X-ray security inspection machine picture object detection method

Abstract

The invention discloses a picture object detection method for an X-ray security inspection machine. The method comprises the following steps: acquiring a picture data set of the X-ray security inspection machine, generating a high-quality X-ray image sample through a generative adversarial network, expanding the data set, and optimizing a training process of an adversarial network generator through reinforcement learning. A multi-task learning model is constructed, object detection, object segmentation and object category identification are jointly trained, and under the condition of no labeled data, feature representation is generated through self-supervised learning to serve as an initial weight to be applied to training of an object detection model; the invention discloses a generative adversarial network. A traditional data enhancement method may have limitation in diversity and authenticity of generated samples. The complex condition in the actual environment can be better simulated through the sample generated by the GAN, so that the robustness of the model in the actual application is improved. And reinforcement learning optimization: utilizing reinforcement learning to optimize the training process of the GAN, and dynamically adjusting the generation strategy, so that the generated image samples are more diversified and real.

Description

Technical Field

[0001] The present invention relates to the technical field of image object detection, and in particular to an image object detection method for an X-ray security inspection machine. Background Art

[0002] exist In optical security inspection systems, the main task of object detection methods is to identify objects in images and classify and analyze them. Problems in current technologies include poor image quality, difficulty in detection due to overlapping objects, real-time issues, and false positive and false negative issues: 1. Image quality issues: Noise and Artifacts: X-ray images may contain noise and artifacts, which affect the accuracy of object detection.

[0003] Uneven contrast: Differences in contrast between objects can make some objects difficult to identify.

[0004] 2. Object Detection Accuracy: Overlapping objects: When multiple objects overlap, it may lead to false detection or missed detection.

[0005] Small Object Detection: Detection of small objects may be limited by image resolution and processing algorithms.

[0006] 3. Computational complexity: Real-time requirements: Complex processing and analysis of images may result in long calculation time, affecting real-time performance.

[0007] 4. Algorithm generalization ability: Insufficient training data: The training of deep learning models requires a large amount of labeled data. Insufficient data will affect the generalization ability of the model.

[0008] 5. Environmental changes: Different security equipment and environmental conditions may cause the model to perform inconsistently in different scenarios.

[0009] 6. False positive and false negative problems: False positive: Misidentifying a non-dangerous item as a dangerous item may result in a false alarm.

[0010] False negative: Dangerous items are misidentified as non-dangerous items, which may lead to missed reports. Summary of the invention

[0011] In order to solve one of the above technical problems, a method for detecting objects in X-ray security inspection machine images is provided, which has higher accuracy, better data processing capability, higher real-time performance and adaptability, and can effectively solve The main technical issues in object detection in optical security inspection machines.

[0012] In order to achieve the above purpose, the technical solution adopted by the present invention is: A first aspect: A method for detecting objects in an X-ray security inspection machine image, comprising the following steps: Obtain an X-ray security inspection machine image dataset, generate high-quality X-ray image samples through a generative adversarial network, expand the dataset, and optimize the training process of the adversarial network generator through reinforcement learning; Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models. Object detection is performed based on the transformer network, long-distance dependencies in the image are processed through the attention mechanism, and the inference path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology; According to the Isolation Forest anomaly detection model, abnormal objects or unusual combinations of objects are identified in X-ray images, and an adaptive decision-making system is constructed based on the identified abnormal objects or unusual combinations of objects.

[0013] Preferred: Obtain an X-ray security inspection machine image dataset, generate high-quality X-ray image samples through a generative adversarial network, expand the dataset, and optimize the training process of the adversarial network generator through reinforcement learning, specifically including: Collecting and arranging an X-ray image data set, the X-ray image data set includes several different types of objects, and preprocessing the X-ray image data set, the preprocessing including normalization and size adjustment; Constructing a GAN (Gan Adversarial Network) model, wherein the GAN model includes a generator and a discriminator, wherein the generator receives noise input and generates simulated X-ray image samples, and the discriminator distinguishes between real images and generated images; Training the model, in each iteration, the generator generates a number of forged X-ray images, the discriminator distinguishes the forged X-ray images from real images, and the adversarial loss function is used to optimize the generator and the discriminator so that the generated images gradually approach the real images; Use the trained generator to generate high-quality X-ray image samples for expanding the dataset, where the expression is: ; in, is the discriminator, is a generator, is a real image, It is noise. is the data distribution, is the noise distribution.

[0014] Preferably, the training process of optimizing the adversarial network generator through reinforcement learning is expressed as: ; in, is at the time step Rewards, It’s an instant reward. is the discount factor.

[0015] Preferred: Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models, including: The multi-task learning model includes sub-task networks for object detection, object segmentation, and object category recognition; According to the multi-task learning model sharing underlying features, the multi-task learning model is trained by a joint loss function; In response to the weighted summation of the losses of each task, the entire multi-task learning model is optimized, the performance of the multi-task learning model is evaluated on the validation set, and the weight parameters are adjusted to obtain the best performance of the multi-task learning model. The expression is: ; in, is the detection loss, is the segmentation loss, is the classification loss, and is the weight coefficient.

[0016] Preferably, the self-supervised learning generates feature representations through rotation prediction, image restoration or contrast learning, constructs a self-supervised model, uses unlabeled X-ray image data, trains the self-supervised model to extract effective feature representations, and uses the feature representations obtained by the self-supervised learning as initial weights for the training of the object detection model. The expression is: ; in, is the reconstruction loss, is the contrast loss.

[0017] Preferred: Object detection is performed based on the transformer network, long-distance dependencies in the image are processed through the attention mechanism, and the reasoning path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology, including: Perform object detection based on the DETR transformer network model, and use the attention mechanism to find the location and bounding box of objects in the global image range; Through the self-attention mechanism, the DETR transformer network model focuses on important areas in the image and handles long-distance dependencies; Use some labeled data for training, and evaluate and adjust on the validation set. The self-attention mechanism expression is: ; ; in, is the query matrix, is the key matrix, is the value matrix, is the dimension of the key.

[0018] Preferably: the dynamic reasoning technology includes: Dynamically select the optimal reasoning path according to the demand dynamics of the actual detection task of the current input image, which includes the complexity and required accuracy of the image; Different computing resources are allocated to each reasoning path, and reinforcement learning or heuristic algorithms are used to dynamically select the reasoning path. The expression is: ; in, is the path efficiency, is the path accuracy.

[0019] Preferably: according to the Isolation Forest anomaly detection model, an abnormal object or an unusual combination of objects is identified in the X-ray image, and an adaptive decision system is constructed according to the identification of the abnormal object or the unusual combination of objects, specifically including: Building an anomaly detection model to identify abnormal objects or unusual combinations of objects, wherein the abnormal objects include weapons, explosives, flammable liquid containers, and live animals, and the unusual combinations of objects include electronic devices and suspicious items, clothing and prohibited tools, and metal objects and liquids; Train anomaly detection models using historical and simulated anomaly data and deploy them in live systems; In the X-ray image stream, the anomaly detection model analyzes and detects abnormal objects in real time, and triggers an alarm mechanism or starts manual review. The anomaly detection model expression is: ; in, is the depth of the outlier, is the number of samples, and Depth is the depth of the tree.

[0020] Preferred: Adaptively adjust the detection strategy according to the type and quantity of objects detected in real time and the abnormal situation data in the historical time period; Adaptively adjust the detection threshold according to the real-time detection results, and optimize the rules and parameters of the decision-making system through the feedback loop. The formula is: Base Threshold+Adjustment Factor; Among them, Base Threshold is the basic threshold and Adjustment Factor is the adjustment factor.

[0021] The second aspect: A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor implements the X-ray security inspection machine image object detection method.

[0022] Compared with the prior art, the present invention has the following beneficial effects: The present invention generates adversarial networks (GAN): Traditional data augmentation methods may have limitations in the diversity and authenticity of generated samples. The samples generated by GAN can better simulate the complex situations in the actual environment, thereby improving the robustness of the model in practical applications.

[0023] Reinforcement learning optimization: Using reinforcement learning to optimize the GAN training process can dynamically adjust the generation strategy to make the generated image samples more diverse and realistic. This method can improve the image generation quality through continuous feedback and solve the sample diversity problem in the traditional GAN ​​training process.

[0024] Multi-task learning: By jointly training multiple tasks (such as object detection, segmentation, and classification), the model can improve the recognition ability of different types of objects while learning shared features. This approach not only improves the detection accuracy of the model, but also enhances the performance of the model in complex scenarios.

[0025] Self-supervised learning: Generating feature representations through self-supervised learning in the absence of labeled data can further improve the performance of the model, especially when data is scarce. This method utilizes the potential information of unlabeled data to make up for the data shortage problem in traditional supervised learning.

[0026] Transformer Networks: Traditional convolutional neural networks (CNNs) may perform poorly when dealing with long-range dependencies. Transformer Networks can effectively capture long-range dependencies in images through an attention mechanism, thereby improving the ability to detect complex objects and scenes.

[0027] Dynamic reasoning: By dynamically selecting the reasoning path according to the actual detection task, the computational burden is reduced and the efficiency of the system is improved. This method can adjust the reasoning strategy according to specific task requirements and optimize the use of computing resources.

[0028] Real-time anomaly detection: Using anomaly detection models (such as IsolationForest) can identify unusual objects and anomalies in real time, reduce false positives and false negatives, and improve system reliability.

[0029] Adaptive decision-making: Adaptive decision-making based on real-time data and historical records can adjust detection strategies according to environmental changes and improve the adaptability and flexibility of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The present invention is a flow chart of the method for detecting objects in X-ray security inspection machine images. DETAILED DESCRIPTION

[0031] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0032] Embodiment 1: Reference Figure 1 As shown, a method for detecting objects in an X-ray security inspection machine image includes the following steps: Obtain an X-ray security inspection machine image dataset, generate high-quality X-ray image samples through a generative adversarial network, expand the dataset, and optimize the training process of the adversarial network generator through reinforcement learning; Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models. Object detection is performed based on the transformer network, long-distance dependencies in the image are processed through the attention mechanism, and the inference path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology; According to the Isolation Forest anomaly detection model, abnormal objects or unusual combinations of objects are identified in X-ray images, and an adaptive decision-making system is constructed based on the identified abnormal objects or unusual combinations of objects.

[0033] Specifically: obtain the X-ray security inspection machine image data set, generate high-quality X-ray image samples through the generative adversarial network, expand the data set, and optimize the training process of the adversarial network generator through reinforcement learning, including: Collecting and arranging an X-ray image data set, the X-ray image data set includes several different types of objects, and preprocessing the X-ray image data set, the preprocessing including normalization and size adjustment; Many different types of objects (such as metals, liquids, electronic devices, etc.); Constructing a GAN (Gan Adversarial Network) model, wherein the GAN model includes a generator and a discriminator, wherein the generator receives noise input and generates simulated X-ray image samples, and the discriminator distinguishes between real images and generated images; Training the model, in each iteration, the generator generates a number of forged X-ray images, the discriminator distinguishes the forged X-ray images from real images, and the adversarial loss function is used to optimize the generator and the discriminator so that the generated images gradually approach the real images; Use the trained generator to generate high-quality X-ray image samples for expanding the dataset, where the expression is: ; in, is the discriminator, is a generator, is a real image, It is noise. is the data distribution, is the noise distribution.

[0034] Combination of GAN and reinforcement learning: GAN is used to generate high-quality X-ray image samples, and reinforcement learning is used to optimize the generation strategy, which improves the diversity and authenticity of the data set and thus enhances the robustness of the model. Specifically: The training process of optimizing the adversarial network generator through reinforcement learning is expressed as: ; in, is at the time step Rewards, It’s an instant reward. is the discount factor.

[0035] Specifically: Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models. Specifically, it includes: The multi-task learning model includes sub-task networks for object detection, object segmentation, and object category recognition; According to the multi-task learning model sharing underlying features, the multi-task learning model is trained by a joint loss function; In response to the weighted summation of the losses of each task, the entire multi-task learning model is optimized, the performance of the multi-task learning model is evaluated on the validation set, and the weight parameters are adjusted to obtain the best performance of the multi-task learning model. The expression is: ; in, is the detection loss, is the segmentation loss, is the classification loss, and is the weight coefficient. Application of multi-task learning and self-supervised learning: Jointly training detection, segmentation and recognition tasks, combined with self-supervised learning technology, improves detection accuracy and model generalization ability.

[0036] Specifically: the self-supervised learning generates feature representations through rotation prediction, image restoration or contrastive learning, builds a self-supervised model, uses unlabeled X-ray image data, trains the self-supervised model to extract effective feature representations, and uses the feature representations obtained by self-supervised learning as initial weights for the training of the object detection model. The expression is: ; in, is the reconstruction loss, is the contrast loss.

[0037] Specifically: Object detection is performed based on the transformer network, long-distance dependencies in the image are processed through the attention mechanism, and the reasoning path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology, including: Perform object detection based on the DETR transformer network model, and use the attention mechanism to find the location and bounding box of objects in the global image range; Through the self-attention mechanism, the DETR transformer network model focuses on important areas in the image and handles long-distance dependencies; Use some labeled data for training, and evaluate and adjust on the validation set. The self-attention mechanism expression is: ; ; in, is the query matrix, is the key matrix, is the value matrix, is the dimension of the key. Using transformer networks to handle long-range dependencies in images improves the accuracy and efficiency of object detection, especially in complex scenes.

[0038] Specifically: the dynamic reasoning technology includes: Dynamically select the optimal reasoning path according to the demand dynamics of the actual detection task of the current input image, which includes the complexity and required accuracy of the image; Different computing resources are allocated to each reasoning path, and reinforcement learning or heuristic algorithms are used to dynamically select the reasoning path. The expression is: ; in, is the path efficiency, is the path accuracy.

[0039] Specifically: Based on the Isolation Forest anomaly detection model, abnormal objects or unusual combinations of objects are identified in X-ray images, and an adaptive decision system is built based on the identification of abnormal objects or unusual combinations of objects, including: Building an anomaly detection model to identify abnormal objects or unusual combinations of objects, wherein the abnormal objects include weapons, explosives, flammable liquid containers, and live animals, and the unusual combinations of objects include electronic devices and suspicious items, clothing and prohibited tools, and metal objects and liquids; Aspects of unusual objects include: Prohibited Items: Weapons: Guns and their parts, including the metal bodies of pistols and rifles, can be identified as abnormal objects in X-ray images due to their unique shapes and density characteristics, even after being disguised. Knives, such as folding knives and daggers, have metal blades with a certain length and shape, which will appear different from ordinary luggage items in the image and are easy to detect.

[0040] Explosives related: For example, detonators have a relatively regular small tubular metal shape; if the gunpowder components commonly found in homemade explosives are carried, they may appear as block or powder aggregates with certain density and shape characteristics. These are abnormal objects that are strictly prohibited during security checks.

[0041] Flammable liquid containers: For example, large-capacity containers of flammable liquids such as gasoline and alcohol, whose appearance and internal liquids are different from those of conventionally allowed items under X-ray images, will be judged as abnormal objects because these items pose a great safety hazard in an aircraft transportation environment.

[0042] Live animals (not carried in accordance with regulations): If a passenger attempts to carry a pet or other small animal in luggage, its bones, body contours and other features that are different from those of ordinary luggage items can be seen in the X-ray image, and it is considered an abnormal object. For example, if a kitten is hidden in a backpack, the bone shape of its spine, limbs and other bones and the contours of its soft tissues can be distinguished in the security X-ray image, and under normal circumstances, live animals should not appear in such luggage.

[0043] Out-of-specification lithium batteries: Lithium batteries can present specific images in images based on their internal structure and external packaging. However, if the specifications of lithium batteries exceed the standards allowed by airlines for carry-on or checked baggage, such as ultra-large capacity power banks, they will be regarded as abnormal objects because large-capacity lithium batteries pose safety risks such as overheating and fire during flight.

[0044] Unusual aspects of object combinations include: The combination of electronic devices and suspicious items: For example, in a laptop bag, in addition to the normal images of laptop components, if there are also unidentified blocks connected by wires, and the blocks have suspicious shapes and density characteristics similar to the combination of batteries and detonators, this combination is very unusual and is very likely to involve the disguised carrying of improvised explosive devices, which requires special inspection during security checks.

[0045] Combination of clothing and prohibited tools: If metal tools of a certain length and shape, such as screwdrivers and wrenches, are mixed in between a pile of clothing, and these tools do not meet the requirements for the types of tools that can be carried on board, the combination of such clothing and tools that should not be there is unusual from the X-ray image, and may be a combination of items prepared for improper behavior such as intending to damage aircraft equipment.

[0046] Unreasonable density distribution combination inside luggage: Even if the items in normal luggage are placed in a messy way, their density distribution is relatively reasonable. For example, when common items such as clothes and books are mixed, the overall density is relatively uniform and the image characteristics of each area have certain regularities. However, if a part of a suitcase has extremely high density, such as a large amount of metal accumulation, and another part is almost empty, this is a very contrasting density distribution, and it cannot be explained by normal and reasonable luggage items. This combination is unusual and may contain prohibited large metal products and other abnormal objects, which requires further inspection and verification.

[0047] Train anomaly detection models using historical and simulated anomaly data and deploy them in live systems; In the X-ray image stream, the anomaly detection model analyzes and detects abnormal objects in real time, and triggers an alarm mechanism or starts manual review. The anomaly detection model expression is: ; in, is the depth of the outlier, is the number of samples, and Depth is the depth of the tree. Dynamic reasoning and real-time anomaly detection: Dynamic reasoning technology reduces the computational burden, the real-time anomaly detection model identifies anomalies and reduces false positives and false negatives, and the adaptive decision system provides flexible detection strategy adjustments, which improves the overall performance of the system.

[0048] Specifically: According to the type and number of objects detected in real time and the abnormal situation data in the historical time period, the detection strategy is adaptively adjusted; according to the real-time detection results, the detection threshold is adaptively adjusted, and the rules and parameters of the decision-making system are optimized through the feedback loop. The formula is: Base Threshold+Adjustment Factor; Among them, Base Threshold is the basic threshold and Adjustment Factor is the adjustment factor.

[0049] 1. Combination of GAN and reinforcement learning: Generative Adversarial Networks (GAN): Traditional data augmentation methods may have limitations in the diversity and authenticity of generated samples. The samples generated by GAN can better simulate the complex situations in the actual environment, thereby improving the robustness of the model in practical applications.

[0050] Reinforcement learning optimization: Using reinforcement learning to optimize the GAN training process can dynamically adjust the generation strategy to make the generated image samples more diverse and realistic. This method can improve the image generation quality through continuous feedback and solve the sample diversity problem in the traditional GAN ​​training process.

[0051] For example, in airport security, GAN can generate a variety of different types of dangerous goods images to enhance the diversity of the data set. Combined with reinforcement learning, the model can adaptively adjust during the generation process to produce samples that conform to the actual situation, thereby improving the object detection system's ability to identify complex objects.

[0052] 2. Application of multi-task learning and self-supervised learning: Multi-task learning: By jointly training multiple tasks (such as object detection, segmentation, and classification), the model can improve the recognition ability of different types of objects while learning shared features. This approach not only improves the detection accuracy of the model, but also enhances the performance of the model in complex scenarios.

[0053] Self-supervised learning: Generating feature representations through self-supervised learning in the absence of labeled data can further improve the performance of the model, especially when data is scarce. This method utilizes the potential information of unlabeled data to make up for the data shortage problem in traditional supervised learning.

[0054] For example, for the task of object detection in optical images, multi-task learning can simultaneously optimize the ability to identify the location, boundary, and category of objects. In the case of unlabeled data, self-supervised learning technology can generate useful feature representations through contrastive learning, further improving the accuracy of the detection system.

[0055] 3. Introduction of Transformer Network: Transformer Networks: Traditional convolutional neural networks (CNNs) may perform poorly when dealing with long-range dependencies. Transformer Networks can effectively capture long-range dependencies in images through an attention mechanism, thereby improving the ability to detect complex objects and scenes.

[0056] For example, in light images, the transformer network can effectively process global contextual information in the image, such as detecting the relative position and structure of complex objects inside a package. This method is more advantageous than traditional CNN in processing complex scenes and can improve the overall detection accuracy of the system.

[0057] 4. Dynamic reasoning and real-time anomaly detection: Dynamic reasoning: By dynamically selecting the reasoning path according to the actual detection task, the computational burden is reduced and the efficiency of the system is improved. This method can adjust the reasoning strategy according to specific task requirements and optimize the use of computing resources.

[0058] Real-time anomaly detection: Using anomaly detection models (such as IsolationForest) can identify unusual objects and anomalies in real time, reduce false positives and false negatives, and improve system reliability.

[0059] Adaptive decision-making: Adaptive decision-making based on real-time data and historical records can adjust detection strategies according to environmental changes and improve the adaptability and flexibility of the system.

[0060] For example, when dealing with high-volume security inspection tasks, dynamic reasoning technology can adjust the calculation path according to real-time conditions to optimize processing efficiency. Real-time anomaly detection models can quickly identify and process abnormal objects, reducing false positives and false negatives. Adaptive decision systems can automatically adjust detection strategies according to different environmental conditions (such as peak traffic periods) to improve overall detection performance.

[0061] In summary: Compared with the prior art, the present invention not only has significant progress in the diversity, accuracy and efficiency of data processing, but can also effectively solve the main technical problems in object detection in traditional optical security inspection machines and improve the performance and reliability of the overall system.

[0062] Embodiment 2 A sort of The method for detecting objects in optical security inspection machine images is applied to deep learning and multimodal fusion, and includes the following steps: Step 1: Integrate data from different Datasets of optical security inspection machines and different environments to improve the generalization ability of the model; the datasets include data from different angles, different exposure conditions, and different object types; use synthetic image generation technology (such as using GAN generators) to expand the dataset and simulate rare objects and abnormal situations (such as objects in different backgrounds and environments); apply image enhancement techniques such as rotation, scaling, translation, color transformation, etc. to improve the model's robustness to image deformation and noise; Geometric transformations: such as rotation, flipping, scaling, etc., make the model robust to objects of different angles and scales.

[0063] Color jitter: Enhance the model's adaptability under different lighting conditions by adjusting brightness, contrast, saturation, etc.

[0064] Noise addition: Add Gaussian noise, salt and pepper noise, etc. to the image so that the model can correctly identify objects under noise interference; Geometric transformation matrix: ; Noise addition: ; in, For the enhanced image, The mean is , the standard deviation is Noise Synthetic sample generation: Generate synthetic samples based on existing object categories, simulating various occlusion situations, object overlaps, and lighting conditions; Multimodal fusion: through integration Light images and other sensor data enhance the comprehensiveness and accuracy of object detection. This innovation uses multimodal information to improve the recognition of complex scenes and objects.

[0065] Step 2: Joint detection and classification: Use a multi-task learning network (such as YOLOv5, Efficient Det) to perform object detection and classification simultaneously to improve detection accuracy and classification accuracy; introduce a fine-grained feature learning module to extract more detailed features of objects to enhance the ability to distinguish similar objects; introduce a dynamic convolution layer to dynamically adjust the convolution kernel weights based on the input data to adapt to the characteristics of different objects; Integrate attention mechanisms (such as SE blocks or CBAM) to enhance focus on key information and automatically optimize important areas of feature maps; Dynamic convolution and attention mechanism: Dynamic convolution and attention mechanism are used to improve the ability to recognize object details, enabling the model to adaptively adjust under different conditions and enhancing the robustness of detection; Step 3: Fusion of X-ray images with other sensor data: Integrate other sensor data (such as millimeter wave radar, infrared images) with Light images, combined with multimodal information through deep fusion networks (such as Transformer fusion modules), provide more comprehensive object information; using multi-angle shooting The optical images are fused to reconstruct the three-dimensional information of the object and improve the recognition and classification of the object; Real-time optimization and feedback mechanism: Real-time optimization and automated training systems enable the model to be continuously improved during actual operation, solving the problem that traditional models cannot be updated in a timely manner; Based on contextual information (such as background objects and environmental information of the detection area), use context-aware networks (such as Transformer models) to improve the accuracy of object detection; Context perception and relationship modeling: Through context perception analysis and target relationship modeling, the recognition ability of complex scenes and object combinations is improved, and false detection and missed detection are reduced; In complex scenes, the relationship between targets is modeled through relational networks to improve the recognition ability of complex object combinations; Step 4: Use lightweight deep learning models (such as Mobile Net and Efficient Net) to optimize computing performance, ensure fast processing in a real-time environment, implement a dynamic adjustment mechanism, automatically adjust model parameters and processing strategies according to actual operating conditions (such as processing speed and detection accuracy), integrate user feedback mechanisms, collect false detections and missed detections in real time, and dynamically adjust and optimize model parameters; Automated training system: Build an automated model training and update system to automatically update the model based on real-time data and user feedback to maintain continuous improvement in detection capabilities.

[0066] Specifically, 1. Innovation of multimodal fusion: Existing optical security inspection systems typically rely solely on light images for object detection and recognition. This approach can be limited in scenes with occluded, overlapping, or complex objects.

[0067] Multimodal data fusion: By integrating X-ray images with other sensor data (such as millimeter-wave radar and infrared images), more comprehensive object information is provided. This fusion can significantly improve detection accuracy. For example, millimeter-wave radar can penetrate most materials and help detect objects obscured by X-ray images; infrared images can help identify objects with abnormal temperatures.

[0068] Multi-view fusion: Combining light images taken from different angles to reconstruct the three-dimensional information of an object is rare in existing technologies. This method can more clearly identify complex object combinations and improve recognition accuracy.

[0069] 2. Application of dynamic convolution and attention mechanism: Traditional deep learning models use fixed convolution kernels for feature extraction and may not be able to adapt to changes in different objects and environmental conditions.

[0070] Dynamic convolution: Introducing dynamic convolution layers to dynamically adjust the convolution kernel weights based on input data. This innovation can improve the model's ability to adapt to different object features, making the model more flexible and accurate when dealing with different objects or complex scenes.

[0071] Attention mechanism: Apply attention mechanism (such as SE block or CBAM) to enhance the focus on key information and automatically optimize the important areas of the feature map. This mechanism can significantly improve the model's ability to extract key features and reduce interference from background noise.

[0072] 3. Real-time optimization and automated training system: Existing optical security inspection systems usually rely on static models during operation and cannot be optimized and updated in real time according to actual conditions.

[0073] Real-time processing and optimization: Using lightweight models and dynamic adjustment strategies, fast processing and optimization in real-world environments are achieved. This enables the system to respond to environmental changes in real time and improve detection efficiency. For example, the system can automatically adjust model parameters based on actual operating conditions to ensure efficient performance under high load.

[0074] Automated training system: Build an automated model training and update system to automatically update the model based on real-time data and user feedback. This approach solves the problem of traditional models requiring manual updates, allowing the system to continuously improve and adapt to new detection needs.

[0075] 4. Context awareness and relationship modeling: Traditional methods usually ignore the relationships between objects and the scene context, and have limited ability to recognize object combinations in complex scenes.

[0076] Context-aware optimization: Use context-aware networks (such as Transformer models) to analyze the scene background and provide more accurate object recognition. For example, the model can adjust the detection strategy based on the scene background information to distinguish between background objects and main objects.

[0077] Target relationship modeling: Modeling the relationship between targets through relational networks improves the ability to identify complex combinations of objects. For example, when multiple objects are found in luggage, the system can identify their relationships, thereby improving the accuracy of identifying potential threats.

[0078] For example: Practical application scenarios: In airport security checks, traditional optical security inspection machines may have difficulty identifying complex combinations of objects in luggage, such as mixtures of metal objects with batteries, liquids, etc. After the introduction of multimodal data fusion technology, millimeter-wave radar can penetrate liquid containers, and infrared images can detect abnormal battery heating, thereby improving the ability to identify potentially dangerous items.

[0079] Complex scene processing: In a piece of luggage, using dynamic convolution and attention mechanisms can help the model better identify occluded objects. For example, a large package contains multiple small items, and the attention mechanism can automatically focus on the key features of these small items, even if they are partially occluded by other items.

[0080] Real-time adjustment: During peak hours, traditional systems may experience reduced efficiency due to slow processing speeds. With real-time optimization and automated training systems, models can dynamically adjust to cope with high loads, improving overall processing speed and accuracy and avoiding performance bottlenecks during busy hours.

[0081] Embodiment 3 An embodiment of the present application also provides a computer-readable storage medium, which can be set in a server to store at least one instruction or at least one program related to an X-ray security inspection machine image object detection method in a method embodiment, and the at least one instruction or the at least one program is loaded and executed by the processor to implement an X-ray security inspection machine image object detection method provided in the above method embodiment.

[0082] Optionally, in this embodiment, the storage medium may be located in at least one of the multiple network servers of the computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0083] Embodiment 4 The embodiment of the present invention further provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes an X-ray security inspection machine image object detection method provided in the above various optional embodiments.

[0084] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0086] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0087] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

[0088] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting objects in X-ray security inspection machine images, characterized in that: The following steps are involved: Obtain an X-ray security inspection machine image dataset, generate high-quality X-ray image samples through a generative adversarial network, expand the dataset, and optimize the training process of the adversarial network generator through reinforcement learning; Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models. Object detection is performed based on the transformer network, long-distance dependencies in the image are processed through the attention mechanism, and the inference path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology; According to the Isolation Forest anomaly detection model, abnormal objects or unusual combinations of objects are identified in X-ray images, and an adaptive decision-making system is constructed based on the identified abnormal objects or unusual combinations of objects.

2. The method according to claim 1, characterized in that: Obtain an X-ray security inspection machine image dataset, generate high-quality X-ray image samples through a generative adversarial network, expand the dataset, and optimize the training process of the adversarial network generator through reinforcement learning, including: Collecting and arranging an X-ray image data set, the X-ray image data set includes several different types of objects, and preprocessing the X-ray image data set, the preprocessing including normalization and size adjustment; Constructing a GAN model, the GAN model comprising a generator and a discriminator, wherein the generator receives noise input and generates simulated X-ray image samples, and the discriminator distinguishes between real images and generated images; Training the model, wherein in each iteration, the generator generates a number of forged X-ray images, the discriminator distinguishes the forged X-ray images from real images, and the generator and the discriminator are optimized by an adversarial loss function; The trained generator is used to generate high-quality X-ray image samples for expanding the dataset.

3. The method according to claim 2, characterized in that: The training process of optimizing the adversarial network generator through reinforcement learning is expressed as: ； in, is at the time step Rewards, It’s an instant reward. is the discount factor.

4. The method according to claim 1, characterized in that: Build a multi-task learning model to jointly train object detection, object segmentation, and object category recognition. In the absence of labeled data, generate feature representations as initial weights through self-supervised learning and apply them to the training of object detection models. Specifically, The multi-task learning model includes sub-task networks for object detection, object segmentation, and object category recognition; According to the multi-task learning model sharing underlying features, the multi-task learning model is trained by a joint loss function; In response to the weighted summation of the losses of each task, the entire multi-task learning model is optimized, the performance of the multi-task learning model is evaluated on the validation set, and the weight parameters are adjusted to obtain the best performance of the multi-task learning model.

5. The method according to claim 4, characterized in that: The self-supervised learning generates feature representations through rotation prediction, image restoration or contrastive learning, constructs a self-supervised model, uses unlabeled X-ray image data, trains the self-supervised model to extract effective feature representations, and uses the feature representations obtained by the self-supervised learning as initial weights for the training of the object detection model.

6. The method according to claim 1, characterized in that: Object detection is performed based on the transformer network. The long-distance dependencies in the image are processed through the attention mechanism. The inference path is dynamically selected according to the needs of the actual detection task through dynamic reasoning technology, including: Perform object detection based on the DETR transformer network model, and use the attention mechanism to find the location and bounding box of objects in the global image range; Through the self-attention mechanism, the DETR transformer network model focuses on important areas in the image and handles long-distance dependencies; Use some labeled data for training and evaluate and adjust on the validation set.

7. The method according to claim 6, characterized in that: The dynamic reasoning technology includes: Dynamically select the optimal reasoning path according to the demand dynamics of the actual detection task of the current input image, which includes the complexity and required accuracy of the image; Different computing resources are allocated to each reasoning path, and reinforcement learning or heuristic algorithms are used to dynamically select the reasoning path.

8. The method according to claim 1, characterized in that: According to the Isolation Forest anomaly detection model, abnormal objects or unusual combinations of objects are identified in X-ray images, and an adaptive decision system is built based on the identification of abnormal objects or unusual combinations of objects, including: Building an anomaly detection model to identify abnormal objects or unusual combinations of objects, wherein the abnormal objects include weapons, explosives, flammable liquid containers, and live animals, and the unusual combinations of objects include electronic devices and suspicious items, clothing and prohibited tools, and metal objects and liquids; Train anomaly detection models using historical and simulated anomaly data and deploy them in live systems; In the X-ray image stream, the anomaly detection model analyzes and detects abnormal objects in real time and triggers an alarm mechanism or initiates manual review.

9. The method according to claim 8, characterized in that: The construction of the adaptive decision-making system includes: adaptively adjusting the detection strategy according to the type and quantity of objects detected in real time and the abnormal situation data in the historical time period; adaptively adjusting the detection threshold according to the real-time detection results, and optimizing the rules and parameters of the decision-making system through a feedback loop.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the X-ray security inspection machine image object detection method according to any one of claims 1 to 9.

Images