Satellite manufacturing intelligent detection method, system, equipment and medium

Abstract

The invention discloses a satellite manufacturing intelligent detection method, system and device and a medium, and the method comprises the steps: determining a to-be-detected region of a to-be-detected target element based on a detection database, and obtaining an initial mapping image of the target element through an augmented reality AR space registration technology; obtaining sensing data of a to-be-detected area of the to-be-detected target element through a preset sensing device; inputting the feature data of the target element after the perception data preprocessing into a pre-trained deep neural network model to obtain a detection and recognition result of the target element output by the deep neural network model; and overlapping and displaying the detection and recognition result of the target element on the surface of the initial mapping image of the target element in a visual layer form through augmented reality (AR) equipment to obtain a detection result image, and completing the detection of the target element. According to the method, the detection efficiency and the detection accuracy of the satellite manufacturing element are effectively improved, and the quality guarantee of the satellite manufacturing element is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and more specifically, to an intelligent inspection method, system, equipment, and medium for satellite manufacturing. Background Technology

[0002] With the rapid development of the aerospace industry, satellite manufacturing is showing a trend towards mass production, high precision, and personalization. Traditional quality inspection methods rely on manual visual inspection, gauge measurement, or 2D document standard comparison, which suffers from long information transfer chains, low operational efficiency, and high error rates, making it difficult to meet the demands of high-paced manufacturing processes for intelligent, visualized, and data-driven inspection processes. Augmented reality (AR) technology has been widely used in industrial manufacturing in recent years, which can integrate virtual models with the real environment to provide operators with intuitive task assistance. However, most existing AR-based inspection methods rely on static models or manually labeled standards, failing to achieve real-time perception of component status and high-dimensional data comparison. In addition, the lack of multi-sensor fusion mechanisms results in insufficient robustness and accuracy in identifying complex defects.

[0003] Therefore, there is an urgent need to study a novel detection method that deeply integrates AR visualization with multimodal perception technology to build an intelligent detection system, thereby improving the quality control capabilities in the satellite manufacturing process. Summary of the Invention

[0004] This specification provides a smart inspection method, system, device, and medium for satellite manufacturing, to overcome at least one technical problem existing in related technologies.

[0005] According to a first aspect of the embodiments of this specification, a smart inspection method for satellite manufacturing is provided, comprising: A testing database is constructed by associating the three-dimensional structural models, process parameters, testing standards, and historical defect cases of satellite components. The testing database categorizes and saves the three-dimensional structural models, process parameters, and testing standards of satellite components according to type and batch. For the target component to be inspected, the inspection information of the target component is determined according to the inspection standards and historical defect cases corresponding to the type of the target component in the inspection database. The inspection area of ​​the target component is determined according to the inspection information and the three-dimensional structural model of the target component. The inspection area and the target component are associated and mapped in the inspection scene of the target component using augmented reality (AR) spatial registration technology to obtain the initial mapping image of the target component. The initial mapping image includes the three-dimensional image of the target component and the position label of the inspection area. The sensing data of the detection area of ​​the target element to be detected is obtained through a preset sensing device, which is arranged in the detection scene of the target element. The sensory data of the target element to be detected is preprocessed to obtain the feature data of the target element. The feature data of the target element is used as the input of a pre-trained deep neural network model to obtain the detection and recognition result of the target element output by the deep neural network model. The deep neural network model calculates the matching and recognition result of the element with the preset defect category based on the input feature data of the element. The detection and recognition results of the target element are overlaid on the surface of the initial mapped image of the target element in the form of a visual layer using an augmented reality (AR) device to obtain the detection result image, thus completing the detection of the target element, and the detection result is added to the detection database.

[0006] Optionally, the step of acquiring sensing data of the detection area of ​​the target element to be detected through a preset sensing device includes: A 3D structural reconstruction image of the area to be detected is obtained using a structured light depth camera; Infrared thermal images of the area to be detected are obtained using an infrared thermal imager. Laser ranging data of the area to be detected is obtained using a laser rangefinder. Force data of the area to be detected is obtained through micro-force sensors, tactile arrays, and acoustic sensors.

[0007] Optionally, the step of preprocessing the perceptual data of the target element to be detected to obtain the feature data of the target element, and using the feature data of the target element as input to a pre-trained deep neural network model to obtain the detection and recognition result of the target element output by the deep neural network model includes: After preprocessing the infrared thermal image, 3D structural reconstruction image, laser ranging data, and force sensing data of the target element to be detected, the infrared thermal image, 3D structural reconstruction image, laser ranging data, and force sensing data are respectively input into the pre-trained deep neural network model as different modal data. The intermediate layer of the deep neural network model is used to fuse features of various modal data through an attention mechanism. Key defect features are extracted by a convolutional encoder and output as probability distribution data according to categories, including overheating, misalignment, cracks, and loosening. Based on the probability distribution data of different types of defects, the defect distribution in the area to be detected of the target component is obtained, and the detection and identification results are obtained.

[0008] Optionally, the deep neural network model is trained through the following steps: A training sample set is generated based on the defect category and the corresponding component samples. The training sample set contains multiple training sample groups, each of which contains a defect category and multiple component samples. Each component sample corresponds to a matching tag. If a component sample in a component sample group matches the corresponding defect type, the matching tag of the component sample is recorded as a first tag value. If a component sample in a component sample group does not match the corresponding defect category, the matching tag of the component sample is recorded as a second tag value, which is different from the first tag value. The deep neural network model is trained using the training sample set, and the deep neural network model is used to calculate whether the input component sample matches the defect category.

[0009] Optionally, the step of overlaying the detection and recognition results of the target element onto the surface of the initial mapped image of the target element in the form of a visual layer using an augmented reality (AR) device to obtain a detection result image, thereby completing the detection of the target element, and adding the detection results to the detection database includes: Based on the detection and recognition results of the target element and the position marking of the area to be detected on the target element, the detection result image is obtained by superimposing it on the surface of the initial mapping image of the target element in the form of color coding, text label, or thermal image through an augmented reality (AR) device. The initial mapping image of the target element, the perception data, and the detection results are saved to the detection database.

[0010] Optionally, after the steps of preprocessing the perceptual data of the target element to be detected to obtain the feature data of the target element, using the feature data of the target element as input to a pre-trained deep neural network model, and obtaining the detection and recognition result of the target element output by the deep neural network model, the method further includes: Supervised learning training of a deep neural network model is performed using pre-obtained soft labels and historical case training sets, enabling the deep neural network model to make dynamic judgments and classifications based on real-time perceived data.

[0011] Optionally, the step of preprocessing the sensing data of the target element to be detected to obtain the feature data of the target element includes: The edge computing unit of the sensing device performs noise reduction preprocessing on the sensing data of the target element to be detected, and obtains the feature data of the target element.

[0012] According to a second aspect of the embodiments of this specification, a satellite manufacturing intelligent inspection system is provided, including a database module, an association mapping module, a data acquisition module, a defect identification module, and a result mapping module, wherein... The database module is configured to associate the three-dimensional structural models, process parameters, testing standards, and historical defect cases of satellite components to construct a testing database. The testing database will classify and save the three-dimensional structural models, process parameters, and testing standards of satellite components according to type and batch. The association mapping module is configured to, for a target element to be detected, determine the detection information of the target element according to the detection standards and historical defect cases corresponding to the type of the target element in the detection database, determine the detection area of ​​the target element according to the detection information and the three-dimensional structural model of the target element, and perform association mapping between the detection area and the target element in the detection scene of the target element using augmented reality (AR) spatial registration technology to obtain an initial mapping image of the target element. The initial mapping image includes a three-dimensional image of the target element and the position label of the detection area. The data acquisition module is configured to acquire sensing data of the detection area of ​​the target element to be detected through a preset sensing device, wherein the sensing device is arranged in the detection scene of the target element. The defect identification module is configured to preprocess the perception data of the target component to be detected to obtain the feature data of the target component, use the feature data of the target component as the input of a pre-trained deep neural network model, and obtain the detection and identification result of the target component output by the deep neural network model. The deep neural network model calculates the matching and identification result of the component with the preset defect category based on the input feature data of the component. The result mapping module is configured to overlay the detection and recognition results of the target element onto the surface of the initial mapping image of the target element in the form of a visual layer through an augmented reality (AR) device, thereby obtaining a detection result image, completing the detection of the target element, and adding the detection results to the detection database.

[0013] According to a third aspect of the embodiments of this specification, a computing device is provided, including a storage device and a processor, the storage device being used to store a computer program, and the processor running the computer program to cause the computing device to perform the steps of the intelligent detection method for satellite manufacturing.

[0014] According to a fourth aspect of the embodiments of this specification, a storage medium is provided that stores a computer program used in the computing device, which, when executed by a processor, implements the steps of the intelligent detection method for satellite manufacturing.

[0015] The beneficial effects of the embodiments in this specification are as follows: This specification provides an intelligent inspection method, system, device, and medium for satellite manufacturing. The inspection method retrieves corresponding standards and cases from a database to improve the standardization and consistency of standard implementation. It integrates AR technology and multimodal perception technology to achieve rapid identification, standard comparison, and visual feedback of components across multiple physical characteristics, such as size, temperature field, and structural integrity. Through multi-source fusion of infrared, structured light, and laser, it enhances the ability to identify latent defects, such as internal cracks, misalignments, and loosening, significantly improving inspection accuracy. Based on AR combined with layer semantic annotation, it intuitively displays inspection results and anomaly distribution, enhancing the visualization of the inspection process. In summary, the inspection method in this specification effectively improves inspection efficiency, accuracy, and intelligence, greatly enhancing quality control capabilities in the satellite manufacturing process.

[0016] The innovative aspects of the embodiments in this specification include: 1. This specification employs an intelligent detection mechanism that integrates AR and multimodal perception. By integrating multiple sensors such as 3D vision, infrared thermal imaging, laser ranging, and force / tactile sensing, it achieves multidimensional information perception of key satellite components. After the data from each modality is processed in real time by the edge computing module, a deep neural network performs multi-source feature fusion and anomaly identification to accurately locate complex physical defects such as cracks, loosening, and misalignment. At the same time, the database automatically matches the corresponding detection standards and defect templates to achieve dynamic calling and updating of detection rules, which is one of the innovative points of the embodiments in this specification.

[0017] 2. This specification adopts a virtual-real fusion visual interaction enhancement mechanism, which uses spatial registration technology to achieve high-precision overlay of AR visualization layers and actual parts. The detection results are presented intuitively in the AR device in a semantic form. For example, the defect type and location are marked in real time through color coding, text labels, hot zone images, etc., and supervised learning training is carried out in combination with historical detection data. Prompts and suggestions are generated based on real-time data, which is one of the innovative points of the embodiments of this specification. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or related technologies of this specification, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an intelligent inspection method for satellite manufacturing provided in one embodiment of this specification. Figure 2This is a schematic diagram of the structure of an intelligent inspection system for satellite manufacturing provided in one embodiment of this specification; Figure 3 This is a schematic diagram of the structure of a computing device provided in one embodiment of this specification; Figure 4 This is a schematic diagram of the structure of a storage medium provided in one embodiment of this specification. Detailed Implementation

[0020] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be noted that the terms "comprising" and "having," and any variations thereof, in the embodiments and drawings of this specification are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0022] This specification discloses an intelligent inspection method, system, equipment, and medium for satellite manufacturing, which will be described in detail below.

[0023] Figure 1 This is a flowchart illustrating an embodiment of an intelligent inspection method for satellite manufacturing provided in this specification. Figure 1 As shown, a smart inspection method for satellite manufacturing includes: S110. The three-dimensional structural model, process parameters, testing standards and historical defect cases of satellite components are associated to construct a testing database. The testing database classifies and saves the three-dimensional structural model, process parameters and testing standards of satellite components according to type and batch.

[0024] S120. For the target component to be inspected, the inspection information of the target component is determined according to the inspection standards and historical defect cases corresponding to the type of the target component in the inspection database. The inspection area of ​​the target component is determined according to the inspection information and the three-dimensional structural model of the target component. The inspection area and the target component are associated and mapped in the inspection scene of the target component using augmented reality (AR) spatial registration technology to obtain the initial mapping image of the target component. The initial mapping image includes the three-dimensional image of the target component and the position label of the inspection area.

[0025] The AR spatial registration technology is achieved by fusing structured light positioning with an inertial navigation unit (IMU), with a registration accuracy of ±0.1mm.

[0026] S130. Acquire sensing data of the detection area of ​​the target element to be detected through a preset sensing device, wherein the sensing device is arranged in the detection scene of the target element. The sensing device includes a structured light depth camera, an infrared thermal imager, a laser rangefinder, a micro-force sensor and tactile array, and an acoustic sensor.

[0027] In a specific embodiment, step S130, which involves acquiring sensing data of the detection area of ​​the target element to be detected through a preset sensing device, includes: S132. Obtain a three-dimensional structural reconstruction image of the area to be detected using a structured light depth camera. The three-dimensional structural reconstruction image is used for the surface and spatial structure reconstruction of the target element.

[0028] S134. Obtain an infrared thermal image of the area to be detected using an infrared thermal imager. The infrared thermal image is used to detect material thermal anomalies of the target component.

[0029] S136. Acquire laser ranging data of the area to be detected using a laser rangefinder. The laser ranging data is used for the size measurement of the target component.

[0030] S138. Force data of the area to be detected is obtained through a micro-force sensor, a tactile array, and an acoustic sensor. The micro-force sensor and the tactile array are used to detect the stability of the connector, and the acoustic sensor is used for non-contact structural integrity assessment.

[0031] S140. Preprocess the perception data of the target element to be detected to obtain the feature data of the target element. Use the feature data of the target element as the input of a pre-trained deep neural network model to obtain the detection and recognition result of the target element output by the deep neural network model. The deep neural network model calculates the matching and recognition result of the element with the preset defect category based on the input feature data of the element.

[0032] In a specific embodiment, S140, the step of preprocessing the perception data of the target element to be detected to obtain the feature data of the target element, and using the feature data of the target element as the input of a pre-trained deep neural network model to obtain the detection and recognition result of the target element output by the deep neural network model, includes: S142. After preprocessing the infrared thermal image, three-dimensional structure reconstruction image, laser ranging data and force sensing data of the target element to be detected, the infrared thermal image, three-dimensional structure reconstruction image, laser ranging data and force sensing data are respectively input into the pre-trained deep neural network model as different modal data. S144. The intermediate layer of the deep neural network model is used to perform feature fusion on the data of each modality through the attention mechanism. The key defect features are extracted by the convolutional encoder and output as probability distribution data according to the categories, including overheating, misalignment, cracks and loosening. S146. Based on the probability distribution data of different types of defects, the defect distribution of the target component in the area to be detected is obtained, and the detection and identification results are obtained.

[0033] The deep neural network model is trained through the following steps: A training sample set is generated based on the defect category and the corresponding component samples. The training sample set contains multiple training sample groups, each of which contains a defect category and multiple component samples. Each component sample corresponds to a matching tag. If a component sample in a component sample group matches the corresponding defect type, the matching tag of the component sample is recorded as a first tag value. If a component sample in a component sample group does not match the corresponding defect category, the matching tag of the component sample is recorded as a second tag value, which is different from the first tag value. The deep neural network model is trained using the training sample set, and the deep neural network model is used to calculate whether the input component sample matches the defect category.

[0034] The step of preprocessing the sensing data of the target element to be detected to obtain the feature data of the target element includes: The edge computing unit of the sensing device performs noise reduction preprocessing on the sensing data of the target element to be detected, and obtains the feature data of the target element.

[0035] After step S140, which involves preprocessing the perceptual data of the target element to be detected to obtain feature data of the target element, using the feature data of the target element as input to a pre-trained deep neural network model, and obtaining the detection and recognition result of the target element output by the deep neural network model, the method further includes: S148. Supervised learning training of the deep neural network model is performed using pre-obtained soft labels and historical case training sets, so that the deep neural network model can make dynamic judgments and classification labels based on real-time perceived data.

[0036] S150. The detection and recognition results of the target element are overlaid on the surface of the initial mapping image of the target element in the form of a visual layer through an augmented reality (AR) device to obtain the detection result image, thus completing the detection of the target element, and the detection result is added to the detection database.

[0037] In specific implementation, S150, the step of overlaying the detection and recognition results of the target element onto the surface of the initial mapped image of the target element in the form of a visual layer using an augmented reality (AR) device to obtain a detection result image, completing the detection of the target element, and adding the detection results to the detection database, includes: S152. Based on the detection and recognition results of the target element and the position marking of the area to be detected on the target element, the detection result image is obtained by superimposing it on the surface of the initial mapping image of the target element in at least one of the following forms: color coding, text label, and hot zone image, using an augmented reality (AR) device. S154. Save the initial mapping image, perception data and detection results of the target element to the detection database.

[0038] The augmented reality (AR) display device is a head-mounted display or tablet computer that supports gesture interaction and voice control.

[0039] In practical production, this method is applied to the inspection of solar panel deployment mechanism components. The 3D models of each component, such as hinges, shafts, motors, and connecting holes, are imported into a database along with inspection standards. Component feature descriptions based on historical defect cases are established, including early warning rules and common defect patterns. Operators wear AR headsets and use a spatial registration system to identify the actual position of the deployment mechanism. The system automatically loads the inspection scene for that component, displaying all inspection points, such as high-risk wear areas, heat sources, and electrical connectors. Next, a structured light system scans the hinge gaps to determine if they exceed limits; an infrared camera detects overheating areas after motor operation; laser ranging measures the deployment angle and displacement errors; and a micro-sensor detects changes in shaft stiffness. Furthermore, a convolutional neural network compares the infrared images with typical overheating defect maps to identify potential motor anomalies. The abnormal locations are automatically highlighted on the AR interface, and a "replacement suggestion" is provided. The inspection process is automatically recorded and a structured report is generated, including a timestamp, operator ID, anomaly number, and suggested remedial measures, which is then uploaded to the Manufacturing Execution System (MES).

[0040] When this method is applied to the inspection of thermal control components in the attitude control propulsion module, it identifies the thermal control multilayer heat insulation sheet and the heating wire area, and establishes positioning registration through natural feature recognition; it uses a thermal imager to identify whether the heat distribution is uniform after power-on, and whether there is thermal short circuit or thermal resistance; it compares features with historical thermal anomaly cases and provides image comparison and risk prediction.

[0041] Figure 2 This is a schematic diagram of the structure of an intelligent inspection system for satellite manufacturing, provided as an embodiment of this specification. Figure 2As shown, a satellite manufacturing intelligent inspection system 200 includes a database module 210, an association mapping module 220, a data acquisition module 230, a defect identification module 240, and a result mapping module 250, wherein... The database module 210 is configured to associate the three-dimensional structural model, process parameters, testing standards and historical defect cases of satellite components to construct a testing database. The testing database classifies and saves the three-dimensional structural model, process parameters and testing standards of satellite components according to type and batch.

[0042] The association mapping module 220 is configured to, for a target element to be detected, determine the detection information of the target element according to the detection standards and historical defect cases corresponding to the type of the target element in the detection database, determine the detection area of ​​the target element according to the detection information and the three-dimensional structural model of the target element, and perform association mapping between the detection area and the target element in the detection scene of the target element using augmented reality (AR) spatial registration technology to obtain an initial mapping image of the target element. The initial mapping image includes a three-dimensional image of the target element and the position label of the detection area.

[0043] The data acquisition module 230 is configured to acquire sensing data of the detection area of ​​the target element to be detected through a preset sensing device, wherein the sensing device is arranged in the detection scene of the target element.

[0044] The defect identification module 240 is configured to preprocess the perception data of the target element to be detected to obtain the feature data of the target element, use the feature data of the target element as the input of a pre-trained deep neural network model, and obtain the detection and identification result of the target element output by the deep neural network model. The deep neural network model calculates the matching and identification result of the element with the preset defect category based on the input feature data of the element.

[0045] The result mapping module 250 is configured to overlay the detection and recognition results of the target element onto the surface of the initial mapping image of the target element in the form of a visual layer through an augmented reality (AR) device, thereby obtaining a detection result image, completing the detection of the target element, and adding the detection results to the detection database.

[0046] The inspection system is integrated with the Manufacturing Execution System (MES) to achieve closed-loop management of inspection data and problem tracking.

[0047] Figure 3 This is a schematic diagram of the structure of a computing device provided in one embodiment of this specification. Figure 3As shown, a computing device 300 includes a storage device 310 and a processor 320. The storage device 310 stores a computer program, and the processor 320 runs the computer program to enable the computing device to perform the steps of the intelligent detection method for satellite manufacturing.

[0048] Figure 4 This is a schematic diagram of the structure of a storage medium provided in one embodiment of this specification. For example... Figure 4 As shown, a storage medium 400 stores a computer program 410 used in the computing device, which, when executed by a processor, implements the steps of the intelligent detection method for satellite manufacturing.

[0049] In summary, the embodiments of this specification provide a satellite manufacturing intelligent inspection method, system, equipment, and medium. This method integrates augmented reality visualization and multimodal perception for satellite manufacturing quality inspection, constructing an inspection method that integrates multi-dimensional spatial information overlay, standard invocation, real-time comparison, and intelligent recognition. It breaks through the limitations of existing AR inspection systems and has high inspection efficiency, accuracy, and intelligence. The embodiments of this specification are applicable to the manufacturing inspection of various components such as structures, electrical systems, and thermal control systems of various spacecraft, including communication satellites, remote sensing satellites, and navigation satellites, and have extremely high industrial application value and promotion prospects.

[0050] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of one embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention.

[0051] Those skilled in the art will understand that the modules in the apparatus of the embodiments can be distributed in the apparatus of the embodiments as described in the embodiments, or they can be located in one or more devices different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart inspection method for satellite manufacturing, characterized in that, include: A testing database is constructed by associating the three-dimensional structural models, process parameters, testing standards, and historical defect cases of satellite components. The testing database categorizes and saves the three-dimensional structural models, process parameters, and testing standards of satellite components according to type and batch. For the target component to be inspected, the inspection information of the target component is determined according to the inspection standards and historical defect cases corresponding to the type of the target component in the inspection database. The inspection area of ​​the target component is determined according to the inspection information and the three-dimensional structural model of the target component. The inspection area and the target component are associated and mapped in the inspection scene of the target component using augmented reality (AR) spatial registration technology to obtain the initial mapping image of the target component. The initial mapping image includes the three-dimensional image of the target component and the position label of the inspection area. The sensing data of the detection area of ​​the target element to be detected is obtained through a preset sensing device, which is arranged in the detection scene of the target element. The sensory data of the target element to be detected is preprocessed to obtain the feature data of the target element. The feature data of the target element is used as the input of a pre-trained deep neural network model to obtain the detection and recognition result of the target element output by the deep neural network model. The deep neural network model calculates the matching and recognition result of the element with the preset defect category based on the input feature data of the element. The detection and recognition results of the target element are overlaid on the surface of the initial mapped image of the target element in the form of a visual layer using an augmented reality (AR) device to obtain the detection result image, thus completing the detection of the target element, and the detection result is added to the detection database.

2. The method according to claim 1, characterized in that, The step of acquiring sensing data of the detection area of ​​the target element to be detected through a preset sensing device includes: A 3D structural reconstruction image of the area to be detected is obtained using a structured light depth camera; Infrared thermal images of the area to be detected are obtained using an infrared thermal imager. Laser ranging data of the area to be detected is obtained using a laser rangefinder. Force data of the area to be detected is obtained through micro-force sensors, tactile arrays, and acoustic sensors.

3. The method according to claim 2, characterized in that, The steps of preprocessing the perceptual data of the target element to be detected to obtain the feature data of the target element, using the feature data of the target element as input to a pre-trained deep neural network model, and obtaining the detection and recognition result of the target element output by the deep neural network model include: After preprocessing the infrared thermal image, 3D structural reconstruction image, laser ranging data, and force sensing data of the target element to be detected, the infrared thermal image, 3D structural reconstruction image, laser ranging data, and force sensing data are respectively input into the pre-trained deep neural network model as different modal data. The intermediate layer of the deep neural network model is used to fuse features of various modal data through an attention mechanism. Key defect features are extracted by a convolutional encoder and output as probability distribution data according to categories, including overheating, misalignment, cracks, and loosening. Based on the probability distribution data of different types of defects, the defect distribution in the area to be detected of the target component is obtained, and the detection and identification results are obtained.

4. The method according to claim 1, characterized in that, The deep neural network model is trained through the following steps: A training sample set is generated based on the defect category and the corresponding component samples. The training sample set contains multiple training sample groups, each of which contains a defect category and multiple component samples. Each component sample corresponds to a matching tag. If a component sample in a component sample group matches the corresponding defect type, the matching tag of the component sample is recorded as a first tag value. If a component sample in a component sample group does not match the corresponding defect category, the matching tag of the component sample is recorded as a second tag value, which is different from the first tag value. The deep neural network model is trained using the training sample set, and the deep neural network model is used to calculate whether the input component sample matches the defect category.

5. The method according to claim 1, characterized in that, The steps of overlaying the detection and recognition results of the target element onto the surface of the initial mapped image of the target element in the form of a visual layer using an augmented reality (AR) device to obtain a detection result image, completing the detection of the target element, and adding the detection results to the detection database include: Based on the detection and recognition results of the target element and the position marking of the area to be detected on the target element, the detection result image is obtained by superimposing it on the surface of the initial mapping image of the target element in the form of color coding, text label, or thermal image through an augmented reality (AR) device. The initial mapping image of the target element, the perception data, and the detection results are saved to the detection database.

6. The method according to claim 1, characterized in that, After the steps of preprocessing the perceptual data of the target element to be detected to obtain the feature data of the target element, using the feature data of the target element as input to a pre-trained deep neural network model, and obtaining the detection and recognition result of the target element output by the deep neural network model, the method further includes: Supervised learning training of a deep neural network model is performed using pre-obtained soft labels and historical case training sets, enabling the deep neural network model to make dynamic judgments and classifications based on real-time perceived data.

7. The method according to claim 1, characterized in that, The step of preprocessing the sensing data of the target element to be detected to obtain the feature data of the target element includes: The edge computing unit of the sensing device performs noise reduction preprocessing on the sensing data of the target element to be detected, and obtains the feature data of the target element.

8. A satellite manufacturing intelligent inspection system, characterized in that, It includes a database module, an association mapping module, a data acquisition module, a defect identification module, and a result mapping module, among which... The database module is configured to associate the three-dimensional structural models, process parameters, testing standards, and historical defect cases of satellite components to construct a testing database. The testing database will classify and save the three-dimensional structural models, process parameters, and testing standards of satellite components according to type and batch. The association mapping module is configured to, for a target element to be detected, determine the detection information of the target element according to the detection standards and historical defect cases corresponding to the type of the target element in the detection database, determine the detection area of ​​the target element according to the detection information and the three-dimensional structural model of the target element, and perform association mapping between the detection area and the target element in the detection scene of the target element using augmented reality (AR) spatial registration technology to obtain an initial mapping image of the target element. The initial mapping image includes a three-dimensional image of the target element and the position label of the detection area. The data acquisition module is configured to acquire sensing data of the detection area of ​​the target element to be detected through a preset sensing device, wherein the sensing device is arranged in the detection scene of the target element. The defect identification module is configured to preprocess the perception data of the target component to be detected to obtain the feature data of the target component, use the feature data of the target component as the input of a pre-trained deep neural network model, and obtain the detection and identification result of the target component output by the deep neural network model. The deep neural network model calculates the matching and identification result of the component with the preset defect category based on the input feature data of the component. The result mapping module is configured to overlay the detection and recognition results of the target element onto the surface of the initial mapping image of the target element in the form of a visual layer through an augmented reality (AR) device, thereby obtaining a detection result image, completing the detection of the target element, and adding the detection results to the detection database.

9. A computing device, characterized in that, The device includes a storage device and a processor, the storage device being used to store a computer program, and the processor running the computer program to cause the computing device to perform the steps of the method according to any one of claims 1-7.

10. A storage medium, characterized in that, It stores a computer program used in the computing device of claim 9, which, when executed by a processor, implements the steps of the method of any one of claims 1-7.

Images