A method and system for detecting infrared stealth coatings on transparent aerospace components.

Abstract

This invention provides a method and system for detecting infrared stealth coatings on transparent aerospace components. The invention relates to the field of aerospace transparent component detection technology. The method includes: acquiring an irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component; constructing a thermal map; processing the thermal map using a graph neural network to obtain high-risk, suspicious thermal regions; clustering multiple suspicious edge thermal points to obtain K clusters; generating suspicious edge thermal defect lines based on the K clusters; determining multiple suspicious intermediate test points based on the multiple suspicious thermal regions and suspicious edge thermal defect lines; and determining the detection result of the infrared stealth coating on the transparent aerospace component based on the high-risk, suspicious thermal regions, suspicious edge thermal defect lines, and multiple suspicious intermediate test points. This method can efficiently and accurately identify global defects in the infrared stealth coating of transparent aerospace components and quantify defect risks.

Description

Technical Field

[0001] This invention relates to the field of aviation transparent parts inspection technology, specifically to a method and system for detecting infrared stealth coatings on aviation transparent parts. Background Technology

[0002] Transparent components in aviation are core parts of aircraft, such as cockpit windshields and nose cones. The infrared stealth coating process for these components is susceptible to factors such as process technology, substrate, and environment, resulting in hidden defects like uneven thickness and poor edge adhesion. These defects directly compromise infrared stealth performance and threaten flight safety. Current inspection technologies for infrared stealth coatings on aviation transparent components mainly include visual inspection, ultrasonic inspection, and conventional infrared thermography. Visual inspection can only identify macroscopic surface defects, with an extremely low detection rate for minute defects within the coating and at its edges. Ultrasonic inspection is easily affected by the optical material properties of transparent components, limiting its accuracy and making large-area rapid scanning difficult. Conventional infrared thermography often uses a fixed-frequency illumination mode, capturing only the static thermal distribution on the coating surface. It cannot effectively distinguish between normal thermal fluctuations during the spraying process and abnormal thermal signals caused by defects, especially lacking the ability to identify defects in transition zones between suspicious thermal areas and edge defects, as well as defects along minute diffusion paths, leading to missed or false detections. Meanwhile, existing detection technologies lack systematic coverage of defect-related areas, often focusing only on isolated abnormal points or regions. This neglects the spatial correlation and defect diffusion patterns between suspected thermal areas, edge defects, and transition areas, resulting in detection results that cannot fully reflect the overall quality status of the coating. Furthermore, traditional detection methods rely on human experience for defect judgment, which is highly subjective and lacks standardized judgment criteria, making it difficult to meet the high-precision, high-reliability, and automated requirements of aerospace equipment for infrared stealth coating detection.

[0003] Therefore, how to efficiently and accurately identify defects across the entire infrared stealth coating of transparent aerospace components and quantify defect risks is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem addressed by this invention is how to efficiently and accurately identify defects across the entire infrared stealth coating of transparent aerospace components and quantify the risk of these defects.

[0005] According to a first aspect, the present invention provides a method for detecting an infrared stealth coating on an aerospace transparent component, comprising: acquiring an illumination video of the initial infrared frequency of the infrared stealth coating being sprayed on the aerospace transparent component to be detected; determining multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency of the infrared stealth coating being sprayed on the aerospace transparent component to be detected; acquiring an illumination video of the first detection infrared frequency of each suspicious thermal region; constructing a thermal map, wherein the thermal map includes multiple suspicious thermal regions and multiple edges between the multiple suspicious thermal regions, and the node features of each suspicious thermal region include the illumination video of the initial infrared frequency of each suspicious thermal region, and the first detection infrared frequency of each suspicious thermal region. The first detection infrared frequency irradiation video of the thermal region is used, and the edges between nodes represent the distance information of the suspected thermal region. The thermal map is processed using a graph neural network to obtain high-risk suspected thermal regions. K clusters are obtained based on the multiple suspected edge thermal point information. Suspicious edge thermal defect lines are generated based on the K clusters. Multiple suspicious intermediate test points are determined based on the multiple suspected thermal regions and the suspicious edge thermal defect lines. Each suspicious intermediate test point is a test point in the region between a suspected thermal region and a suspicious edge thermal defect line. The detection results of the infrared stealth coating on the aviation transparent component are determined based on the high-risk suspected thermal regions, the suspicious edge thermal defect lines, and the multiple suspicious intermediate test points.

[0006] In one possible implementation, the determination of the detection results of the infrared stealth coating of the aerospace transparent component based on the high-risk suspected thermal region, the suspected edge thermal defect line, and the multiple suspected intermediate test points includes: determining multiple high-risk suspected points in the high-risk suspected thermal region based on the high-risk suspected thermal region; determining high-risk suspected points in the suspected edge thermal defect line based on the suspected edge thermal defect line; and acquiring phase-locked thermal images of the multiple high-risk suspected points in the high-risk suspected thermal region, the high-risk suspected points in the suspected edge thermal defect line, and the multiple suspected intermediate test points. Based on phase-locked thermal imaging of multiple high-risk and suspicious points in the high-risk and suspicious thermal areas, phase-locked thermal imaging of high-risk and suspicious points in the suspicious edge thermal defect lines, phase-locked thermal imaging of multiple suspicious intermediate test points, information on multiple suspicious edge thermal points, irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be inspected, and irradiation video of the first detection infrared frequency of each suspicious thermal area, a defect probability distribution map of the infrared stealth coating is generated; the detection result of the infrared stealth coating of the aviation transparent part is determined based on the defect probability distribution map of the infrared stealth coating.

[0007] In one possible implementation, determining multiple high-risk suspicious points of the high-risk suspicious thermal region based on the high-risk suspicious thermal region includes: acquiring an illumination video of the high-risk suspicious thermal region at a second detection infrared frequency; and determining multiple high-risk suspicious points of the high-risk suspicious thermal region based on the illumination video of the high-risk suspicious thermal region at the second detection infrared frequency.

[0008] In one possible implementation, determining the high-risk suspicious points of the suspicious edge thermal defect lines based on the suspicious edge thermal defect lines includes: acquiring a sequence of suspicious edge thermal defect lines of multiple consecutive frames at the initial infrared frequency during the spraying process; and determining the high-risk suspicious points of the suspicious edge thermal defect lines based on the multiple suspicious edge thermal point information and the sequence of suspicious edge thermal defect lines of multiple consecutive frames at the initial infrared frequency during the spraying process.

[0009] According to a second aspect, the present invention provides a detection system for infrared stealth coatings on transparent aerospace components, comprising: an acquisition module for acquiring an illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be detected; a first determination module for determining multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be detected; a video acquisition module for acquiring an illumination video of the first detection infrared frequency of each suspicious thermal region; and a thermal map construction module for constructing a thermal map, wherein the thermal map includes multiple suspicious thermal regions and multiple edges between the multiple suspicious thermal regions, and the node features of each suspicious thermal region include the illumination video of the initial infrared frequency of each suspicious thermal region and the information of each suspicious thermal region. The first detection module uses infrared frequency illumination video, with the edges between nodes representing distance information to suspicious thermal regions. A first processing module processes the thermal map using a graph neural network to obtain high-risk suspicious thermal regions. A clustering analysis module clusters the multiple suspicious edge thermal points to obtain K clusters. A generation module generates suspicious edge thermal defect lines based on the K clusters. A second determination module determines multiple suspicious intermediate test points based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines; each suspicious intermediate test point is a test point in the region between a suspicious thermal region and a suspicious edge thermal defect line. A result determination module determines the detection result of the infrared stealth coating on the aerospace transparent component based on the high-risk suspicious thermal regions, the suspicious edge thermal defect lines, and the multiple suspicious intermediate test points.

[0010] In one possible implementation, the result determination module is further configured to: determine multiple high-risk suspicious points in the high-risk suspicious thermal area based on the high-risk suspicious thermal area; determine high-risk suspicious points in the suspicious edge thermal defect line based on the suspicious edge thermal defect line; acquire phase-locked thermal images of multiple high-risk suspicious points in the high-risk suspicious thermal area, phase-locked thermal images of high-risk suspicious points in the suspicious edge thermal defect line, and phase-locked thermal images of multiple suspicious intermediate test points; generate a defect probability distribution map of the infrared stealth coating based on the phase-locked thermal images of multiple high-risk suspicious points in the high-risk suspicious thermal area, the phase-locked thermal images of high-risk suspicious points in the suspicious edge thermal defect line, the phase-locked thermal images of the multiple suspicious intermediate test points, the information of the multiple suspicious edge thermal points, the irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be inspected, and the irradiation video of the first detection infrared frequency of each suspicious thermal area; and determine the detection result of the infrared stealth coating of the aviation transparent part based on the defect probability distribution map of the infrared stealth coating.

[0011] In one possible implementation, determining multiple high-risk suspicious points of the high-risk suspicious thermal region based on the high-risk suspicious thermal region includes: acquiring an illumination video of the high-risk suspicious thermal region at a second detection infrared frequency; and determining multiple high-risk suspicious points of the high-risk suspicious thermal region based on the illumination video of the high-risk suspicious thermal region at the second detection infrared frequency.

[0012] In one possible implementation, determining the high-risk suspicious points of the suspicious edge thermal defect lines based on the suspicious edge thermal defect lines includes: acquiring a sequence of suspicious edge thermal defect lines of multiple consecutive frames at the initial infrared frequency during the spraying process; and determining the high-risk suspicious points of the suspicious edge thermal defect lines based on the multiple suspicious edge thermal point information and the sequence of suspicious edge thermal defect lines of multiple consecutive frames at the initial infrared frequency during the spraying process.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring an illumination video of the initial infrared frequency during the spraying of an infrared stealth coating on an aerospace transparent part to be inspected; determining multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the aerospace transparent part to be inspected; acquiring an illumination video of the first detection infrared frequency of each suspicious thermal region; constructing a thermal map, the thermal map including multiple suspicious thermal regions and multiple edges between the multiple suspicious thermal regions, and nodes of each suspicious thermal region. The features include an initial infrared frequency illumination video for each suspicious thermal region, an illumination video of the first detected infrared frequency for each suspicious thermal region, and the edges between nodes representing distance information between suspicious thermal regions. High-risk suspicious thermal regions are obtained by processing the thermal map using a graph neural network. K clusters are obtained based on the multiple suspicious edge thermal point information. Suspicious edge thermal defect lines are generated based on the K clusters. Multiple suspicious intermediate test points are determined based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines. Each suspicious intermediate test point is a test point in the region between a suspicious thermal region and a suspicious edge thermal defect line. The detection results of the infrared stealth coating on the aerospace transparent component are determined based on the high-risk suspicious thermal regions, the suspicious edge thermal defect lines, and the multiple suspicious intermediate test points.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the aforementioned method for detecting infrared stealth coatings on transparent aerospace components. The method includes: acquiring an illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be detected; determining multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be detected; acquiring an illumination video of the first detection infrared frequency of each suspicious thermal region; constructing a thermal map, the thermal map including multiple suspicious thermal regions and multiple edges between the multiple suspicious thermal regions, and the node features of each suspicious thermal region including each The initial infrared frequency illumination video of each suspicious thermal region, the first detection infrared frequency illumination video of each suspicious thermal region, and the edges between nodes represent the distance information of the suspicious thermal regions; high-risk suspicious thermal regions are obtained by processing the thermal map based on a graph neural network; K clusters are obtained based on the multiple suspicious edge thermal point information; suspicious edge thermal defect lines are generated based on the K clusters; multiple suspicious intermediate test points are determined based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines, and each suspicious intermediate test point is a test point in the region between a suspicious thermal region and a suspicious edge thermal defect line; the detection results of the infrared stealth coating of the aviation transparent part are determined based on the high-risk suspicious thermal regions, the suspicious edge thermal defect lines, and the multiple suspicious intermediate test points.

[0015] This invention provides a method and system for detecting infrared stealth coatings on transparent aerospace components. The method includes acquiring an illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component; determining multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component; acquiring an illumination video of the first detection infrared frequency for each suspicious thermal region; and constructing a thermal map, wherein the thermal map includes multiple suspicious thermal regions and multiple edges between the multiple suspicious thermal regions, and the node features of each suspicious thermal region include the illumination video of the initial infrared frequency of each suspicious thermal region and the illumination video of the first detection infrared frequency of each suspicious thermal region. The method involves: using the edges between nodes to represent distance information of suspected thermal regions; processing the thermal map using a graph neural network to obtain high-risk suspected thermal regions; clustering the multiple suspected edge thermal points to obtain K clusters; generating suspected edge thermal defect lines based on the K clusters; determining multiple suspected intermediate test points based on the multiple suspected thermal regions and the suspected edge thermal defect lines, where each suspected intermediate test point is a test point in the region between a suspected thermal region and a suspected edge thermal defect line; and determining the detection results of the infrared stealth coating on aerospace transparent parts based on the high-risk suspected thermal regions, the suspected edge thermal defect lines, and the multiple suspected intermediate test points. This method can efficiently and accurately identify global defects in the infrared stealth coating of aerospace transparent parts and quantify defect risks. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a method for detecting infrared stealth coatings on transparent aerospace components, provided in an embodiment of the present invention.

[0017] Figure 2 A schematic diagram of an aerospace transparent component provided in an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of an infrared imaging device provided in an embodiment of the present invention;

[0019] Figure 4 A schematic flowchart illustrating the detection results of infrared stealth coating on aerospace transparent parts, provided for an embodiment of the present invention;

[0020] Figure 5 This invention provides a schematic flowchart for determining multiple high-risk and suspicious points in a high-risk and suspicious thermal area.

[0021] Figure 6 This invention provides a schematic flowchart for determining high-risk suspicious points on suspicious edge thermal defect lines.

[0022] Figure 7This is a schematic diagram of a detection system for infrared stealth coatings on transparent aerospace components, provided in an embodiment of the present invention. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0024] In this embodiment of the invention, the following are provided: Figure 1 The method for detecting infrared stealth coatings on transparent aerospace components includes steps S1 to S9:

[0025] Step S1: Obtain the initial infrared frequency irradiation video of the infrared stealth coating being applied to the transparent aerospace component to be tested.

[0026] Aviation transparent components are transparent structural parts installed on aircraft to provide a field of vision or optical windows, such as cockpit windshields, nose cones, or optoelectronic windows. Aviation transparent components are made of high-strength optical materials. Figure 2 A schematic diagram of an aerospace transparent component provided in an embodiment of the present invention;

[0027] The video of the initial infrared frequency during the application of the infrared stealth coating on the transparent aviation component under test is a video of the application process obtained by an infrared imaging device during the application of the infrared stealth coating on the transparent aviation component. The video is generated by irradiating the transparent aviation component with the initially set infrared frequency. Figure 3 This is a schematic diagram of an infrared imaging device provided in an embodiment of the present invention.

[0028] The video of the initial infrared frequency irradiation during the application of infrared stealth coating on the transparent aviation component under test can record the thermal radiation distribution of the coating in the initial stage of application and the infrared response state that changes over time.

[0029] Step S2: Based on the irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be tested, determine multiple suspicious thermal areas, multiple suspicious edge thermal point information, and the first detection infrared frequency.

[0030] In some embodiments, an illumination analysis model can be used to determine multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency. The illumination analysis model is a recurrent neural network. The input to the illumination analysis model is an illumination video of the initial infrared frequency during the application of the infrared stealth coating to the aircraft transparent part to be inspected, and the output of the illumination analysis model is multiple suspicious thermal regions, multiple suspicious edge thermal point information, and the first detection infrared frequency.

[0031] Recurrent Neural Networks (RNNs) are deep learning models used to process sequential data. Their core structure includes a recursive update mechanism for hidden states and effectively captures temporal dependencies. In video analysis tasks, RNNs can model dynamic changes between frames and continuously learn the patterns of dynamic change in the data by progressively processing each frame's data and passing state information.

[0032] Multiple suspicious thermal areas were identified by an irradiation analysis model as areas with abnormal infrared thermal distributions and potential coating defects.

[0033] Multiple suspicious thermal areas indicate the locations and ranges where there may be uneven coating thickness or material defects.

[0034] Multiple suspicious edge thermal points were identified by the irradiation analysis model as points located at the edge of the coating and exhibiting abnormal thermal behavior.

[0035] Suspicious edge thermal point information includes the point's pixel coordinates, the curve of thermal value change with spraying time, thermal diffusion radius, the offset distance between the point and the coating edge baseline, and the difference between the thermal value and the surrounding normal area.

[0036] Suspicious edge thermal points are specific locations where defects frequently occur, particularly at the edges of the coating. As the boundary between coating processes, the edges are prone to defects due to uneven adhesion, drying differences, and other factors, often manifesting as localized abnormal heat dissipation. Identifying suspicious edge thermal points can prevent omissions or misjudgments caused by confusion with the main body area and provides accurate data for subsequent precise analysis of edge defects.

[0037] The first detection infrared frequency is determined by the illumination analysis model and used for subsequent targeted detection of suspicious thermal areas.

[0038] The video of the initial infrared frequency during the application of the infrared stealth coating on the transparent aerospace component under inspection provides a complete thermal evolution process of the coating in its initial state. The illumination analysis model can analyze the temperature change trend of pixels in the video over time, thereby identifying abnormal areas and edge points that do not conform to the thermal characteristics of normal spraying, and inferring the first detection infrared frequency that is more suitable for capturing such defects based on the abnormal characteristics.

[0039] Recurrent neural networks (RNNs) can process video illumination at initial infrared frequencies frame by frame by memorizing time-series data. They can focus on the temporal dimension of temperature changes in the video to identify dynamic thermal anomaly patterns, thereby accurately locating multiple suspicious thermal regions and edge thermal points, and calculating the most sensitive first detection infrared frequency.

[0040] Step S3: Obtain the irradiation video of the first detection infrared frequency for each suspected thermal area.

[0041] The video of each suspected thermal area being illuminated at the first detection infrared frequency is obtained by using an infrared camera to capture each suspected thermal area individually at the first detection infrared frequency.

[0042] High-resolution thermal details of each suspected thermal area can be recorded in the video of the first detection infrared frequency of each area at a specific sensitive frequency.

[0043] Step S4: Construct a heat map. The heat map includes multiple suspected heat regions and multiple edges between the suspected heat regions. The node features of each suspected heat region include the illumination video of the initial infrared frequency of each suspected heat region and the illumination video of the first detection infrared frequency of each suspected heat region. The edges between the nodes are the distance information of the suspected heat regions.

[0044] A thermal map is a graphical data structure used to describe thermally anomalous regions on a coating surface and their interrelationships. Thermal maps can systematically represent the topological connections and attribute associations between spatially dispersed suspicious regions.

[0045] The thermal map uses each suspected thermal region as a node and the distance information between suspected thermal regions as the edges between nodes. The node features integrate the illumination video of the initial infrared frequency and the illumination video of the first detected infrared frequency of the corresponding suspected thermal region.

[0046] The initial infrared frequency irradiation video of the suspected thermal area refers to the video of the portion corresponding to a single suspected thermal area in the initial infrared frequency irradiation video during the spraying of the infrared stealth coating on the transparent aerospace component to be inspected.

[0047] By constructing a heat map, the relevant features and inter-regional distance correlations of multiple suspected heat regions can be integrated into a unified map structure, which facilitates subsequent analysis by graph neural networks.

[0048] Step S5: Process the thermal map based on a graph neural network to obtain high-risk and suspected thermal regions.

[0049] Graph Neural Networks (GNNs) are deep learning models that can run directly on graphs. GNNs capture graph dependencies by passing messages between nodes. They can aggregate information from neighboring nodes to update the representation of the current node, thus combining the graph's topology and node features. GNNs can handle non-Euclidean spatial data and effectively mine spatial associations and feature correlations between nodes.

[0050] High-risk suspected thermal areas are identified by graph neural networks as areas with a very high probability of coating defects among multiple suspected thermal areas.

[0051] Nodes in a heatmap represent multiple suspicious thermal regions, and edges represent the distance relationships between these regions. Each suspicious thermal region node possesses rich node features, namely, the illumination video of the initial infrared frequency and the illumination video of the first detected infrared frequency for each suspicious thermal region. Graph neural networks excel at processing complex heatmap data. By aggregating the features of a node itself and those of its neighboring nodes, graph neural networks can achieve deep extraction and updating of node features, effectively capturing complex relationships between nodes. This processing method preserves the core thermal features of suspicious thermal regions and the inter-regional correlation information. Heatmaps provide complete node and edge feature information, and graph neural networks are capable of processing this type of heatmap data, extracting high-risk related features to identify high-risk suspicious thermal regions.

[0052] Graph neural networks (GNNs) can analyze the spatial distribution patterns and feature interactions between different suspicious heatmaps by aggregating neighborhood information of each node. Through graph convolution operations, GNNs can combine local video features with the global topology to calculate the risk weight of each node, ultimately identifying high-risk suspicious heatmaps.

[0053] Step S6: Cluster the multiple suspicious edge heat points to obtain K clusters.

[0054] The clustering method described is K-means clustering, an iterative clustering analysis algorithm. K-means clustering divides data into K categories using a pre-defined value of K. The core idea of ​​K-means clustering is to minimize the sum of the squared distances from each point within a cluster to the cluster center, thereby achieving high similarity within clusters and low similarity between clusters.

[0055] The K clusters are K sets of data obtained by grouping multiple suspicious edge heat points using the K-means clustering algorithm based on features such as the similarity of heat point location, the difference of heat peak value, and the similarity of heat change curve trends. Each cluster contains multiple suspicious edge heat point information with similar features.

[0056] The process of clustering multiple suspicious edge heat points using the K-means clustering algorithm is as follows: First, K suspicious edge heat points are randomly selected as initial cluster centers. Then, the comprehensive distance between each suspicious edge heat point and each initial cluster center is calculated in multiple feature dimensions, such as three-dimensional spatial coordinates and the curve of heat value change with spraying time, so that each heat point is assigned to the cluster containing the cluster center with the closest comprehensive distance. Next, the mean of all heat points in each cluster in each feature dimension is recalculated to determine new cluster centers. The above allocation and update process is repeated until the change of cluster centers is less than a preset threshold, and finally K clusters are obtained.

[0057] By clustering multiple suspicious edge heatmaps into several groups, suspicious edge heatmaps with similar characteristics can be categorized and integrated, thus forming groups with clear feature correlations from scattered heatmap information. This clustering method can highlight the common features and differences between different categories of heatmaps, reduce the interference of isolated outliers on subsequent analysis, simplify data dimensions, focus on core suspicious features, and thus provide a structured and targeted data foundation for subsequent generation of suspicious edge heatmap defect lines based on clusters.

[0058] Step S7: Generate suspicious edge thermal defect lines based on the K clusters.

[0059] In some embodiments, a defect determination model can be used to generate suspected edge thermal defect lines. The defect determination model is a deep neural network. The input to the defect determination model is the K clusters, and the output of the defect determination model is the suspected edge thermal defect line.

[0060] Deep neural networks (DNNs) are neural network models composed of multiple hidden layers. Through their deep network structure, DNNs can extract and abstract data features layer by layer, and can mine complex high-order features from raw data. The neurons in each layer of a deep neural network are connected by weights, and the weight parameters are continuously adjusted through the backpropagation algorithm, thereby achieving accurate learning and prediction of the target task.

[0061] Suspicious edge thermal defect lines are lines that characterize potential defects in the edge portion of the infrared stealth coating of aerospace transparent parts, obtained by fitting the data through deep neural network analysis based on the information of suspicious edge thermal points in K clusters.

[0062] Suspicious edge thermal defect lines can intuitively characterize the spatial distribution and extension trend of coating edge defects, and provide accurate geometric reference for subsequent location of high-risk edge defect points and delineation of defect-related areas.

[0063] K clusters are generated by clustering suspicious edge thermal points with similar characteristics. They integrate common features such as location from the thermal point and thermal changes, and can also eliminate interference from isolated outliers. The spatial connection relationship between clusters and the progressive law of thermal features enable the model to fit local defect segments and connect them to form complete lines, accurately depicting the extension trajectory of edge defects.

[0064] Deep neural networks can construct a feature matrix for each cluster. Then, through convolutional and pooling layers, the feature matrix is ​​compressed in dimension and enhanced with key information to uncover the spatial distribution patterns of thermal points within a cluster and the synergistic laws of thermal characteristics. Simultaneously, deep neural networks utilize fully connected layers to establish inter-cluster connections and analyze the spatial connections and progressive trends of thermal characteristics between different clusters, thereby determining whether potential defect extension paths exist between clusters. Based on intra-cluster clustering patterns and inter-cluster connections, the deep neural network's regression prediction module can perform curve fitting on thermal points within each cluster, generating local defect segments that conform to the cluster's characteristics. Finally, the model integrates all local defect segments corresponding to all clusters, performing smooth connections and optimization corrections based on inter-cluster connections, while eliminating contradictory segments and completing broken nodes, ultimately forming continuous and accurate lines identifying suspicious edge thermal defects.

[0065] Step S8: Based on the multiple suspected thermal regions and the suspected edge thermal defect lines, multiple suspected intermediate test points are determined. Each suspected intermediate test point is a test point in the region between a suspected thermal region and a suspected edge thermal defect line.

[0066] In some embodiments, a test point determination model can be used to identify multiple suspicious intermediate test points. The test point determination model is a Transformer model. The inputs to the test point determination model are the multiple suspicious thermal regions and the suspicious edge thermal defect lines, and the output of the test point determination model is the multiple suspicious intermediate test points.

[0067] The Transformer model is a deep learning model based on the self-attention mechanism. Through this attention mechanism, the Transformer model can capture the global dependencies between elements in the input sequence. It includes an encoder and decoder structure and can process data in parallel. The Transformer model performs exceptionally well in handling long-range dependencies and complex contextual relationships.

[0068] Suspicious intermediate test points are specific detection locations located in the transition area between the suspicious thermal region and the suspicious edge thermal defect line, determined by the test point determination model.

[0069] Suspicious intermediate test points are key detection locations in the transitional areas between multiple suspicious thermal regions and suspicious edge thermal defect lines, filling the spatial monitoring gap between these two core detection objects. Suspicious intermediate test points, along with suspicious thermal regions and suspicious edge thermal defect lines, form a complete detection chain, thereby improving the overall coverage and accuracy of infrared stealth coating defect detection for aerospace transparent parts.

[0070] Multiple suspicious thermal regions mark the core potential areas where defects may exist, and suspicious edge thermal defect lines clarify the characterization trend of potential defects at the coating edge. Multiple suspicious thermal regions and suspicious edge thermal defect lines define the key spatial range where defects may spread and associate, enabling the model to focus on the correlation between multiple suspicious thermal regions and suspicious edge thermal defect lines to accurately determine multiple suspicious intermediate test points. This ensures that the layout of multiple suspicious intermediate test points can conform to the defect distribution logic and also fully cover key risk areas.

[0071] The Transformer model, relying on a self-attention mechanism, can simultaneously extract and analyze features related to the spatial coordinate range of multiple suspicious thermal regions, the core location of thermal anomalies, and the extension trajectory and curvature variation characteristics of suspicious edge thermal defect lines. The model calculates the Euclidean distance between the geometric center of each suspicious thermal region and each sampling point on the defect line, and combines this with the coupling coefficient between the intensity of the regional thermal anomaly and the thermal characteristics of the defect line, thereby quantifying the spatial correlation strength between the region and the line. Subsequently, the model can identify transitional regions where the correlation strength falls within a preset range, and accurately determine the positions with the highest spatial correlation strength in these regions as multiple suspicious intermediate test points.

[0072] In some embodiments, determining multiple suspicious intermediate test points based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines includes steps S21 to S23:

[0073] Step S21: Based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines, determine the potential diffusion path information of defects, the key segments covered by the path, and the thermal gradient extreme value zone information of multiple intermediate transition regions.

[0074] In some embodiments, deep neural networks can be used to determine information on potential defect diffusion paths, key segments covered by the paths, and thermal gradient extremum zones in multiple intermediate transition regions.

[0075] Defect potential propagation path information is the path information that a defect may propagate from a suspected thermal region to the edge defect line, determined by a deep neural network. This information includes the coordinates of the path's direction, the coordinates of key turning points, and its extent.

[0076] The potential diffusion path of a defect can intuitively reflect the likely diffusion trajectory of the defect.

[0077] The critical segment of the path coverage is the core segment that is most likely to become the inevitable path for defect propagation in the potential defect propagation path and is crucial for defect monitoring, as determined by a deep neural network.

[0078] The information on the extreme value zone of the thermal gradient in the intermediate transition region is determined by a deep neural network. It identifies the zone with the most dramatic changes in thermal gradient within the transition area between multiple suspected thermal regions and suspected edge thermal defect lines. This information includes the location and extent of the extreme value zone, as well as the magnitude of the thermal change.

[0079] The extreme value zone of the thermal gradient in the intermediate transition region is a key area where defects are prone to grow or spread.

[0080] Deep neural networks can extract features such as the spatial location and thermal intensity distribution of multiple suspicious thermal regions, as well as features such as the extension trajectory and key inflection points of suspicious edge thermal defect lines. Then, by constructing the spatial association logic between multiple suspicious thermal regions and suspicious edge thermal defect lines, the logic of defect propagation path can be deduced, the core segment of the path can be located, and the extreme value distribution of thermal gradient in the transition zone can be identified. Finally, the model can derive information on the potential diffusion path of defects, the key segments covered by the path, and the extreme value zone of thermal gradient in multiple intermediate transition areas.

[0081] Step S22: Based on the potential diffusion path information of the defects, the key segments covered by the path, and the thermal gradient extreme value zone information of the multiple intermediate transition regions, determine multiple key monitoring transition zones, the risk level of each key monitoring transition zone, the correlation between each key monitoring transition zone and adjacent suspicious thermal regions, the edge radiation coverage of each key monitoring transition zone and adjacent suspicious edge thermal defect lines, and the similarity between each key monitoring transition zone and each adjacent key monitoring transition zone.

[0082] In some embodiments, deep neural networks may be used to determine multiple key monitoring transition zones, the risk level of each key monitoring transition zone, the correlation between each key monitoring transition zone and adjacent suspected thermal areas, the edge radiation coverage of each key monitoring transition zone and adjacent suspected edge thermal defect lines, and the similarity between each key monitoring transition zone and each adjacent key monitoring transition zone.

[0083] Multiple key monitoring transition zones are highly correlated areas of defect propagation identified from intermediate transition zones using deep neural networks.

[0084] The risk level of each key monitoring transition zone is a quantitative index determined by a deep neural network, representing the degree of defect risk in each key monitoring transition zone.

[0085] The correlation between each key monitoring transition zone and its adjacent suspected thermal areas is a numerical indicator determined by a deep neural network, which measures the degree of correlation between a single key monitoring transition zone and its nearest suspected thermal area.

[0086] The edge radiation coverage of each key monitoring transition zone and its adjacent suspected edge thermal defect line is a numerical indicator determined by a deep neural network, which measures the completeness of coverage of a single key monitoring transition zone within the radiation range affected by the corresponding adjacent suspected edge thermal defect line.

[0087] The similarity between each key monitoring transition zone and each adjacent key monitoring transition zone is a numerical indicator determined by a deep neural network, which measures the degree of overlap in core features between a single key monitoring transition zone and other adjacent key monitoring transition zones.

[0088] Deep neural networks can learn the directional characteristics of potential defect diffusion paths, the core location features of key segments covered by the paths, and the distribution characteristics of thermal gradient extreme zones in multiple intermediate transition areas, while also uncovering the inherent relationships between these information. By modeling the mapping logic between input information and target output, and quantifying the defect risk of each transition zone, its association attributes with adjacent suspicious thermal areas and adjacent suspicious edge thermal defect lines, as well as the degree of overlap of inter-regional features, deep neural networks can accurately deduce multiple key monitoring transition zones, the risk level of each key monitoring transition zone, the correlation between each key monitoring transition zone and adjacent suspicious thermal areas, the edge radiation coverage of each key monitoring transition zone and adjacent suspicious edge thermal defect lines, and the similarity between each key monitoring transition zone and each adjacent key monitoring transition zone.

[0089] Step S23: Based on the multiple key monitoring transition zones, the risk level of each key monitoring transition zone, the correlation between each key monitoring transition zone and adjacent suspicious thermal areas, the edge radiation coverage of each key monitoring transition zone and adjacent suspicious edge thermal defect lines, and the similarity between each key monitoring transition zone and each adjacent key monitoring transition zone, multiple suspicious intermediate test points are determined.

[0090] In some embodiments, a deep neural network can be used to identify multiple suspicious intermediate test points.

[0091] Deep neural networks, through multi-layer perception and feature fusion mechanisms, can accurately capture the spatial location attributes of multiple key monitoring transition zones, the quantitative indicators of the risk level of each key monitoring transition zone, the degree of correlation between each key monitoring transition zone and adjacent suspected thermal areas, the edge radiation coverage status of each key monitoring transition zone and adjacent suspected edge thermal defect lines, and the feature overlap between each key monitoring transition zone and its neighbors. The model can deeply analyze the inherent correlation logic of various input information. By learning the reasonable distribution patterns of suspected intermediate test points under different input conditions, deep neural networks can establish a precise correspondence between input information and target point selection, thereby enabling the selection of representative core points that can comprehensively cover monitoring needs.

[0092] Step S9: Determine the detection results of the infrared stealth coating of the aviation transparent part based on the high-risk suspected thermal area, the suspected edge thermal defect line, and the multiple suspected intermediate test points.

[0093] In some embodiments, Figure 4 This is a flowchart illustrating the process for determining the detection result of the infrared stealth coating on an aerospace transparent component, as provided in an embodiment of the present invention. The determination of the detection result includes steps S31 to S35:

[0094] Step S31: Based on the high-risk and suspicious thermal areas, determine multiple high-risk and suspicious points in the high-risk and suspicious thermal areas.

[0095] In some embodiments, Figure 5 This invention provides a flowchart illustrating the process of determining multiple high-risk and suspicious points within a high-risk and suspicious thermal area, comprising steps S41-S42:

[0096] Step S41: Obtain the irradiation video of the high-risk, suspicious thermal area at the second detection infrared frequency.

[0097] The second detection infrared frequency is an infrared band frequency for secondary re-examination of high-risk areas, obtained by weighting the initially set infrared frequency and the first detection infrared frequency.

[0098] The video of the high-risk and suspected thermal area being illuminated at the second detection infrared frequency is obtained by using an infrared camera to capture the high-risk and suspected thermal area at the second detection infrared frequency.

[0099] Step S42: Based on the illumination video of the high-risk and suspicious thermal area at the second detection infrared frequency, determine multiple high-risk and suspicious points of the high-risk and suspicious thermal area.

[0100] In some embodiments, a suspicious point determination model can be used to identify multiple high-risk suspicious points within a high-risk suspicious thermal region. The suspicious point determination model is a recurrent neural network. The input to the suspicious point determination model is the video of the high-risk suspicious thermal region being illuminated at a second detection infrared frequency, and the output of the suspicious point determination model is multiple high-risk suspicious points within the high-risk suspicious thermal region.

[0101] High-risk and suspicious points are the pixel locations with the most significant thermal anomalies and the highest defect risk within the high-risk and suspicious thermal areas identified by the suspicious point determination model after analyzing the video of the high-risk and suspicious thermal areas irradiated at the second detection infrared frequency.

[0102] The video recording of the high-risk, suspicious thermal area underwent illumination at the second detection infrared frequency, revealing subtle thermal dynamic changes at specific sensitive frequencies. A recurrent neural network (RNN), leveraging its ability to memorize time-series data, can process the video frame-by-frame. The RNN analyzes minute fluctuations and abnormal diffusion patterns of thermal values ​​over time within the video, thereby capturing transient or persistent thermal instabilities within the high-risk area and ultimately pinpointing multiple high-risk, suspicious points.

[0103] Step S32: Based on the suspected edge thermal defect line, determine the high-risk suspected points of the suspected edge thermal defect line.

[0104] In some embodiments, Figure 6 This is a schematic flowchart illustrating a process for determining high-risk suspicious points of a suspected edge thermal defect line according to an embodiment of the present invention. The determination of high-risk suspicious points of a suspected edge thermal defect line includes steps S51-S52:

[0105] Step S51: Obtain the sequence information of suspicious edge thermal defect lines in multiple consecutive frames of the initial infrared frequency during the spraying process.

[0106] The sequence information of suspicious edge thermal defect lines in multiple consecutive frames at the initial infrared frequency during the spraying process is the thermal change data of the corresponding suspicious edge thermal defect line positions in multiple consecutive frames of video taken at the initial infrared frequency during the coating spraying process.

[0107] The information on the suspected edge thermal defect line sequence includes the change in the average thermal intensity of the suspected edge thermal defect line, the fluctuation range of the difference between the thermal value and the average value of the surrounding normal area, and the degree of continuous coverage of the thermal anomaly signal of the suspected edge thermal defect line.

[0108] The degree of continuous coverage of thermal anomaly signals is the ratio of the total length of continuous line segments whose thermal values ​​exceed the threshold of the normal area within the extension range of the suspected edge thermal defect line to the total length of the defect line.

[0109] Step S52: Based on the multiple suspicious edge thermal point information and the sequence information of suspicious edge thermal defect lines of multiple consecutive frames of the initial infrared frequency during the spraying process, determine the high-risk suspicious points of the suspicious edge thermal defect lines.

[0110] In some embodiments, a defect sequence analysis model can be used to identify high-risk suspicious points of suspicious edge thermal defect lines. The defect sequence analysis model is a Transformer model, and its input includes the information of the plurality of suspicious edge thermal points and the sequence information of suspicious edge thermal defect lines from multiple consecutive frames at the initial infrared frequency during the spraying process. The output of the defect sequence analysis model is the high-risk suspicious points of the suspicious edge thermal defect lines.

[0111] High-risk suspicious points on suspicious edge thermal defect lines are specific locations on suspicious edge thermal defect lines that are identified by defect sequence analysis models and have significant infrared thermal anomalies and a very high probability of having coating defects.

[0112] Multiple suspicious edge thermal point information includes core features such as pixel coordinates and thermal value variation curves over spraying time, accurately reflecting the basic attributes of edge anomaly points. The sequence of multiple consecutive frames of suspicious edge thermal defect lines at the initial infrared frequency during spraying reveals the dynamic evolution of the defect lines. The Transformer model, through its self-attention mechanism, can simultaneously focus on the spatial coordinates and thermal features of multiple suspicious edge thermal point information, as well as the dynamic evolution of the suspicious edge thermal defect line sequence information. The Transformer model can capture the correlation between suspicious edge thermal point information and the sequence of suspicious edge thermal defect lines at the initial infrared frequency during spraying, thereby analyzing the intensity, continuity, and trend of thermal anomalies at various locations on the defect lines. This allows for the screening of specific points with significant thermal anomalies and extremely high defect risk, ultimately identifying high-risk suspicious points along the suspicious edge thermal defect lines.

[0113] Step S33: Obtain phase-locked thermal images of multiple high-risk and suspicious points in the high-risk and suspicious thermal area, phase-locked thermal images of high-risk and suspicious points on suspicious edge thermal defect lines, and phase-locked thermal images of multiple suspicious intermediate test points.

[0114] Phase-locked thermal imaging of multiple high-risk and suspicious points in a high-risk and suspicious thermal area is phase map and amplitude map imaging data obtained by applying periodic heat source excitation to multiple high-risk and suspicious points in a high-risk and suspicious thermal area using a phase-locked infrared thermal imager.

[0115] Phase-locked thermal imaging of high-risk and suspicious points on suspicious edge thermal defect lines is phase and amplitude imaging data obtained by applying periodic heat source excitation to high-risk and suspicious points on suspicious edge thermal defect lines using a phase-locked infrared thermal imager.

[0116] Phase-locked thermal imaging of suspicious intermediate test points is obtained by using a phase-locked infrared thermal imager to excite the suspicious intermediate test points with a periodic heat source, and the resulting phase and amplitude images are obtained.

[0117] Lock-in thermal imaging, through periodic heat source excitation and phase amplitude analysis, can more accurately capture multiple high-risk and suspicious points in high-risk and suspicious thermal areas, high-risk and suspicious points on suspicious edge thermal defect lines, and microscopic thermal anomaly features of multiple suspicious intermediate test points. Moreover, its detection accuracy is significantly higher than that of conventional infrared imaging, thus effectively identifying thermal response differences caused by minute defects inside the coating.

[0118] Step S34: Based on phase-locked thermal imaging of multiple high-risk and suspicious points in the high-risk and suspicious thermal areas, phase-locked thermal imaging of high-risk and suspicious points in the suspicious edge thermal defect lines, phase-locked thermal imaging of multiple suspicious intermediate test points, information on multiple suspicious edge thermal points, irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be inspected, and irradiation video of the first detection infrared frequency of each suspicious thermal area, a defect probability distribution map of the infrared stealth coating is generated.

[0119] In some embodiments, a defect probability calculation model can be used to generate a defect probability distribution map of the infrared stealth coating. The defect probability calculation model is a generative adversarial network (GAN). The inputs to the model are phase-locked thermal images of multiple high-risk suspicious points in the high-risk suspicious thermal region, phase-locked thermal images of high-risk suspicious points in the suspicious edge thermal defect line, phase-locked thermal images of multiple suspicious intermediate test points, information on multiple suspicious edge thermal points, an illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the aerospace transparent part to be inspected, and an illumination video of the first detection infrared frequency for each suspicious thermal region. The output of the defect probability calculation model is the defect probability distribution map of the infrared stealth coating.

[0120] Generative Adversarial Networks (GANs) are deep learning models consisting of a generator and a discriminator. The generator is responsible for capturing the data distribution and generating new data samples, while the discriminator is responsible for distinguishing whether the input data is real or generated by the generator. Through game-like training between the generator and the discriminator, GANs can learn the complex distribution characteristics of the data and generate high-quality, highly realistic results.

[0121] The defect probability distribution map of infrared stealth coating is an image that is output by the defect probability calculation model and can intuitively show the probability of defects existing at different locations on the infrared stealth coating of aviation transparent parts. Each location in the image corresponds to a defect probability value.

[0122] Phase-locked thermal imaging data can accurately capture the phase shift and amplitude variation characteristics of thermal radiation caused by minute defects inside the coating at various points. Information on multiple suspicious edge thermal points includes pixel coordinates, thermal value variation curves over spraying time, and other data, thus clarifying the spatial location and dynamic thermal evolution of abnormal edge points. The initial infrared frequency illumination video during the spraying of the infrared stealth coating on the aerospace transparent component under inspection fully records the overall thermal radiation distribution and thermal response process of the coating in the initial stage of spraying. The first detection infrared frequency illumination video of each suspicious thermal area can clearly present the local high-resolution thermal details of each suspicious area at the sensitive detection frequency. These input data provide comprehensive feature data related to coating defects from specific dimensions such as micro-defect thermal response, spatial and dynamic characteristics of points, overall thermal evolution of the coating, and fine thermal performance of local areas, providing data basis for the accurate analysis of defect risks at various locations using a defect probability calculation model.

[0123] The generator in a generative adversarial network (GAN) can perform feature fusion on various input phase-locked thermal images, thermal videos, and location information, learn the mapping relationship between coating defects and thermal features, and generate preliminary coating defect probability distribution data. The discriminator can continuously optimize the accuracy of the generated results, and through adversarial learning mechanisms, mine the mapping relationship between deep defect features and surface thermal anomalies from complex infrared thermal images and video data, ultimately enabling the generator to generate a high-precision defect probability distribution map of the infrared stealth coating.

[0124] Step S35: Determine the detection result of the infrared stealth coating of the aviation transparent part based on the defect probability distribution map of the infrared stealth coating.

[0125] In some embodiments, a quality assessment model can be used to determine the detection results of the infrared stealth coating on the aerospace transparent component. The quality assessment model is a deep neural network. The input to the quality assessment model is the defect probability distribution map of the infrared stealth coating, and the output of the quality assessment model is the detection result of the infrared stealth coating on the aerospace transparent component.

[0126] The inspection results for the infrared stealth coating of transparent aerospace components are determined by analyzing the defect probability distribution map of the infrared stealth coating using a quality assessment model. The inspection results for the infrared stealth coating of transparent aerospace components include both pass and fail categories.

[0127] Deep neural networks, leveraging their multi-layered feature extraction and logical reasoning capabilities, can accurately transform coating defect probability distribution maps into judgment results. The defect probability distribution map of an infrared stealth coating quantifies the spatial distribution of defect risks on the coating surface in two-dimensional data. The deep neural network receives the coating defect probability distribution map data through its input layer and then utilizes multiple hidden layers to extract features and recognize patterns in high-probability defect areas within the distribution map. The deep neural network can analyze the geometry, connectivity, and cumulative probability density of high-risk areas to determine whether the severity of the defects will impair infrared stealth performance. The model uses fully connected layers to map the extracted complex two-dimensional distribution features to a binary classification space and calculates the final classification confidence based on the learned judgment logic. If the overall quality characteristics shown by the coating defect probability distribution map meet the preset quality standards for infrared stealth coatings on aerospace transparent parts, the deep neural network will classify it as qualified; otherwise, it will classify it as unqualified.

[0128] Based on the same inventive concept Figure 7 This is a schematic diagram of a detection system for infrared stealth coatings on transparent aerospace components, provided by an embodiment of the present invention. The detection system includes:

[0129] The acquisition module 61 is used to acquire the initial infrared frequency irradiation video of the infrared stealth coating being sprayed on the transparent aviation part to be inspected.

[0130] The first determining module 62 is used to determine multiple suspicious thermal areas, multiple suspicious edge thermal point information, and the first detection infrared frequency based on the irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be detected.

[0131] The video acquisition module 63 is used to acquire the irradiation video of the first detection infrared frequency for each suspicious thermal area;

[0132] The graph construction module 64 is used to construct a heat map. The heat map includes multiple suspicious heat regions and multiple edges between the multiple suspicious heat regions. The node features of each suspicious heat region include the illumination video of the initial infrared frequency of each suspicious heat region and the illumination video of the first detection infrared frequency of each suspicious heat region. The edges between the nodes are the distance information of the suspicious heat regions.

[0133] The first processing module 65 is used to process the thermal map based on a graph neural network to obtain high-risk and suspected thermal regions.

[0134] Clustering analysis module 66 is used to cluster the multiple suspicious edge heat points to obtain K clusters;

[0135] Generation module 67 is used to generate suspicious edge thermal defect lines based on the K clusters;

[0136] The second determining module 68 is used to determine multiple suspicious intermediate test points based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines. The suspicious intermediate test points are test points in the region between each suspicious thermal region and the suspicious edge thermal defect lines.

[0137] The result determination module 69 is used to determine the detection results of the infrared stealth coating of the aviation transparent part based on the high-risk suspected thermal area, the suspected edge thermal defect line, and the multiple suspected intermediate test points.

[0138] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for detecting infrared stealth coatings on transparent aerospace components, characterized in that, include: Acquire the initial infrared frequency irradiation video during the application of infrared stealth coating to the transparent aerospace component to be inspected; Based on the illumination video of the initial infrared frequency during the application of the infrared stealth coating on the transparent aerospace component to be inspected, multiple suspicious thermal regions, multiple suspicious edge thermal points, and a first detection infrared frequency are determined. This determination includes using an illumination analysis model to identify the multiple suspicious thermal regions, multiple suspicious edge thermal points, and the first detection infrared frequency. The illumination analysis model is a recurrent neural network, and the input to the illumination analysis model is the initial infrared frequency during the application of the infrared stealth coating on the transparent aerospace component to be inspected. The irradiation video of the frequency, the output of the irradiation analysis model is multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency. The first detection infrared frequency is the infrared frequency determined by the irradiation analysis model for subsequent targeted detection of suspicious thermal regions. The irradiation video of the initial infrared frequency of the infrared stealth coating sprayed on the aviation transparent part to be tested provides the complete thermal evolution process of the coating in the initial state. The irradiation analysis model can analyze the temperature change trend of the pixels in the video over time, thereby identifying abnormal regions and edge points that do not match the normal spraying thermal characteristics, and inferring the first detection infrared frequency that is more suitable for capturing such defects based on the abnormal characteristics. Acquire the first detection infrared frequency irradiation video of each suspected thermal area; A heat map is constructed, which includes multiple suspected heat regions and multiple edges between the suspected heat regions. The node features of each suspected heat region include the illumination video of the initial infrared frequency of each suspected heat region and the illumination video of the first detected infrared frequency of each suspected heat region. The edges between the nodes are the distance information of the suspected heat regions. High-risk and suspected thermal regions are obtained by processing the thermal map using a graph neural network. K clusters are obtained based on the information of the multiple suspicious edge heat points; Based on the K clusters, generate suspicious edge thermal defect lines; Based on the multiple suspected thermal regions and the suspected edge thermal defect lines, multiple suspected intermediate test points are determined. Each suspected intermediate test point is a test point in the region between a suspected thermal region and a suspected edge thermal defect line. The detection results of the infrared stealth coating on the aviation transparent parts are determined based on the high-risk suspected thermal areas, the suspected edge thermal defect lines, and the multiple suspected intermediate test points.

2. The detection method for infrared stealth coating on aerospace transparent parts as described in claim 1, characterized in that, The detection results for determining the infrared stealth coating of the aerospace transparent component based on the high-risk suspected thermal areas, the suspected edge thermal defect lines, and the multiple suspected intermediate test points include: Based on the aforementioned high-risk and suspicious thermal areas, multiple high-risk and suspicious points within the high-risk and suspicious thermal areas were identified; Based on the suspected edge thermal defect lines, high-risk suspected points of the suspected edge thermal defect lines are identified; Acquire phase-locked thermal images of multiple high-risk and suspicious points in high-risk and suspicious thermal areas, phase-locked thermal images of high-risk and suspicious points on suspicious edge thermal defect lines, and phase-locked thermal images of multiple suspicious intermediate test points; Based on phase-locked thermal imaging of multiple high-risk and suspicious points in the high-risk and suspicious thermal areas, phase-locked thermal imaging of high-risk and suspicious points in the suspicious edge thermal defect lines, phase-locked thermal imaging of multiple suspicious intermediate test points, information on multiple suspicious edge thermal points, irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be inspected, and irradiation video of the first detection infrared frequency of each suspicious thermal area, a defect probability distribution map of the infrared stealth coating is generated. The detection results of the infrared stealth coating on the aerospace transparent parts are determined based on the defect probability distribution map of the infrared stealth coating.

3. The detection method for infrared stealth coatings on transparent aerospace components as described in claim 2, characterized in that, The multiple high-risk and suspicious points identified based on the high-risk and suspicious thermal areas include: Acquire video of high-risk, suspicious thermal areas being irradiated at the second detection infrared frequency; Based on the irradiation video of the high-risk and suspicious thermal area at the second detection infrared frequency, multiple high-risk and suspicious points of the high-risk and suspicious thermal area are determined.

4. The detection method for infrared stealth coating on aerospace transparent parts as described in claim 2, characterized in that, The determination of high-risk suspicious points based on the suspicious edge thermal defect line includes: Acquire a sequence of suspicious edge thermal defect lines in multiple consecutive frames at the initial infrared frequency during the spraying process; Based on the information of the multiple suspicious edge thermal points and the sequence information of the suspicious edge thermal defect lines of multiple consecutive frames of the initial infrared frequency during the spraying process, high-risk suspicious points of the suspicious edge thermal defect lines are determined.

5. A detection system for infrared stealth coatings on transparent aerospace components, characterized in that, include: The acquisition module is used to acquire the initial infrared frequency irradiation video of the infrared stealth coating being applied to the transparent aerospace component to be inspected. The first determining module is used to determine multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be inspected. The determination of multiple suspicious thermal regions, multiple suspicious edge thermal point information, and the first detection infrared frequency based on the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be inspected includes: using an illumination analysis model to determine the multiple suspicious thermal regions, multiple suspicious edge thermal point information, and the first detection infrared frequency. The illumination analysis model is a recurrent neural network, and the input of the illumination analysis model is the illumination video of the initial infrared frequency during the spraying of the infrared stealth coating on the transparent aerospace component to be inspected. The initial infrared frequency of the irradiation video, the output of the irradiation analysis model is multiple suspicious thermal regions, multiple suspicious edge thermal point information, and a first detection infrared frequency. The first detection infrared frequency is the infrared frequency determined by the irradiation analysis model for subsequent targeted detection of suspicious thermal regions. The irradiation video of the initial infrared frequency of the infrared stealth coating sprayed on the aviation transparent part to be tested provides the complete thermal evolution process of the coating in the initial state. The irradiation analysis model can analyze the temperature change trend of the pixels in the video over time, thereby identifying abnormal regions and edge points that do not match the normal spraying thermal characteristics, and inferring a first detection infrared frequency that is more suitable for capturing such defects based on the abnormal characteristics. The video acquisition module is used to acquire the irradiation video of the first detection infrared frequency for each suspicious thermal area; The heat map construction module is used to construct a heat map, which includes multiple suspected heat regions and multiple edges between the suspected heat regions. The node features of each suspected heat region include the illumination video of the initial infrared frequency of each suspected heat region and the illumination video of the first detection infrared frequency of each suspected heat region. The edges between the nodes are the distance information of the suspected heat regions. The first processing module is used to process the heat map based on a graph neural network to obtain high-risk and suspected heat regions; The clustering analysis module is used to cluster the multiple suspicious edge heat points to obtain K clusters; The generation module is used to generate suspicious edge thermal defect lines based on the K clusters; The second determining module is used to determine multiple suspicious intermediate test points based on the multiple suspicious thermal regions and the suspicious edge thermal defect lines. The suspicious intermediate test points are test points in the region between each suspicious thermal region and the suspicious edge thermal defect lines. The result determination module is used to determine the detection results of the infrared stealth coating of the aviation transparent part based on the high-risk suspected thermal area, the suspected edge thermal defect line, and the multiple suspected intermediate test points.

6. The detection system for infrared stealth coatings on transparent aerospace components as described in claim 5, characterized in that, The result determination module is also used for: Based on the aforementioned high-risk and suspicious thermal areas, multiple high-risk and suspicious points within the high-risk and suspicious thermal areas were identified; Based on the suspected edge thermal defect lines, high-risk suspected points of the suspected edge thermal defect lines are identified; Acquire phase-locked thermal images of multiple high-risk and suspicious points in high-risk and suspicious thermal areas, phase-locked thermal images of high-risk and suspicious points on suspicious edge thermal defect lines, and phase-locked thermal images of multiple suspicious intermediate test points; Based on phase-locked thermal imaging of multiple high-risk and suspicious points in the high-risk and suspicious thermal areas, phase-locked thermal imaging of high-risk and suspicious points in the suspicious edge thermal defect lines, phase-locked thermal imaging of multiple suspicious intermediate test points, information on multiple suspicious edge thermal points, irradiation video of the initial infrared frequency during the spraying of the infrared stealth coating on the aviation transparent part to be inspected, and irradiation video of the first detection infrared frequency of each suspicious thermal area, a defect probability distribution map of the infrared stealth coating is generated. The detection results of the infrared stealth coating on the aerospace transparent parts are determined based on the defect probability distribution map of the infrared stealth coating.

7. The detection system for infrared stealth coatings on transparent aerospace components as described in claim 6, characterized in that, The multiple high-risk and suspicious points identified based on the high-risk and suspicious thermal areas include: Acquire video of high-risk, suspicious thermal areas being irradiated at the second detection infrared frequency; Based on the irradiation video of the high-risk and suspicious thermal area at the second detection infrared frequency, multiple high-risk and suspicious points of the high-risk and suspicious thermal area are determined.

8. The detection system for infrared stealth coatings on transparent aerospace components as described in claim 6, characterized in that, The determination of high-risk suspicious points based on the suspicious edge thermal defect line includes: Acquire a sequence of suspicious edge thermal defect lines in multiple consecutive frames at the initial infrared frequency during the spraying process; Based on the information of the multiple suspicious edge thermal points and the sequence information of the suspicious edge thermal defect lines of multiple consecutive frames of the initial infrared frequency during the spraying process, high-risk suspicious points of the suspicious edge thermal defect lines are determined.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the detection method for infrared stealth coatings on aerospace transparent parts as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the detection method for infrared stealth coatings on aerospace transparent parts as described in any one of claims 1 to 4.

Images