood, and a third coarse recognition based on deep neural network learning. The first recognition result includes the corresponding first coarse recognition result, second coarse recognition result, and third coarse recognition result.

[0006] Based on the first identification result, the groove area identification result data is obtained;

[0007] The groove area identification result data is interactively judged with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component.

[0008] As some optional embodiments of this application, the step of performing coarse recognition on the three-dimensional model image of the aerospace component that meets preset conditions to obtain a coarse recognition result includes:

[0009] Sample images are obtained based on the three-dimensional model image of the aircraft component, wherein the sample images include a first sample image, a second sample image, and a third sample image, and the grayscale values ​​of the first sample image, the second sample image, and the third sample image are different.

[0010] The first sample image is subjected to the first coarse recognition based on mathematical morphological pattern difference evaluation to obtain the first coarse recognition result;

[0011] The second sample image is subjected to the second coarse recognition based on the neighborhood feature patch attributes of the slot region to obtain the second coarse recognition result;

[0012] The third sample image is subjected to the third coarse recognition based on deep neural network learning to obtain the third coarse recognition result.

[0013] As some optional embodiments of this application, the step of obtaining sample images based on the three-dimensional model images of the aircraft components includes:

[0014] Each feature facet is assigned a different grayscale value based on its facet number in the 3D model image of the aircraft component.

[0015] The first sample image is obtained by cropping the three-dimensional model image of the aircraft component;

[0016] Set the grayscale values ​​of all feature patches in the three-dimensional model image of the aircraft component to the first preset grayscale value;

[0017] Extract the three-dimensional model image of the aircraft component to obtain the second sample image;

[0018] All feature patches in the three-dimensional model image of the aircraft component are set to a second preset grayscale value, wherein the second preset grayscale value is the same as the grayscale value of the background image in the three-dimensional image of the aircraft component;

[0019] The three-dimensional model image of the aircraft component is extracted to obtain a third sample image.

[0020] As some optional embodiments of this application, the step of performing the first coarse recognition based on mathematical morphological pattern difference evaluation on the first sample image to obtain the first coarse recognition result includes:

[0021] Obtain pixel information from the first sample image;

[0022] The pixel information is filled into a grayscale difference lookup table and a pixel coordinate position lookup table. The grayscale difference lookup table is constructed based on the grayscale value of each feature patch and the number of pixels. The pixel coordinate position lookup table is constructed based on the grayscale value of each feature patch and the position of the pixels.

[0023] Obtain the gradient image corresponding to each feature patch based on the grayscale difference lookup table and the pixel coordinate point position lookup table;

[0024] The area and aspect ratio information of the corresponding feature patch are obtained from each gradient image.

[0025] The type of each feature patch is determined based on the area information and the aspect ratio information of the feature patch, and recorded as the first coarse recognition result.

[0026] As some optional embodiments of this application, the step of obtaining the gradient image based on the grayscale difference lookup table and the pixel coordinate point position lookup table includes:

[0027] Based on the data in each row of the grayscale difference lookup table, obtain n corresponding color images, where n is a positive integer and equal to the number of rows in the grayscale difference lookup table;

[0028] Perform grayscale and binarization processing on n color images to obtain the gradient value of each pixel;

[0029] n gradient images are obtained based on the gradient value of each pixel, wherein the size of the gradient image is the same as the size of the first sample image, the background gray level of the gradient image is 0, and the foreground gray level of the gradient image is the gradient value of the corresponding pixel.

[0030] As some optional embodiments of this application, the step of obtaining the area information and aspect ratio information of the corresponding feature patch based on each gradient image includes:

[0031] Based on the nearest neighbor principle of chessboard distance, the pixels with non-zero gray values ​​in each gradient image are connected end to end to obtain the corresponding closed contour line.

[0032] Project each of the closed contour lines onto the X-axis and Y-axis of a Cartesian coordinate system to obtain the corresponding X-axis projection extrema and Y-axis projection extrema.

[0033] The minimum bounding rectangle is obtained according to each X-axis projection extreme value and Y-axis projection extreme value, wherein the tilt angle of the minimum bounding rectangle is 0.

[0034] The area and aspect ratio information of the corresponding feature facet are obtained based on each minimum bounding rectangle.

[0035] As some optional embodiments of this application, the step of performing the second coarse recognition on the second sample image based on the groove region feature neighborhood feature patch attributes to obtain the second coarse recognition result includes:

[0036] Based on the size of the second sample image, grayscale traversal is performed on each pixel of the second sample image at a preset step.

[0037] The search stops when the gray value of the traversed pixel is not equal to the first preset gray value, until all pixels of the second sample image have been traversed, so as to obtain all target pixels with four-neighbor reachability.

[0038] The corresponding contour image is obtained according to the coordinate information of the target pixel in each row of the first index table. The first index table is constructed according to the coordinate information of the target pixel, and the target pixels located in the same feature patch are stored in the same row of the first index table.

[0039] The deep learning classification network model is trained based on each contour image to obtain weight parameters that meet preset conditions.

[0040] The deep learning classification network model is initialized according to the weight parameters that meet the preset conditions;

[0041] The second sample image is loaded into the deep learning classification network model to obtain a second coarse recognition result.

[0042] As some optional embodiments of this application, the step of obtaining the corresponding contour image based on the pixel data values ​​of each row of the first index table includes:

[0043] The first index table is sorted according to a preset sorting rule;

[0044] The scale value of the feature patch is obtained based on the data in each row of the first index table;

[0045] Construct a feature neighborhood image based on the scale value;

[0046] A second index table is established, wherein the second index table is established based on the first index table and the minimum X-axis coordinate value and minimum Y-axis coordinate value in each row of the first index table;

[0047] Contours are drawn in the feature neighborhood image according to the second index table to obtain a contour line image.

[0048] As some optional embodiments of this application, the step of performing the third coarse recognition based on deep neural network learning on the third sample image to obtain the third coarse recognition result includes:

[0049] The third sample image is annotated according to preset annotation rules to obtain the first annotation file;

[0050] Data augmentation is performed on the third sample image and the annotation file to obtain an augmented sample image and an augmented annotation file;

[0051] The enhanced annotation file is then converted to a new format to obtain the target annotation file.

[0052] The deep detection network is trained with a preset number of training iterations based on the training samples to obtain weight parameter files corresponding to different training iterations. The training samples include augmented sample images and target annotation files.

[0053] Initialize the detection network model according to the weight parameter file;

[0054] The third sample image is input into the detection network model to obtain the third coarse recognition result.

[0055] As some optional embodiments of this application, the step of obtaining the slot area identification result data based on the first identification result includes:

[0056] Construct a joint analysis decision function, wherein the joint analysis decision function is as follows:

[0057]

[0058] In the formula, x1 is the first coarse recognition result, w1 is the weight of the first coarse recognition, x2 is the second coarse recognition result, w1 is the weight of the second coarse recognition, x3 is the first coarse recognition result, w3 is the weight of the third coarse recognition, and the sum of w1, w2 and w3 is less than or equal to 1.

[0059] Based on the joint analysis decision function and the first coarse identification result, the second coarse identification result, and the third coarse identification result, the groove area identification result data is obtained.

[0060] As some optional embodiments of this application, the step of interactively judging the groove area recognition result data with the three-dimensional model image of the aerospace component to obtain the final recognition result includes:

[0061] The corresponding identification slot area is obtained based on the identification result data of each slot area;

[0062] The length of the corresponding slot region is obtained based on each of the identified slot regions and the three-dimensional model image of the aircraft component.

[0063] If the vertical distance between the feature patches in the recognition slot region is less than or equal to the length of the corresponding slot region, then the corresponding slot region recognition result data is recorded as the final recognition result.

[0064] Otherwise, remove the slot area identification result data.

[0065] As some optional embodiments of this application, after the step of interactively judging the groove area recognition result data with the data corresponding to the groove area of ​​the part in the three-dimensional model image to obtain the final recognition result, the method further includes:

[0066] The identified groove area is then back-mapped into the 3D model image of the aircraft component.

[0067] To address the aforementioned technical problems, this application also proposes an aircraft component slot area identification device, characterized in that the device comprises:

[0068] The coarse recognition module is used to perform coarse recognition on the three-dimensional model image of the aircraft parts to obtain coarse recognition results. The coarse recognition includes a first coarse recognition, a second coarse recognition, and a third coarse recognition. The coarse recognition results include the corresponding first coarse recognition results, second coarse recognition results, and third coarse recognition results.

[0069] The refined identification module is used to refine the first coarse identification result, the second coarse identification result, and the third coarse identification result according to the joint analysis decision function, so as to obtain the trench area identification result data;

[0070] The interactive judgment module is used to interactively judge the groove area recognition result data with the data corresponding to the groove area of ​​the part in the three-dimensional model image to obtain the final recognition result.

[0071] To address the aforementioned technical problems, this application also proposes: an aircraft component slot area identification device, comprising at least one processor, at least one memory, and computer program instructions stored in the memory, which implement the method described above when the computer program instructions are executed by the processor.

[0072] To address the aforementioned technical problems, this application also proposes: a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implements the method described above.

[0073] The method for identifying groove areas in aircraft parts described in this application involves performing a first identification on a 3D model image of the aircraft part to obtain a first identification result. This first identification involves identifying groove areas in the 3D model image of the aircraft part. The first identification includes a first coarse identification, a second coarse identification, and a third coarse identification. The first identification result includes the corresponding first coarse identification result, second coarse identification result, and third coarse identification result. Based on the first identification result, groove area identification result data is obtained. The groove area identification result data is then interactively judged with the 3D model image of the aircraft part to obtain the groove areas in the 3D model image of the aircraft part. It can be seen that the solution of this application can automatically identify groove areas in the aforementioned image using the 3D model image of the aircraft part, without manual intervention, reducing labor costs and improving the efficiency of groove area identification. Furthermore, by using a joint analysis decision function to refine the coarse identification result and interactively judging the groove area identification result data with the 3D model image of the aircraft part, the accuracy of groove area identification is ensured. Attached Figure Description

[0074] Figure 1 This is a flowchart illustrating the method for identifying the slot area of ​​aircraft parts as described in the embodiments of this application;

[0075] Figure 2 This is a schematic diagram of the process for obtaining three coarse recognition results as described in the embodiments of this application;

[0076] Figure 3 This is a schematic diagram of the first sample image in an embodiment of this application;

[0077] Figure 4 This is a schematic diagram of the second sample image in the embodiments of this application;

[0078] Figure 5 This is a schematic diagram of the third sample image described in the embodiments of this application;

[0079] Figure 6 This is a schematic diagram of the process of reverse mapping the identified slot area as described in the embodiments of this application;

[0080] Figure 7 This is a schematic diagram of the aviation component slot area identification device described in the embodiments of this application;

[0081] Figure 8 This is a schematic diagram of the aviation component slot area identification device described in the embodiments of this application;

[0082] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0083] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0084] In the aerospace industry, common frame and beam structural components primarily feature surfaces such as webs, sidewalls, fillets, bottom corners, and rib tops. The groove region, composed of these feature surfaces, represents a higher-order three-dimensional geometric element and is one of the smallest machining units in CNC machining. Taking groove machining as an example, existing toolpath generation software requires manual selection of the driving and guiding surfaces formed by the feature surfaces corresponding to the groove. Selecting the associated surface information corresponding to the groove region and designing the corresponding machine tool machining parameters yields the corresponding CNC machining program. Manually selecting the feature surfaces corresponding to the groove requires highly skilled machining program designers, especially for aerospace manufacturing parts, which places even greater demands on the workforce and increases labor costs compared to ordinary production companies.

[0085] Based on this, such as Figure 1 As shown, an embodiment of this application proposes a method for identifying slot areas in aircraft parts, including the following steps:

[0086] S1. Perform a first recognition on the three-dimensional model image of the aircraft component to obtain a first recognition result. The first recognition is to identify the groove area of ​​the three-dimensional model image of the aircraft component. The first recognition includes a first coarse recognition based on mathematical morphological pattern difference evaluation, a second coarse recognition based on the feature neighborhood patch attributes of the groove region, and a third coarse recognition based on deep neural network learning. The first recognition result includes the corresponding first coarse recognition result, second coarse recognition result, and third coarse recognition result.

[0087] As can be seen, the aerospace component groove area recognition method described in this application directly performs a first recognition on the three-dimensional model image of the aerospace component to obtain a first recognition result. The aerospace component can be a beam-type part or a frame-type part in the aircraft structural components; the three-dimensional image of the aerospace component can be an image designed based on the three-dimensional modeling software CATIA; the coarse recognition includes a first coarse recognition, a second coarse recognition, and a third coarse recognition. The recognition strategies of the first coarse recognition, the second coarse recognition, and the third coarse recognition are different. The coarse recognition includes, but is not limited to, a coarse recognition method based on generalized groove layer processing features, a coarse recognition method based on convolutional neural networks, a coarse recognition method based on mathematical morphological pattern difference evaluation, a coarse recognition method based on groove region feature neighborhood feature patch attributes, and a coarse recognition method based on deep neural network learning. The coarse recognition result includes the corresponding first coarse recognition result, second coarse recognition result, and third coarse recognition result. By performing coarse recognition on the three-dimensional model image of the aerospace component using three coarse recognition methods based on different recognition strategies, the recognition defects caused by a single recognition strategy are avoided.

[0088] S2. Based on the first identification result, obtain the groove area identification result data;

[0089] Since the first identification result includes a first coarse identification result, a second coarse identification result, and a third coarse identification result obtained by different identification strategies, it is necessary to analyze the above three results to obtain the groove area identification result data, thereby improving the accuracy of groove area identification for aerospace parts.

[0090] S3. Interact with the groove area identification result data and the three-dimensional model image of the aircraft part to obtain the groove area in the three-dimensional model image of the aircraft part.

[0091] By using the above method, the groove area recognition result data is interactively judged with the three-dimensional model image of the aircraft component. The corresponding adjacency relationship of the identified groove area and its corresponding surface is found in the three-dimensional model image of the aircraft component. Based on the adjacency relationship, the final recognition result can be obtained. This realizes the automatic recognition of the groove area of ​​the aircraft component without manual intervention, reduces labor costs, and improves the efficiency of groove area recognition of the aircraft component.

[0092] In order to accurately identify the features of the groove area in the subsequent identification process, as one of the optional methods of this application, in the CATIA software, the background grayscale of the three-dimensional model image of the aerospace component is set to (255,255,255), the line width is set to 3:0.5mm, the line type is set to 1, and the parameters of scattering, reflection and diffusion are set to 1.

[0093] As some optional embodiments of this application, the step of performing coarse recognition on the three-dimensional model image of the aerospace component that meets preset conditions to obtain a coarse recognition result includes:

[0094] S11. Obtain sample images based on the three-dimensional model image of the aircraft component, wherein the sample images include a first sample image, a second sample image, and a third sample image, and the grayscale values ​​of the first sample image, the second sample image, and the third sample image are different.

[0095] Specifically, the coarse recognition process includes three coarse recognition methods with different recognition strategies. Different recognition strategies have different requirements for the grayscale values ​​of the sample images. Therefore, the obtained sample images include a first sample image, a second sample image, and a third sample image with different grayscale values.

[0096] S12. Perform the first coarse recognition based on the mathematical morphological pattern difference evaluation on the first sample image to obtain the first coarse recognition result;

[0097] S13. Perform the second coarse recognition on the second sample image based on the neighborhood feature patch attributes of the slot region feature to obtain the second coarse recognition result;

[0098] S14. Perform the third coarse recognition based on deep neural network learning on the third sample image to obtain the third coarse recognition result.

[0099] As can be seen, after acquiring the sample images, different coarse recognition strategies are applied to the first sample image, the second sample image, and the third sample image. In this embodiment, the coarse recognition includes a first coarse recognition based on mathematical morphological pattern difference evaluation, a second coarse recognition based on the features of the groove region feature neighborhood patch attributes, and a third coarse recognition based on deep neural network learning, in order to avoid the error of a single coarse recognition strategy.

[0100] As some optional embodiments of this application, the step of obtaining sample images based on the three-dimensional model images of the aircraft components includes:

[0101] S111. Set different grayscale values ​​for each feature facet according to the facet number of the feature facet in the three-dimensional model image of the aircraft component;

[0102] S112. Extract the three-dimensional model image of the aircraft component to obtain the first sample image;

[0103] like Figure 3 The diagram shows a schematic of the first sample image, which is used for the first coarse recognition based on mathematical morphological pattern difference evaluation. First, the patch number of each feature patch in the 3D model image of the aircraft component is obtained using CATIA software. Then, different grayscale values ​​are assigned to each feature patch based on the patch number, ensuring that no patches with the same grayscale exist in the 3D model of the aircraft component. The grayscale value is determined according to the following formula:

[0104]

[0105] In the formula, sum represents the total number of faces contained in the 3D model image of the aerospace component; s n The nth facet in the randomized order of the part represents the facet; S represents all faces in the 3D model image of the aerospace component; h R (s n ), h G (s n ), hB(s n ) represent the s of facet number n respectively. n The gray values ​​of the R, G, and B channels of the feature patch; a, b, and c represent the gray values ​​of the feature patch in the initial iteration of the R, G, and B channels, respectively; Floor represents the floor operation.

[0106] S113. Set the grayscale values ​​of all feature patches in the three-dimensional model image of the aircraft component to the first preset grayscale value;

[0107] S114. Extract the three-dimensional model image of the aircraft component to obtain a second sample image;

[0108] like Figure 4 The image shown is a schematic diagram of the second sample image. The second coarse recognition is based on the neighborhood patch attributes of the groove region features. The neighborhood is determined by combining the relationship between a patch and its neighboring patches when identifying a patch. This feature is reflected in the feature line segments where the patch intersects with the neighboring patches. Among all the patches with the same gray level, the features of different patches differ according to their corresponding contours. Therefore, it is necessary to set the gray level of all feature patches in the 3D model image of the aerospace parts to the same gray level. In a specific embodiment, the first preset gray level is (210, 29, 164).

[0109] S115. Set all feature patches in the three-dimensional model image of the aircraft component to a second preset grayscale value, wherein the second preset grayscale value is the same as the grayscale value of the background image in the three-dimensional image of the aircraft component.

[0110] S116. Extract the three-dimensional model image of the aircraft component to obtain a third sample image.

[0111] like Figure 5 The image shown is a schematic diagram of the third sample image. The third coarse recognition based on deep neural network learning mainly identifies the slot region. The deep neural network learns the grayscale, shape and other information of the feature patches corresponding to the slot region to make a decision on whether it belongs to the slot region. The difference between the background and the foreground is reflected by grayscale and contour lines. Different foreground patches are reflected by contour lines in the image. Therefore, setting the grayscale of all patches to the same grayscale as the background can facilitate the recognition efficiency and accuracy of the third coarse recognition.

[0112] As some optional embodiments of this application, the step of performing the first coarse recognition based on mathematical morphological pattern difference evaluation on the first sample image to obtain the first coarse recognition result includes:

[0113] S121. Obtain pixel information in the first sample image;

[0114] The pixel information includes the pixel coordinates and grayscale information. In one specific embodiment, the pixel information can be obtained through OpenCV.

[0115] S122. Fill the pixel information into the grayscale difference lookup table and the pixel coordinate position lookup table, wherein the grayscale difference lookup table is constructed based on the grayscale value of each feature patch and the number of pixels, and the pixel coordinate position lookup table is constructed based on the grayscale value of each feature patch and the position of the pixels.

[0116] The data in each row of the grayscale difference comparison table represents all pixels with the same grayscale value. The first, second, and third columns of the grayscale difference comparison table represent the grayscale values ​​corresponding to the red R channel, green G channel, and blue B channel of the feature patch, respectively. The other columns represent the number of pixels contained in the feature patch.

[0117] The pixel coordinate point location lookup table stores the coordinates of pixels that satisfy the same grayscale value relationship in each row. The first column is the horizontal coordinate value, the second column is the vertical coordinate value, and the row number corresponds to the column number of the corresponding grayscale difference lookup table.

[0118] S123. Obtain the gradient image corresponding to each feature patch according to the gray level difference comparison table and the pixel coordinate point position comparison table.

[0119] Image gradient calculation measures the rate of change in an image. For the edges of an image, the grayscale values ​​change significantly, resulting in larger gradient values. Grayscale difference lookup tables and pixel coordinate point location lookup tables store the coordinates and grayscale values ​​of each pixel for each feature patch. The first-order gradient of each pixel can be calculated. Using the calculated gradient value of each pixel as the grayscale value of the corresponding point in the gradient image yields the gradient image. The gradient image better displays the contours of the feature patches, facilitating their recognition.

[0120] As some optional embodiments of this application, the step of obtaining the gradient image based on the grayscale difference lookup table and the pixel coordinate point position lookup table includes:

[0121] S1231. Based on the data of each row in the grayscale difference lookup table, obtain n corresponding color images, where n is a positive integer and equal to the number of rows in the grayscale difference lookup table, the background grayscale of the color image is (0,0,0), and the foreground grayscale is equal to the corresponding row in the grayscale difference lookup table;

[0122] S1232. Perform grayscale and binarization processing on n color images to obtain the gradient value of each pixel.

[0123] The generated n color images are converted to grayscale using a loop, and the grayscale conversion is performed according to the following formula:

[0124]

[0125] In the formula, hn,new(x,y) represents the gray value of the single-channel pixel at the corresponding pixel at (x,y) after processing;

[0126] After grayscale processing, the image is binarized according to the following formula:

[0127] Then h n,new (x,y)=255else then h n,new (x,y)=0

[0128] Calculate the first-order gradient of all pixels in the image. The magnitude of the gradient is approximated by the square root of the sum of the squares of the gradients of the pixels in the X and Y directions. The gradient of a pixel in the X-axis direction can be expressed as:

[0129]

[0130] Using the neighborhood difference form, it can be further expressed as:

[0131]

[0132] Similarly, the gradient of a pixel along the Y-axis can be expressed as:

[0133]

[0134] In summary, the gradient value at any point in a color image can be expressed as:

[0135]

[0136] S1233. Obtain n gradient images based on the gradient value of each pixel, wherein the size of the gradient image is the same as the size of the first sample image, the background gray level of the gradient image is 0, and the foreground gray level of the gradient image is the gradient value of the corresponding pixel.

[0137] S124. Obtain the area information and aspect ratio information of the corresponding feature patch based on each gradient image;

[0138] As some optional embodiments of this application, the step of obtaining the area information and aspect ratio information of the corresponding feature patch based on each gradient image includes:

[0139] S1241. Based on the nearest neighbor principle of chessboard distance, connect the pixels with non-zero gray values ​​in each gradient image to obtain the corresponding closed contour line.

[0140] The chessboard distance between any two points Point1(x1,y1) and Point2(x2,y2) in the gradient image is represented as follows:

[0141] dist8(Point1,Point2)=max(|x1-x2|,|y1-y2|)

[0142] Here, 8 represents the 8-neighborhood relationship of a pixel in the image.

[0143] S1242. Project each of the closed contour lines onto the X-axis and Y-axis of a Cartesian coordinate system to obtain the corresponding X-axis projection extrema and Y-axis projection extrema.

[0144] Establish a Cartesian coordinate system with its origin located at the bottom left corner, opposite to the origin at the top left corner of the gradient image. The scale unit of the coordinate system along the X and Y axes is 1 pixel, corresponding to the pixel measurement method.

[0145] S1243. Obtain the corresponding minimum bounding rectangle based on each X-axis projection extreme value and Y-axis projection extreme value, wherein the tilt angle of the minimum bounding rectangle is 0.

[0146] Gradient images can clearly display the contour of each feature patch. Based on the contour, the area and aspect ratio information of the feature patch can be obtained. Projecting the contour lines along the X and Y axes, a projected straight line segment l can be obtained on the X-axis. X A projected straight line segment l can be obtained on the Y-axis. Y l X It consists of a series of x values, with the maximum value being x. max The corresponding minimum value is x min l Y It consists of a series of y values, with the maximum value being y. max The corresponding minimum value is y min Based on the maximum and minimum values ​​of the projections of the contour lines onto the coordinate axes, construct the minimum bounding rectangle rect with a zero tilt angle for the contour line region. The area of ​​the rectangle is Size. rect (x max x min y max y min )=(x max -x min )*(y max -y min The corresponding aspect ratio is 'ratio'. rect (length, width) = length / width, where length represents the longer side in the rectangle, and width represents the shorter side. The longer and shorter sides are calculated by directly comparing the two values; the larger value is length, and the smaller value is width.

[0147] S1244. Obtain the area information and aspect ratio information of the corresponding feature patch based on each of the minimum enclosing rectangles.

[0148] S125. Determine the type of each feature patch based on the area information and the aspect ratio information of the feature patch, and record it as the first coarse recognition result.

[0149] Specifically, since the image corresponding to the difference evaluation and recognition based on the mathematical morphological model is obtained based on the method of normal vector along the web, it has a large area projection within the visible range. Based on this, the design is based on the area Size of the patch. rect aspect ratio of the patch rect The identification judgment formula (f1) is composed as follows:

[0150] satisfy

[0151] Where k represents the total number of faces in the image; F decision (numslices ) Judgment function; S threshold The set threshold for the area of ​​the web plate; ratio threshold The set aspect ratio threshold for the web plate; S threshold1 The set threshold for the area of ​​the outer sheet excluding the web; ratio threshold1 The set threshold for the aspect ratio of the outer surface of the plate, excluding the web.

[0152] As some optional embodiments of this application, the step of performing the second coarse recognition on the second sample image based on the groove region feature neighborhood feature patch attributes to obtain the second coarse recognition result includes:

[0153] S131. Based on the size of the second sample image, perform grayscale traversal on each pixel of the second sample image with a preset step.

[0154] S132. When the gray value of the traversed pixel is not equal to the first preset gray value, stop the search until all pixels of the second sample image have been traversed, so as to obtain all target pixels with four-neighbor reachability.

[0155] First, based on the size of the second sample image, all pixels within the specified size are used as seed points, and grayscale traversal is performed on the pixels in the four neighboring regions (up, down, left, and right), as shown in Figure XX. The corresponding step size is 1. When the traversed seed point moves to the boundary of the patch, i.e., the grayscale of the corresponding pixel is no longer the first preset grayscale, the search stops until all points are searched to obtain all target pixels with four neighboring reachability, thus obtaining the first index table.

[0156] S133. Obtain the corresponding contour image according to the coordinate information of the target pixel in each row of the first index table, wherein the first index table is constructed according to the coordinate information of the target pixel, and the target pixels located in the same feature patch are stored in the same row of the first index table.

[0157] As some optional embodiments of this application, the step of obtaining the corresponding contour image based on the pixel data values ​​of each row of the first index table includes:

[0158] S1331. Sort the first index table according to the preset sorting rules;

[0159] The preset sorting rule can be based on ascending or descending order of the number of pixels in the same feature patch. In a specific embodiment, the preset sorting rule is based on descending order of the number of pixels in the same feature patch, that is, the feature patch corresponding to the first row of the first index table after processing has the most pixels.

[0160] S1332. Obtain the scale value of the feature patch based on the data of each row of the first index table;

[0161] Data from each row of the first index table is extracted sequentially. Based on the data value of each pixel in a row, the scale value of the four-neighbor independent patches is calculated. This scale value is used as the basis for the size of the newly created feature neighborhood image. The scale value is calculated according to the following formula:

[0162]

[0163]

[0164]

[0165]

[0166] Where slices represent independent image patches within their four-neighborhood in the image; table pixel (num pixel ) represents an index table containing information on all feature patches; x max (num pixel () represents the maximum X-coordinate of all pixels in a given feature patch; x min (num pixel ) represents the minimum X-coordinate of all pixels in a given feature patch; y max (num pixel ) represents the maximum Y-coordinate of all pixels in a given feature patch; y min (num pixel ) represents the minimum Y-coordinate of all pixels in a given feature patch; new_img nea (num pixel ) represents a newly created feature neighborhood image based on the calculated values ​​corresponding to cols and rows.

[0167] S1333. Construct a feature neighborhood image based on the scale value;

[0168] Based on the scale image, a feature neighborhood image is constructed, wherein the size of the feature neighborhood image is determined according to the corresponding scale value, and the background grayscale value of the feature neighborhood image is 0.

[0169] S1334. Establish a second index table, wherein the second index table is established based on the first index table and the minimum X-axis coordinate value and minimum Y-axis coordinate value in each row of the first index table;

[0170] In the second index table, all values ​​remain unchanged except for the X and Y values. The rule for establishing the second index table is to subtract the minimum X-axis coordinate value and the minimum Y-axis coordinate value of the corresponding row from the X and Y values ​​of each row in the first index table.

[0171] S1335. Draw a contour in the feature neighborhood image according to the second index table to obtain a contour line image;

[0172] Based on the data in the second index table, contours are drawn for the contour regions of independent patches in the newly created feature neighborhood image. The contour line width is 2, and the corresponding gray value is 255. The background gray value of the feature neighborhood image is 0, and the gray value of the contour line is 255, which makes the training of the subsequent deep learning classification network model more effective.

[0173] S134. Train the deep learning classification network model based on each contour image to obtain weight parameters that meet preset conditions.

[0174] Based on the obtained contour images, deep learning training samples are constructed. The deep learning samples include web contour images, sidewall contour images, and other contour images (rounded corners, bottom corners, rib tops, etc.). The deep learning classification network model is trained using the deep learning samples, and the weight parameters corresponding to the highest accuracy in the sample validation set are saved to obtain weight parameters that meet the preset conditions.

[0175] S135. Initialize the deep learning classification network model according to the weight parameters that meet the preset conditions;

[0176] S136. Load the second sample image into the deep learning classification network model to obtain the second coarse recognition result.

[0177] After obtaining the weight parameters that meet the preset conditions, the deep learning classification network model is initialized according to the weight parameters. Then, the second sample image is loaded into the deep learning classification network model to obtain the second coarse recognition result.

[0178] As some optional embodiments of this application, the step of performing the third coarse recognition based on deep neural network learning on the third sample image to obtain the third coarse recognition result includes:

[0179] S141. The third sample image is annotated according to the preset annotation rules to obtain the first annotation file;

[0180] When annotating the third sample image, the preset annotation rules include using a rectangle to annotate only complete groove regions in the third sample image; the tilt angle of the rectangle is 0 and it exactly selects the groove region in the image; incomplete, non-closed groove regions in the image are not annotated.

[0181] S142. Perform data augmentation on the third sample image and the annotation file to obtain an enhanced sample image and an enhanced annotation file;

[0182] Data augmentation methods include, but are not limited to, horizontal flipping with a probability of 1.0; vertical flipping with a probability of 1.0; image rotation; affine transformation; and overall image scaling. One or two of the above methods are randomly selected to augment the third sample image and the annotation file to obtain an augmented sample image and an augmented annotation file.

[0183] S143. Convert the format of the enhanced annotation file to obtain the target annotation file;

[0184] The augmented annotation file is in XML format. It can be further converted into COCO data format to obtain the training samples for the deep detection network. The training samples include target annotation files and augmented sample images.

[0185] S144. Train the deep detection network with a preset number of training times based on the training samples to obtain weight parameter files corresponding to different training times, wherein the training samples include augmented sample images and target annotation files.

[0186] S145. Initialize the detection network model according to the weight parameter file;

[0187] S146. Input the third sample image into the detection network model to obtain the third coarse recognition result.

[0188] The deep detection network is trained based on training samples and a preset number of training iterations. One training iteration means that the training samples are completely input into the deep detection network for training. Multiple training iterations reduce randomness and avoid local extrema. In a specific embodiment, the preset number of training iterations is 28. After completing 28 training iterations, 28 weight parameter files are obtained. The weight parameter file corresponding to the training iteration with the smallest loss is selected to initialize the detection network model. The third sample image is then input into the detection network to obtain the third coarse recognition result.

[0189] As some optional embodiments of this application, the step of refining the three coarse identification results according to the joint analysis decision function to obtain the trench area identification result data includes:

[0190] S21. Construct a joint analysis decision function, wherein the joint analysis decision function is as follows:

[0191]

[0192] In the formula, x1 is the first coarse recognition result, w1 is the weight of the first coarse recognition, x2 is the second coarse recognition result, w1 is the weight of the second coarse recognition, x3 is the first coarse recognition result, w3 is the weight of the third coarse recognition, and the sum of w1, w2 and w3 is less than or equal to 1.

[0193] S22. Based on the joint analysis decision function, the first coarse identification result, the second coarse identification result, and the third coarse identification result are refined to obtain the trench area identification result data.

[0194] As some optional embodiments of this application, the step of interactively judging the groove area recognition result data with the three-dimensional model image of the aerospace component to obtain the final recognition result includes:

[0195] S31. Obtain the corresponding identification slot area based on the identification result data of each slot area;

[0196] S32. Obtain the corresponding groove length based on each of the identified groove regions and the three-dimensional model image of the aircraft component;

[0197] S33. If the vertical distance between the feature patches in the recognition slot region is less than or equal to the length of the corresponding slot region, then the corresponding slot region recognition result data is recorded as the final recognition result.

[0198] S34. Otherwise, remove the slot area identification result data.

[0199] The patches belong to the same slot. The corresponding patches of the identified slot area are located in the original 3D model. If the identified patches have a minimum spatial and geometric distance relationship (i.e., the patches are adjacent in the geometric entity or the perpendicular distance between the patches is less than the length of the slot), the identified patch results are highly reliable and should be retained. If the minimum spatial and geometric distance between the identified patches is greater than the length of the corresponding slot, it indicates that the patches corresponding to the slot area have low feasibility and should be removed. The evaluation should then be based on other, more reliable viewpoint image recognition results.

[0200] As some optional embodiments of this application, after the step of interactively judging the groove area recognition result data with the data corresponding to the groove area of ​​the part in the three-dimensional model image to obtain the final recognition result, the method further includes:

[0201] The identified groove area is then back-mapped into the 3D model image of the aircraft component.

[0202] The identified groove area is mapped to the original two-dimensional image, which contains grayscale and positional information of each region of the part. Mapping establishes a correspondence and relationship between feature faces within the groove area; that is, in the original two-dimensional image, one and only one face can be found that satisfies the correspondence.

[0203] like Figure 6 The diagram illustrates the process of back-mapping the recognition results to a 3D image. This involves mapping the results from a 2D image to a 3D image, specifically by attaching the groove region features from the 2D image to the 3D image based on the viewpoint and magnification at the time of capture. Each patch and groove region is unique. Only non-occluded areas satisfy the back-mapping characteristics during the attachment process. Displaying the groove region recognition results obtained after both coarse and fine recognition processes on the 3D model provides a clear visual representation of the recognition effectiveness.

[0204] Please see Figure 7 This invention provides an aircraft component slot area identification device, the device comprising:

[0205] The coarse recognition module is used to perform coarse recognition on the three-dimensional model image of the aircraft parts to obtain coarse recognition results. The coarse recognition includes a first coarse recognition, a second coarse recognition, and a third coarse recognition. The coarse recognition results include the corresponding first coarse recognition results, second coarse recognition results, and third coarse recognition results.

[0206] The refined identification module is used to refine the first coarse identification result, the second coarse identification result, and the third coarse identification result according to the joint analysis decision function, so as to obtain the trench area identification result data;

[0207] The interactive judgment module is used to interactively judge the groove area recognition result data with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component.

[0208] In addition, combined Figure 1 The coordinate compensation method based on central projection described in this embodiment of the invention can be implemented by a coordinate compensation device based on central projection. Figure 8 A schematic diagram of the hardware structure of the coordinate compensation device based on center projection provided in an embodiment of the present invention is shown.

[0209] The coordinate compensation device based on central projection may include at least one processor 301, at least one memory 302, and computer program instructions stored in the shown memory area 302, which implement the method described in the above embodiments when the computer program instructions are executed by the processor 301.

[0210] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0211] Memory 302 may include mass storage for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to a data processing device. In a particular embodiment, memory 302 is a non-volatile solid-state memory. In a particular embodiment, memory 302 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0212] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement any of the coordinate compensation methods based on central projection in the above embodiments.

[0213] In one example, the coordinate compensation device based on central projection may also include a communication interface and a bus. For example, Figure 8 As shown, the processor, memory, and communication interface are connected via a bus to communicate with each other. The communication interface is mainly used to enable communication between the various modules, devices, units, and / or equipment in this embodiment of the invention.

[0214] A bus, including hardware, software, or both, couples components of a center-projection-based coordinate compensation device together. For example, and not limitingly, a bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, a bus may include one or more buses. While specific buses are described and illustrated in embodiments of the invention, the invention contemplates any suitable bus or interconnect.

[0215] Furthermore, in conjunction with the coordinate compensation method based on central projection in the above embodiments, this invention can be implemented using a computer-readable storage medium. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the coordinate compensation methods based on central projection in the above embodiments.

[0216] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0217] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the required tasks. The programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0218] It should also be noted that the exemplary embodiments mentioned in this invention describe methods or systems based on a series of steps or apparatus. However, this invention is not limited to the order of the steps described above; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0219] The above description is merely a specific embodiment of the present invention. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the protection scope of the present invention.

Claims

1. A method for identifying slot areas in aircraft parts, characterized in that, Includes the following steps: The first recognition is performed on the three-dimensional model image of the aircraft component to obtain the first recognition result. The first recognition is to identify the groove area of ​​the three-dimensional model image of the aircraft component. The first recognition includes a first coarse recognition based on mathematical morphological pattern difference evaluation, a second coarse recognition based on the feature patch attributes of the groove region feature neighborhood, and a third coarse recognition based on deep neural network learning. The first recognition result includes the corresponding first coarse recognition result, second coarse recognition result, and third coarse recognition result. Based on the first identification result, the groove area identification result data is obtained; The groove area identification result data is interactively judged with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component; The step of performing the first recognition on the 3D model image of the aircraft to obtain the first recognition result includes: Sample images are obtained based on the three-dimensional model image of the aircraft component, wherein the sample images include a first sample image, a second sample image, and a third sample image, and the grayscale values ​​of the first sample image, the second sample image, and the third sample image are different. The first sample image is subjected to the first coarse recognition based on mathematical morphological pattern difference evaluation to obtain the first coarse recognition result; The second sample image is subjected to the second coarse recognition based on the neighborhood feature patch attributes of the slot region to obtain the second coarse recognition result; The third sample image is subjected to the third coarse recognition based on deep neural network learning to obtain the third coarse recognition result; The step of obtaining the sample image based on the three-dimensional model image of the aircraft component includes: Each feature facet is assigned a different grayscale value based on its facet number in the 3D model image of the aircraft component. The first sample image is obtained by cropping the three-dimensional model image of the aircraft component; Set the grayscale values ​​of all feature patches in the three-dimensional model image of the aircraft component to the first preset grayscale value; Extract the three-dimensional model image of the aircraft component to obtain the second sample image; All feature patches in the three-dimensional model image of the aircraft component are set to a second preset grayscale value, wherein the second preset grayscale value is the same as the grayscale value of the background image in the three-dimensional image of the aircraft component; Extract the three-dimensional model image of the aircraft component to obtain a third sample image; The step of performing the first coarse recognition based on mathematical morphological pattern difference evaluation on the first sample image to obtain the first coarse recognition result includes: Obtain pixel information from the first sample image; The pixel information is filled into a grayscale difference lookup table and a pixel coordinate position lookup table. The grayscale difference lookup table is constructed based on the grayscale value of each feature patch and the number of pixels. The pixel coordinate position lookup table is constructed based on the grayscale value of each feature patch and the position of the pixels. Obtain the gradient image corresponding to each feature patch based on the grayscale difference lookup table and the pixel coordinate point position lookup table; The area and aspect ratio information of the corresponding feature patch are obtained from each gradient image. The type of each feature patch is determined based on the area information and the aspect ratio information of the feature patch, and recorded as the first coarse recognition result. The step of performing the second coarse recognition on the second sample image based on the groove region feature neighborhood feature patch attributes to obtain the second coarse recognition result includes: Based on the size of the second sample image, grayscale traversal is performed on each pixel of the second sample image at a preset step. The search stops when the gray value of the traversed pixel is not equal to the first preset gray value, until all pixels of the second sample image have been traversed, so as to obtain all target pixels with four-neighbor reachability. The corresponding contour image is obtained according to the coordinate information of the target pixel in each row of the first index table. The first index table is constructed according to the coordinate information of the target pixel, and the target pixels located in the same feature patch are stored in the same row of the first index table. The deep learning classification network model is trained based on each contour image to obtain weight parameters that meet preset conditions. The deep learning classification network model is initialized according to the weight parameters that meet the preset conditions; The second sample image is loaded into the deep learning classification network model to obtain a second coarse recognition result.

2. The method for identifying slot areas in aircraft parts according to claim 1, characterized in that, The step of obtaining the gradient image based on the grayscale difference lookup table and the pixel coordinate point position lookup table includes: Based on the data in each row of the grayscale difference lookup table, obtain n corresponding color images, where n is a positive integer and equal to the number of rows in the grayscale difference lookup table; Perform grayscale and binarization processing on n color images to obtain the gradient value of each pixel; n gradient images are obtained based on the gradient value of each pixel, wherein the size of the gradient image is the same as the size of the first sample image, the background gray level of the gradient image is 0, and the foreground gray level of the gradient image is the gradient value of the corresponding pixel.

3. The method for identifying slot areas in aircraft parts according to claim 2, characterized in that, The step of obtaining the area information and aspect ratio information of the corresponding feature patch based on each gradient image includes: Based on the nearest neighbor principle of chessboard distance, the pixels with non-zero gray values ​​in each gradient image are connected end to end to obtain the corresponding closed contour line. Project each of the closed contour lines onto the X-axis and Y-axis of a Cartesian coordinate system to obtain the corresponding X-axis projection extrema and Y-axis projection extrema. The minimum bounding rectangle is obtained according to each X-axis projection extreme value and Y-axis projection extreme value, wherein the tilt angle of the minimum bounding rectangle is 0. The area and aspect ratio information of the corresponding feature facet are obtained based on each minimum bounding rectangle.

4. The method for identifying slot areas in aircraft parts according to claim 1, characterized in that, The step of obtaining the corresponding contour image based on the pixel data values ​​of each row of the first index table includes: The first index table is sorted according to a preset sorting rule; The scale value of the feature patch is obtained based on the data in each row of the first index table; Construct a feature neighborhood image based on the stated scale value; A second index table is established, wherein the second index table is established based on the first index table and the minimum X-axis coordinate value and minimum Y-axis coordinate value in each row of the first index table; Contours are drawn in the feature neighborhood image according to the second index table to obtain a contour line image.

5. The method for identifying the slot area of ​​aircraft parts according to claim 1, characterized in that, The step of performing the third coarse recognition based on deep neural network learning on the third sample image to obtain the third coarse recognition result includes: The third sample image is annotated according to preset annotation rules to obtain the first annotation file; Data augmentation is performed on the third sample image and the annotation file to obtain an augmented sample image and an augmented annotation file; The enhanced annotation file is then converted to a new format to obtain the target annotation file. The deep detection network is trained with a preset number of training iterations based on the training samples to obtain weight parameter files corresponding to different training iterations. The training samples include augmented sample images and target annotation files. Initialize the detection network model according to the weight parameter file; The third sample image is input into the detection network model to obtain the third coarse recognition result.

6. The method for identifying the slot area of ​​aircraft parts according to any one of claims 1-5, characterized in that, The step of obtaining the slot area identification result data based on the first identification result includes: Construct a joint analysis decision function, wherein the joint analysis decision function is as follows: In the formula, x1 is the first coarse recognition result, w1 is the weight of the first coarse recognition, x2 is the second coarse recognition result, w1 is the weight of the second coarse recognition, x3 is the first coarse recognition result, and w3 is the weight of the third coarse recognition. Based on the joint analysis decision function and the first coarse identification result, the second coarse identification result, and the third coarse identification result, the groove area identification result data is obtained.

7. The method for identifying slot areas in aircraft parts according to claim 1, characterized in that, The step of interactively judging the groove area recognition result data with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component includes: The corresponding identification slot area is obtained based on the identification result data of each slot area; The length of the corresponding slot region is obtained based on each of the identified slot regions and the three-dimensional model image of the aircraft component. If the vertical distance between the feature patches in the recognition slot region is less than or equal to the length of the corresponding slot region, the recognition slot region corresponding to the recognition result data of the slot region is retained. Otherwise, remove the slot area identification result data.

8. The method for identifying slot areas in aircraft parts according to claim 1, characterized in that, After the step of interactively judging the groove area identification result data with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component, the method further includes: The identified groove area is then back-mapped into the 3D model image of the aircraft component.

9. An aircraft component slot area identification device, used to implement the aircraft component slot area identification method according to any one of claims 1-8, characterized in that, The device includes: The first recognition module is used to perform a first recognition on the three-dimensional model image of the aircraft part to obtain a first recognition result. The first recognition is to recognize the groove area of ​​the three-dimensional model image of the aircraft part. The first recognition includes a first coarse recognition, a second coarse recognition, and a third coarse recognition. The first recognition result includes the corresponding first coarse recognition result, second coarse recognition result, and third coarse recognition result. The result data acquisition module is used to obtain the groove area identification result data based on the first identification result; The interactive judgment module is used to interactively judge the groove area recognition result data with the three-dimensional model image of the aircraft component to obtain the groove area in the three-dimensional model image of the aircraft component.

10. An aircraft component slot area identification device, characterized in that, include: At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in any one of claims 1-8.

11. A storage medium storing computer program instructions thereon, characterized in that, The method as described in any one of claims 1-8 is implemented when the computer program instructions are executed by the processor.

Images