Infrared thermographic inspection system and method for spacecraft composite material structure products

Abstract

This invention discloses an infrared thermal imaging detection system and method for spacecraft composite material structure products. The system includes: a supporting gantry assembly fixedly mounted on the upper side of an air-float base, forming a rectangular structure; a motion guide rail fixed to the inner side of the gantry assembly; the spacecraft composite material structure product to be tested positioned at the center of the rectangular structure on the upper side of the air-float base, with a belt assembly positioned between the supporting gantry assembly to drive the movement of the spacecraft composite material structure under test; an imaging assembly, a heating assembly, and a laser assembly positioned on the air-float base at their contact points with the supporting gantry assembly; the heating assembly applies a temperature increase during detection; the imaging assembly acquires an infrared thermal image of the adhesive bonding of the spacecraft composite material structure product under test; an electronically controlled host computer processes the infrared thermal image using a pre-trained model to obtain defect locations; and the laser assembly marks the defect locations of the spacecraft composite material structure. This invention can improve detection efficiency.

Description

Technical Field

[0001] This invention relates to the field of aerospace technology, and in particular to an infrared thermal imaging detection system and method for spacecraft composite material structure products. Background Technology

[0002] During the manufacturing, processing, and use of spacecraft composite material structures, various bonding defects, such as delamination and porosity, are prone to occur due to factors such as process technology, material properties, and service conditions. These defects affect the performance of the composite materials, making non-destructive testing and reliability evaluation of the bonding quality of spacecraft composite structures particularly important. Unlike traditional defect detection methods, infrared thermography-based defect detection offers advantages such as non-contact operation, no introduction of additional contaminants, and intuitive detection results. Infrared thermography is more suitable for detecting and evaluating bonding defects during the development of spacecraft composite material structures.

[0003] Composite materials, being multi-layered structures, exhibit complex interfacial heat conduction processes and high surface heat distribution noise. The resulting raw infrared thermograms suffer from low contrast and a lack of depth, making it difficult to distinguish between defect feature data and thermal noise information. Furthermore, existing infrared thermography-based defect detection methods and systems rely on manual judgment to select detection areas. While ensuring thorough defect identification, these methods struggle to balance timeliness and efficiency, particularly for large-scale inspections. These shortcomings, to some extent, limit the application of infrared thermography technology to the detection of adhesive bonding defects in aerospace composite material structures. Summary of the Invention

[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide an infrared thermal imaging detection system and method for spacecraft composite material structure products.

[0005] The technical solution of this invention is:

[0006] In a first aspect, embodiments of the present invention provide an infrared thermal imaging detection system for spacecraft composite material structure products. The system includes: an air-floating base, a supporting gantry assembly, a motion guide rail, a conveyor belt assembly, a heating assembly, an imaging assembly, a laser assembly, an electronically controlled host computer, and the composite material structure of the spacecraft under test.

[0007] The supporting gantry frame assembly is fixedly installed on the upper side of the air-float base, forming a rectangular structure.

[0008] The motion guide rail is fixed to the inside of the gantry assembly so that the composite material structure of the spacecraft under test can move within the motion guide rail;

[0009] The composite material structure product of the spacecraft under test is located at the center of the rectangular structure on the upper side of the air-bearing base, and the conveyor belt assembly is located between the supporting gantry assembly to drive the composite material structure of the spacecraft under test to move.

[0010] The imaging component, the heating component, and the laser component are disposed on the air-floating base at a position where they connect with the supporting gantry assembly;

[0011] The heating component is used to heat and load the adhesive surface of the composite material structure product of the spacecraft under test during testing.

[0012] The imaging component acquires an infrared thermal image of the adhesive bonding surface of the composite material structure product under test based on the temperature change of the bonding surface.

[0013] The electronically controlled host computer calls a pre-trained model to process the infrared thermal image in order to detect the location of defects in the composite material structure product of the spacecraft under test.

[0014] The laser component is used to mark the location of defects in the composite material structure product of the spacecraft under test.

[0015] Optionally, the motion guide rail includes: a first motion guide rail, a second motion guide rail, and a third motion guide rail, wherein the guide rail directions of the first motion guide rail, the second motion guide rail, and the third motion guide rail are perpendicular to each other.

[0016] Secondly, embodiments of the present invention provide an infrared thermal imaging detection method for spacecraft composite material structure products, applied to the detection system described in any of the above claims, the method comprising:

[0017] After entering the working state, the system components are preheated.

[0018] A panoramic scan of the composite material structure product of the spacecraft under test is performed to acquire infrared thermal images of the composite material structure product of the spacecraft under test, and the infrared thermal images are fitted and reconstructed.

[0019] The image information in the fitted and reconstructed infrared thermal image is processed and reconstructed using optimal fusion weights, and the corresponding binary image is obtained.

[0020] Edge detection is performed on the binary image to determine the defect feature data of the composite material structure product of the spacecraft under test;

[0021] The defect feature data is input into a convolutional neural network for detection to obtain the bonding defects of the composite material structure product of the spacecraft under test.

[0022] Optionally, the step of performing edge detection on the binary image to determine the defect feature data of the composite material structure product of the spacecraft under test includes:

[0023] Edge detection is performed on the reconstructed binary image based on the Canny detection algorithm to obtain the edge contour as the boundary line of different defect distribution areas;

[0024] Based on the fact that different defects appear as circles or ellipses in infrared thermal images, the area enclosed by the boundary line is marked.

[0025] By statistically analyzing the pixels enclosed by the boundary line region, the initial defect location area of ​​the composite material structure product of the spacecraft under test is obtained.

[0026] Based on the initial defect location area, the defect feature data is determined.

[0027] Optionally, the step of inputting the defect feature data into a convolutional neural network for detection to obtain the bonding defects of the composite material structure product of the spacecraft under test includes:

[0028] The defect feature data are quantitatively calculated to obtain the defect edge region of the composite material structure product of the spacecraft under test;

[0029] The pre-trained model is called to process the defect edge region to obtain the bonding defects of the composite material structure product of the spacecraft under test.

[0030] The advantages of this invention compared to existing technologies are as follows: This invention improves defect detectability by combining enhanced defect feature data with a convolutional neural network model, achieving automated defect location and discrimination. This reduces the risk of false positives and false negatives associated with human judgment, improving product quality consistency and detection efficiency. During image enhancement, effective fusion after optimized threshold segmentation is employed. Precise qualitative calculation of defect location and region is achieved through pixel extreme value distribution, enabling high-precision extraction of defect feature data. Combined with the laser component in the detection system, high-precision defect location marking can be achieved. Furthermore, this invention designs a high-precision, high-reliability, multi-angle supported gantry-type defect detection system. This system allows for highly stable, multi-degree-of-freedom adjustment of the infrared thermal imaging acquisition equipment. Combined with the transmission component design, it can meet the needs of batch-based, assembly-line defect detection for composite material structures of spacecraft of different sizes. Attached Figure Description

[0031] Figure 1 A schematic diagram of an infrared thermal imaging detection system for a spacecraft composite material structure product provided in an embodiment of the present invention;

[0032] Figure 2A flowchart illustrating the steps of an infrared thermal imaging detection method for a spacecraft composite material structure product provided in an embodiment of the present invention. Detailed Implementation

[0033] Example 1

[0034] Reference Figure 1 The diagram illustrates the structure of an infrared thermal imaging detection system for a spacecraft composite material structure product according to an embodiment of the present invention. Figure 1 As shown, the system may include: an air-floating base 1, a supporting gantry assembly, motion guide rails (3, 4 and 5), a conveyor belt assembly 6, a heating assembly 7, an imaging assembly 8, a laser assembly 9, an electronically controlled host computer 10, and a composite material structure of the spacecraft under test 11.

[0035] The supporting gantry frame assembly is fixedly installed on the upper side of the air-float base, forming a rectangular structure.

[0036] The motion guide rail is fixed to the inside of the gantry assembly so that the composite material structure of the spacecraft under test can move within the motion guide rail;

[0037] The composite material structure product of the spacecraft under test is located at the center of the rectangular structure on the upper side of the air-bearing base, and the conveyor belt assembly is located between the supporting gantry assembly to drive the composite material structure of the spacecraft under test to move.

[0038] The imaging component, the heating component, and the laser component are disposed on the air-floating base at a position where they connect with the supporting gantry assembly;

[0039] The heating component is used to heat and load the adhesive surface of the composite material structure product of the spacecraft under test during testing.

[0040] The imaging component acquires an infrared thermal image of the adhesive bonding surface of the composite material structure product under test based on the temperature change of the bonding surface.

[0041] The electronically controlled host computer calls a pre-trained model to process the infrared thermal image in order to detect the location of defects in the composite material structure product of the spacecraft under test.

[0042] The laser component is used to mark the location of defects in the composite material structure product of the spacecraft under test.

[0043] In one specific implementation of this invention, the motion guide rail includes: a first motion guide rail, a second motion guide rail, and a third motion guide rail, wherein the first motion guide rail, the second motion guide rail, and the third motion guide rail are perpendicular to each other. Figure 1As shown, there are high-strength front-to-back motion guide rails 3, high-strength left-to-right motion guide rails 4, and high-strength up-to-down motion guide rails 5.

[0044] In using such Figure 1 The system shown in the image performs automated batch testing of adhesive bonding in spacecraft composite structures. The specific implementation process is as follows:

[0045] 1. System integration and debugging: Confirm the mechanical motion module and infrared thermal imaging module are functioning normally, and preheat all system components for 10 minutes. First, use a comparison specimen to adjust the layout of the detection parameters and acquisition components, solidifying the detection process and parameters to meet the testing requirements. Install the composite material structure bonding of the spacecraft under test. Based on the size and shape of the bonding, the electrical module controls the host computer to adjust the position of the infrared thermal imaging module to ensure reliable and clear acquisition of the surface condition of the bonding. Automatic zoning is performed according to the inspectable area, and the electrical module controls the configuration and planning of the mechanical motion module's route and speed.

[0046] II. Infrared thermal imaging data of spacecraft composite material structures was acquired using an infrared thermal imaging module. During the acquisition process, the infrared thermal imaging module applied pulsed heat flow to the bonded surface of the spacecraft composite material structure. After the pulsed heat flow, the bonded surface of the spacecraft composite material structure experienced two stages: heating and cooling. By performing polynomial fitting on the temperature change and thermal imaging data, the original thermal image sequence was transformed into a reconstructed thermal image sequence. The reconstruction calculation for the cooling stage is shown in the following formula, where u n Represents the temperature fitting parameters, a n This is the raw thermal image data, where t represents the stage temperature number, and τ... p This indicates the starting temperature for cooling.

[0047]

[0048] Using the above reconstruction calculations, high-frequency noise in infrared thermal image data was filtered out, further improving the signal-to-noise ratio. Subsequent calculations will be performed using the reconstructed infrared thermal image data for further analysis.

[0049] Third, wavelet transform is first performed on the infrared thermal image data. A low-pass filter is constructed using a scaling function to extract low-frequency information from the image, and a high-pass filter is constructed using a wavelet function to extract high-frequency information from the image. The formula for the above information extraction is shown below, where H represents the low-pass filter coefficient matrix, G represents the high-pass filter coefficient matrix, A represents the low-frequency information of the image, and D represents the high-frequency information of the image. As shown in the calculation formula, in the multiple image information extractions, each extraction is based on the previous low-frequency information, and the previous low-frequency information is further downsampled and decomposed into low-frequency information A. l+1 With horizontal vertical diagonal High-frequency information in three directions corresponds to A l+1 , exist Figure 1 The values ​​are LL, LH, and HH, as shown in the following formula:

[0050]

[0051] Optimization thresholding of frequency domain data specifically involves dividing the frequency domain values ​​into two groups based on their magnitude: C0, which corresponds to the portion between the minimum value (min) and the optimal threshold (T), and C1, which corresponds to the portion of gray levels between T+1 and the maximum value (max). The pixel distribution probability and mean of the two gray levels are then calculated, as shown in the following formula:

[0052]

[0053] W1 = 1 - W0

[0054]

[0055]

[0056] Based on this, the variance δ of the two types of regions is calculated. 2 =w0w1(u1-u0) 2 Using this variance as an evaluation index, the inter-class variance is calculated by iterating through the low-frequency coefficients from min to max and assigning values ​​to T. The T value corresponding to the maximum variance is the optimal threshold. The fusion weights for frequency domain information below T are set to 0, and the fusion weights for frequency domain information above T are set to 1. Based on this, the frequency domain signals are weighted and fused, and then inversely transformed into time domain information to reconstruct infrared thermal image data, obtaining foreground data containing defects after background removal. The calculation formula is as follows:

[0057]

[0058]

[0059] IV. Edge detection is performed on the reconstructed image, using the circumscribed regions of different grayscale transformation areas as the initial defect localization regions. Specific calculations include:

[0060] (1) First, edge detection is performed on the reconstructed image based on the Canny detection algorithm, and the detected edge contours are used as the dividing lines of different defect distribution areas;

[0061] (2) Combining the different defects that appear as circles or ellipses in the infrared thermal image, the area enclosed by the boundary line is marked to compensate for the possible loss of area sampling caused by imaging light.

[0062] (3) Statistically count the pixels enclosed by the boundary line region and solve for the center position (x, y) according to the following formula, where g(i,j) represents the pixel value of a point and n and m represent the size of the region.

[0063]

[0064] Within the enclosed area, the pixel furthest from the center is taken as the edge point. The distance between the edge point and the center point is calculated. A circular area is drawn with the center as the center and the distance between the edge point and the center point as the radius. This circular area is used as the initial location area for the defect.

[0065] V. Quantitative calculation of defect location based on pixel value extreme value distribution. The initial defect location area after compensation can be corrected. Based on the initial location area, statistical analysis is performed pixel-by-pixel in the corresponding area of ​​the reconstructed infrared thermal image, where g(i,j) represents a pixel value, and n and m represent the horizontal and vertical coordinate ranges of the area enclosed by the region.

[0066]

[0067] Based on this, calculate the difference A between the pixel value at the center of each edge region and the optimal threshold used in the maximum inter-class comparison; search each edge region and calculate the difference B between each pixel in the region and the center pixel; retain only the corresponding pixels in each edge region where B>0.5A, and finally select the pixels to form the accurate edge region.

[0068] VI. Feature Extraction and Model Training for Edge Regions. The original infrared thermal image corresponding to each precise edge region can be preprocessed. The purpose of preprocessing is to facilitate the training of the deep learning model. Preprocessing mainly includes:

[0069] 1) Perform scale normalization on the local image containing the edge region, and adjust the image to an M×N size using bilateral interpolation;

[0070] 2) Establish a sample dataset. Based on no less than 2,000 defect data, the dataset is expanded by methods such as rotation and adding random noise. The number of samples in the dataset is more than 20,000. The expanded dataset is randomly selected in a ratio of 1:0.5:0.5 to establish training set, validation set and test set respectively.

[0071] 3) Input the training set values ​​into the YOLOv4 model, perform optimization convergence calculation using gradient descent, and use the validation set to perform online detection and correction of the calculation results;

[0072] 4) Test the reliability and effectiveness of the model using the test set. Stop training when the detection accuracy is above 95% to obtain the optimal model.

[0073] VII. Based on an optimally trained deep learning model, defects in the bonding of composite material structures in current spacecraft are detected. All defect areas are normalized to the same size, and the processing results are used as input data to the model. The model outputs coordinates (x, y), defect category, and confidence score for each category as the identification result. Using the coordinates in the image, the three defect categories with the highest confidence, and the confidence scores of these three categories as basic information for each defect, a defect information index database for the bonding of composite material structures in current spacecraft is established.

[0074] 8. Based on the defect information index database, targeted verification and confirmation are conducted according to the actual spacecraft composite material structure bonding inspection requirements for certain key points. The host computer in the electrical module queries and controls the mechanical motion module to move the infrared thermal imaging module to the corresponding defect location, with visual indication provided by a laser component.

[0075] Example 2

[0076] Reference Figure 2 This diagram illustrates a flowchart of the steps involved in an infrared thermal imaging detection method for a spacecraft composite material structure product according to an embodiment of the present invention. This method is applied to the infrared thermal imaging detection system for the spacecraft composite material structure product described in Embodiment 1 above. Figure 2 As shown, the method may include the following steps:

[0077] Step 201: After entering the working state, preheat all components of the system.

[0078] The embodiments of the present invention can be applied to the infrared thermal imaging detection system for spacecraft composite material structure products in the above embodiment one.

[0079] The detection process in this embodiment of the invention may include the following stages:

[0080] Phase 1: Infrared Thermal Image Acquisition. Using the thermal imaging module of the automated defect detection system for spacecraft composite material structures, a panoramic scan and overall infrared thermal image of the adhesive bonding of the spacecraft composite material structure are acquired, and the acquired images are then fitted and reconstructed.

[0081] The second stage is infrared thermal image enhancement. The information in the fitted and reconstructed thermal image is processed with optimal fusion weights and reconstructed. The foreground and background are then described separately in the form of a binary image to segment the defective local area and the defect-free majority, thereby improving the detectability of defects.

[0082] The third stage: quantitative calculation of the defect area. Edge detection is performed using the binary image. Each detected local region is used as the initial calculation location of the defect. The extreme value distribution of pixels in the initial calculation location is calculated. The centroid of the pixel distribution is taken as the center of the defect area. Starting from the center of the area, the difference of pixel values ​​is calculated point by point towards the boundary of the initial calculation location. The difference is filtered by the contrast of the difference between the pixel values ​​at the center of the area and the mean pixel value of the defect-free area, and finally the precise location distribution of the defect area is confirmed.

[0083] Phase Four: Automated Defect Identification. Based on the calculated precise location of the defect area, the corresponding defect feature data is obtained from the local area of ​​the infrared thermographic image at the same location and input into a convolutional neural network for training and detection. Training involves using standard defects on the adhesive bonding of prefabricated spacecraft composite material structures as training samples to optimize the model. Detection utilizes the trained model to analyze and detect local defect areas in the input enhanced infrared thermographic images, achieving automated identification of different defects.

[0084] After entering the working state, the system components can be preheated. Specifically, after confirming that the mechanical motion module and infrared thermal imaging module are working properly, each component can be preheated for 10 minutes. First, a comparison specimen is used to adjust the layout of the detection parameters and acquisition components, and solidify the detection process and parameters to meet the detection requirements. The composite material structure of the spacecraft under test is installed and bonded. Based on the size and shape of the bonded composite material structure, the position of the infrared thermal imaging module is adjusted by the host computer controlled by the electrical module to ensure reliable and clear acquisition of the surface condition of the bonded composite material structure. The system is automatically divided into zones according to the inspectable area, and the route and speed of the mechanical motion module are configured and planned by the electrical module.

[0085] Step 202: Perform a panoramic scan of the composite material structure product of the spacecraft under test to acquire infrared thermal images of the composite material structure product of the spacecraft under test, and fit and reconstruct the infrared thermal images.

[0086] After completing the predictive processing, a panoramic scan of the spacecraft composite material structure product under test can be performed to acquire infrared thermal images of the product, and then the infrared thermal images can be fitted and reconstructed. Specifically, an infrared thermal imaging module can be used to acquire infrared thermal image data of the spacecraft composite material structure product. During the acquisition process, the infrared thermal imaging module applies pulsed heat flow to the bonded surface of the spacecraft composite material structure. After the pulsed heat flow, the bonded surface of the spacecraft composite material structure undergoes two stages: heating and cooling. By performing polynomial fitting on the temperature change and thermal imaging data, the original thermal image sequence is transformed into a reconstructed thermal image sequence, and the cooling stage is reconstructed and calculated.

[0087] Step 203: Perform optimal fusion weight processing and reconstruction on the image information in the fitted and reconstructed infrared thermal image, and obtain the corresponding binary image.

[0088] After reconstruction calculations, the image information in the fitted and reconstructed infrared thermal image can be processed and reconstructed using optimal fusion weights to obtain the corresponding binary image. Specifically, wavelet transform can be performed on the infrared thermal image data first. A low-pass filter is constructed using a scaling function to extract low-frequency information from the image, and a high-pass filter is constructed using a wavelet function to extract high-frequency information from the image. The information in the fitted and reconstructed thermal image is then processed and reconstructed using optimal fusion weights, and the foreground and background are described separately in the form of a binary image to achieve segmentation of the defective local area and the defect-free majority, thereby improving the detectability of defects.

[0089] Step 204: Perform edge detection on the binary image to determine the defect feature data of the composite material structure product of the spacecraft under test.

[0090] After obtaining the binary image, edge detection can be performed on the binary image to determine the defect feature data of the composite material structure product of the spacecraft under test. In this embodiment, edge detection can be performed on the reconstructed binary image based on the Canny detection algorithm to obtain the edge contour as the boundary line of different defect distribution areas; according to the circular or elliptical form of different defects in the infrared thermal image, the range of the area enclosed by the boundary line is calibrated; the pixels enclosed by the boundary line area are counted to obtain the initial defect location area of ​​the composite material structure product of the spacecraft under test; and the defect feature data is determined based on the initial defect location area.

[0091] In practical implementation, edge detection can be performed by combining binary images. The detected local regions are used as the initial calculation positions of the defect locality. The extreme value distribution of pixels in the initial calculation positions is calculated. The centroid of the pixel distribution is used as the center of the defect region. The difference of pixel values ​​is calculated point by point from the center of the region to the boundary of the initial calculation position. The difference is filtered by the contrast of the difference between the pixel values ​​at the center of the region and the average pixel value of the defect-free area, and finally the precise location distribution of the defect region is confirmed.

[0092] Step 205: Input the defect feature data into a convolutional neural network for detection to obtain the bonding defects of the composite material structure product of the spacecraft under test.

[0093] After determining the defect feature data of the composite material structure product of the spacecraft under test, the data can be input into a convolutional neural network for detection to obtain the bonding defects of the composite material structure product. In this embodiment, the defect feature data can be quantitatively calculated to obtain the defect edge region of the composite material structure product; a pre-trained model is then called to process the defect edge region to obtain the bonding defects of the composite material structure product.

[0094] Specifically, based on the calculated precise location of the defect area, the local area of ​​the infrared thermogram at the same location is used as the corresponding defect feature data, which is then input into a convolutional neural network for training and detection. Training involves using standard defects on the adhesive bonding of artificially prefabricated spacecraft composite material structures as training samples to optimize the model. Detection utilizes the trained model to analyze and detect local defect areas in the input enhanced infrared thermogram image, achieving automated identification of different defects.

[0095] The training process for this convolutional neural network can be described in the following steps:

[0096] 1. Perform scale normalization on the local image containing the edge region, and adjust the image to an M×N size using bilateral interpolation;

[0097] 2. Establish a sample dataset. Based on no less than 2,000 defect data, expand the dataset by methods such as rotation and adding random noise. The number of samples in the dataset should be more than 20,000. Randomly select samples from the expanded dataset in a ratio of 1:0.5:0.5 to establish training set, validation set and test set respectively.

[0098] 3. Input the training set into the YOLOv4 model, perform optimization convergence calculation through gradient descent, and use the validation set to perform online detection and correction of the calculation results;

[0099] 4. Test the model's reliability and effectiveness using the test set. Stop training when the detection accuracy is above 95% to obtain the optimal model.

[0100] After obtaining the optimal model through training, defects in the bonding of composite material structures in current spacecraft can be detected based on the optimally trained deep learning model. All defect areas are normalized to the same size, and the processing results are used as input data to the model. The model outputs coordinates (x, y), defect category, and confidence level for each category as the identification result. Using the coordinates in the image, the three defect categories with the highest confidence, and the confidence levels of these three categories as the basic information for each defect, a defect information index database for the bonding of composite material structures in current spacecraft is established.

[0101] Finally, based on the defect information index database, targeted verification and confirmation can be carried out according to the actual spacecraft composite material structure bonding inspection requirements for certain key points. Through queries in the electrical module's host computer, the mechanical motion module is controlled to move the infrared thermal imaging module to the corresponding defect location, and a laser component provides visual indication.

[0102] The specific embodiments described in this application are intended to enable those skilled in the art to gain a more comprehensive understanding of this application, but do not limit this application in any way. Therefore, those skilled in the art should understand that modifications or equivalent substitutions can still be made to this application; and all technical solutions and improvements that do not depart from the spirit and technical essence of this application should be covered within the scope of protection of this patent application.

[0103] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. An infrared thermographic inspection system for spacecraft composite material structural products, characterized in that, The system comprises an air floating base, a supporting gantry assembly, a motion guide rail, a conveying belt assembly, a heating assembly, an imaging assembly, a laser assembly, an electric control host computer and a spacecraft composite material structure to be tested, wherein The supporting gantry assembly is fixedly arranged on the upper side of the air floating base and forms a rectangular structure; The motion guide rail is fixedly connected to the inner side of the gantry assembly so that the heating assembly, the imaging assembly, the laser assembly and the spacecraft composite material structure to be tested move in the motion guide rail; The spacecraft composite material structure product to be tested is arranged at the central position of the rectangular structure on the upper side of the air floating base, and the conveying belt assembly is arranged between the supporting gantry assemblies to drive the spacecraft composite material structure to move; The imaging assembly, the heating assembly and the laser assembly are arranged at the positions of the air floating base where the supporting gantry assemblies are connected; The heating assembly is used to heat the bonding surface of the spacecraft composite material structure product to be tested during detection; The imaging assembly is used to collect and acquire the infrared thermal image of the bonding of the spacecraft composite material structure product to be tested on the basis of the temperature change of the bonding surface of the spacecraft composite material structure product to be tested; The electric control host computer is used to call a pre-trained model to process the infrared thermal image to detect the defect position of the spacecraft composite material structure product to be tested; The laser assembly is used to mark the defect position of the spacecraft composite material structure product to be tested; The electric control host computer performs fitting reconstruction on the infrared thermal image, performs optimal fusion weight processing and reconstruction on the image information in the infrared thermal image after fitting reconstruction, and acquires a corresponding binary image; edge detection is performed on the reconstructed binary image based on a Canny detection algorithm to obtain an edge contour as a boundary line of different defect distribution regions, the range of the region surrounded by the boundary line is calibrated according to the circular or elliptical form of different defects in the infrared thermal image, the pixels in the boundary line region are counted to obtain an initial defect positioning region of the spacecraft composite material structure product to be tested; the defect feature data is determined according to the initial defect positioning region; the defect feature data is quantitatively calculated to obtain a defect edge region of the spacecraft composite material structure product to be tested; the pre-trained model is called to process the defect edge region to obtain the bonding defect of the spacecraft composite material structure product to be tested. The motion guide rail comprises a first motion guide rail, a second motion guide rail and a third motion guide rail, the guide rail directions of the first motion guide rail, the second motion guide rail and the third motion guide rail are perpendicular to each other; the infrared thermal image module is driven to move to the corresponding defect position and is visually indicated by the laser assembly, the electric control host computer adjusts the position of the infrared thermal image module to ensure that the spacecraft composite material structure bonding surface state can be reliably and clearly collected, and the motion route and speed are configured and planned according to the detectable area.

2. A method for infrared thermographic inspection of a spacecraft composite structure product, applied to the inspection system of claim 1, characterized in that, The method comprises: After entering the working state, the preheating treatment is performed on each component of the system; The panoramic scanning is performed on the spacecraft composite material structure product to be tested to collect an infrared thermal image of the spacecraft composite material structure product to be tested, and the infrared thermal image is fitted and reconstructed; Optimal fusion weight processing and reconstruction are performed on image information in the infrared thermal image after fitting and reconstruction, and a corresponding binary image is obtained; Edge detection is performed on the binary image to determine defect feature data of the spacecraft composite material structure product to be tested; The defect feature data is input into a convolutional neural network for detection to obtain a bonding defect of the spacecraft composite material structure product to be tested.

3. The method of claim 2, wherein, The edge detection on the binary image to determine the defect feature data of the spacecraft composite material structure product to be tested comprises: Edge detection is performed on the reconstructed binary image based on a Canny detection algorithm to obtain an edge contour as a boundary line of different defect distribution regions; According to the fact that different defects present in the form of a circle or an ellipse in the infrared thermal image, the range of the region surrounded by the boundary line is calibrated; The pixel points in the region surrounded by the boundary line are counted to obtain an initial defect positioning region of the spacecraft composite material structure product to be tested; The defect feature data is determined according to the initial defect positioning region.

4. The method of claim 3, wherein, The input of the defect feature data into the convolutional neural network for detection to obtain the bonding defect of the spacecraft composite material structure product to be tested comprises: Quantitative calculation is performed on the defect feature data to obtain a defect edge region of the spacecraft composite material structure product to be tested; A pre-trained model is called to process the defect edge region to obtain the bonding defect of the spacecraft composite material structure product to be tested.

Images