Article binding detection method and device

Abstract

The article binding detection method disclosed by the invention comprises the following steps: acquiring a visible light image and an infrared image which are synchronously acquired and pass through pedestrians in a detection area, and performing image registration; according to an actual detection condition, dynamically and adaptively selecting a suitable target detection means to carry out target detection on the registered infrared image; in the registered infrared image, the article exposure proportion of each article area pixel point to the human body area pixel point is obtained, whether the article exposure proportion of each article area is larger than a preset article exposure proportion threshold value or not is judged, and if yes, it is judged that an article is bound in the human body area. According to the method, infrared thermal imaging and visible light images are combined through a dual-light image registration technology, and comprehensive detection of binding and hiding of illegal objects on the body can be realized. A dynamic adaptive technology is adopted, a detection strategy can be intelligently adjusted according to different distances and imaging effects, and the adaptive capacity and the detection precision of various scenes can be remarkably improved by combining article exposure proportion judgment.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent security inspection, and more specifically, to an article binding and hiding detection method and device. Background Art

[0002] Existing "terahertz" electromagnetic wave imaging technology security inspection systems can conduct inspections for hidden items on people. However, their inspection efficiency is relatively low, and it takes 2 - 5 minutes on average to inspect one person. Terahertz electromagnetic waves are between the infrared band in the optical domain and the millimeter wave band in the microwave domain, and are a mysterious electromagnetic wave with a spectral range of 0.1 - 10 THz (terahertz) and a wavelength between 0.03 - 3 mm. Terahertz imaging is mainly based on the interaction between terahertz waves and substances.

[0003] However, generally low power is a major problem faced by terahertz technology. During the process of imaging a two - dimensional object, the energy of the terahertz source will be dispersed to each point on the two - dimensional plane, resulting in lower terahertz energy received by a single detection point. This directly affects the sensitivity and resolution of imaging, making the imaging effect less than ideal, with a resolution often less than 256 * 192 PPI, being not easy to detect, and having great limitations in the field of personal binding and hiding detection applications.

[0004] The sensitivity of the detector is also an important factor restricting the development of terahertz technology. Due to the low energy and long wavelength of terahertz radiation, traditional detectors often have insufficient sensitivity when detecting terahertz waves. This makes it difficult to directly obtain high - quality terahertz images.

[0005] The resolution and penetration power of terahertz imaging technology are relatively weak. Although terahertz waves can penetrate some non - metallic materials, their penetration depth is limited, and they are easily affected by environmental factors during transmission, such as absorption and scattering by substances such as moisture and air, as well as electromagnetic interference. Also, the hardware devices related to terahertz imaging technology are often large in size, occupy a large area, and have high costs. Summary of the Invention

[0006] In view of at least one defect or improvement requirement of the prior art, the present application provides an article binding and hiding detection method and device, which are used to at least solve the technical problem of low terahertz inspection efficiency and achieve real - time detection of human body article binding and hiding.

[0007] To achieve the above object, in a first aspect, the present application provides an article binding and hiding detection method, including:

[0008] Obtaining synchronously collected visible - light images and infrared images of pedestrians passing through the detection area;

[0009] Performing image registration on the infrared image and the visible - light image;

[0010] Dynamically and adaptively select a suitable target detection method according to the actual detection conditions to perform target detection on the registered infrared image.

[0011] In the registered infrared image, respectively obtain the item exposure ratio of each item area pixel point to the corresponding human body area pixel point, and respectively determine whether the item exposure ratio of each item area is greater than a preset item exposure ratio threshold. If it is greater, it is determined that the corresponding human body area hides an item.

[0012] Further, the dynamically and adaptively selecting a suitable target detection method according to the actual detection conditions to perform target detection on the registered infrared image includes:

[0013] If the imaging distance of the registered infrared image is greater than the first preset distance threshold, use the trained deep learning target detection model to perform precise target detection;

[0014] If the imaging distance of the registered infrared image is less than the second preset distance threshold, use traditional image processing methods for auxiliary target detection.

[0015] Further, the acquisition method of the trained deep learning target detection model includes:

[0016] Collect videos of human body hiding items in multiple scenarios, convert the videos into pictures at a preset frame rate interval and remove duplicates to obtain an image dataset;

[0017] Segment and exclude the areas on the images of the image dataset that are not part of the human body, and then perform label annotation of the items;

[0018] After performing data augmentation processing on the image dataset, use the YOLOv8 model for training to obtain the trained deep learning target detection model.

[0019] Further, the discrimination method for the item exposure ratio of the item area pixel point to the corresponding human body area pixel point includes:

[0020] In the registered infrared image, respectively obtain the total number of pixels of each item area;

[0021] Obtain the total number of pixels of the human body area pixels on the corresponding human body area;

[0022] The item exposure ratio of the item area = (the total number of pixels of the pixel points of this item area) / (the total number of pixels of the human body area pixels);

[0023] If the exposure ratio of the item is greater than the preset item exposure ratio threshold, it is determined that the corresponding human body area hides the item; otherwise, it is determined that the corresponding human body area does not hide the item.

[0024] Further, the image registration of the infrared image and the visible light image includes:

[0025] Extract several groups of corresponding feature points from the infrared image and the visible light image and perform normalization processing to obtain normalized control points, construct a matrix representing the set of normalized control points, and expand the set of normalized control points into homogeneous coordinates;

[0026] Use the singular value decomposition method to obtain the inverse matrix of the matrix, judge the linear independence of points, and if the linear independence condition is satisfied, continue with matrix calculations;

[0027] Based on the correspondence between the normalized matrix, the inverse matrix, and the original feature point set, perform matrix operations on the inverse matrix and the matrix to obtain a transformation matrix from the visible light image to the infrared image, and convert the point coordinates in the visible light image to the corresponding positions in the infrared image to align the visible light image with the infrared image.

[0028] Further, after performing data augmentation processing on the image dataset, using the YOLOv8 model for training to obtain the trained deep learning object detection model includes:

[0029] Divide the image dataset after data augmentation processing into a training set and a test set according to a preset ratio;

[0030] Input the training set into the YOLOv8 model to generate object detection results and segmentation results;

[0031] In each round, based on the loss of object detection and the loss of object segmentation, calculate the comprehensive loss value between the predicted image and the annotated image; the loss calculation of object detection includes one or more of position loss, class loss, and confidence loss; the loss calculation of object segmentation includes comparing the difference between the predicted segmentation mask and the true mask;

[0032] Perform gradient backpropagation based on the comprehensive loss value to optimize the parameters of the model, and continuously iterate until the model parameters converge to obtain a deep learning object detection model;

[0033] Use the test set to evaluate the deep learning object detection model, adjust relevant hyperparameters according to the evaluation results, and obtain the trained deep learning object detection model.

[0034] Further, it also includes:

[0035] After determining that the item is hidden in the corresponding human body area, an alarm message is automatically sent.

[0036] Further, the area of the non-human part includes one or more of a backpack, a shoulder bag, a hat, a skirt, and a coat.

[0037] In a second aspect, the present application provides a device, including at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit can execute the steps of the detection method described in any one of the foregoing.

[0038] Further, the device is hoisted and installed.

[0039] Generally speaking, compared with the prior art by the above technical solution conceived by the present application, the following beneficial effects can be obtained:

[0040] (1) By integrating the dual-light image technology, the present application effectively overcomes the limitations of traditional single-modal recognition methods and achieves higher detection accuracy. Through the dual-light image registration technology, the segmentation and detection results of the registered visible light are applied to the infrared image, and then the advantage of infrared thermal imaging in helping to reduce the interference of background objects is used to detect the illegal item hidden in the infrared image, so as to achieve a comprehensive detection of the illegal item hidden on the body. Moreover, the dynamic adaptive technology is adopted, which can intelligently adjust the detection strategy according to different imaging distances and imaging effects. Combined with the discrimination of the item exposure ratio, the adaptability, detection accuracy and detection efficiency for various scenarios can be significantly improved.

[0041] (2) According to different imaging distances and imaging effects, the present application can automatically select the optimal image processing strategy: for the case of long distance and clear imaging, the YOLOv8 object detection deep learning model is used for accurate object detection; for the case of short distance and blurred imaging, traditional image processing methods (such as: binary method) can be used for auxiliary object detection. This dynamic adaptive method significantly improves the adaptability and detection accuracy for various scenarios.

[0042] (3) The present application determines whether there is an illegal item hidden by the item exposure ratio of the total number of pixels in each item area to the total number of pixels in the human body area pixels. If this ratio is greater than the preset threshold, it is determined that there is an illegal item hidden in the human body area; if this ratio is lower than the threshold, it is considered that there is no illegal item hidden. This threshold discrimination method of the item exposure ratio effectively improves the accuracy and reliability of illegal item detection.

[0043] (4) This application segments and excludes areas such as backpacks, shoulder bags, hats, skirts, and coats on the images of the image dataset, and then labels the prohibited items, thereby effectively reducing the misidentification problem and improving the detection accuracy of prohibited items.

[0044] (5) The device of this application can actively receive the infrared radiation of an object to achieve non-intrusive identification in the passage; and it supports hoisting and does not occupy floor space. Description of the Drawings

[0045] To more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 It is the core flowchart of an item hiding detection method provided by an embodiment of this application;

[0047] Figure 2 It is the schematic flowchart of a body dangerous goods (prohibited items) hiding detection method based on dual-light image registration provided by an embodiment of this application;

[0048] Figure 3 It is the actual effect diagram of item hiding detection provided by an embodiment of this application;

[0049] Figure 4 It is the block diagram of a device suitable for implementing the detection method described above provided by an embodiment of this application. Detailed Embodiments

[0050] In order to make the purpose, technical solutions, and advantages of this application clearer, the following further details this application in combination with the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain this application and are not used to limit this application. In addition, the technical features involved in the various embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0051] The terms "first", "second", or "nth" in the specification, claims, or drawings of this application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include steps or units not listed, or may optionally further include other steps or units inherent to these processes, methods, products, or devices.

[0052] As described in the background art of the specification, the existing "terahertz" electromagnetic wave imaging technology has defects such as low power, low detection efficiency, insufficient sensitivity, relatively weak resolution and penetration power, and the related hardware devices are often large in size, occupy a large area, and have high costs. In view of this, the present application proposes an article binding and hiding detection method and device, which can at least solve the technical problem of unsatisfactory imaging caused by small terahertz resolution, and achieve comprehensive detection of high adaptability, high accuracy and high detection efficiency for illegal article binding and hiding on the human body.

[0053] The general technical idea of the embodiments of the present application is as follows: First, the visible light image and the infrared image are aligned through image registration technology, and then different recognition algorithms are performed on the registered visible light image and infrared image, and algorithm recognition is performed using the different advantages of the dual-light images, and the recognition results of the dual-light images affect each other. Since the visible light image has many and obvious features, algorithms for human detection and human segmentation can be performed on the visible light image; and since the infrared image is formed according to the temperature difference of the heat distribution, the detection of illegal article binding and hiding is performed on the infrared image. In addition, the embodiments of the present application adopt dual-light image registration modeling, and the segmentation and detection results of the registered visible light are applied to the infrared image, and then the advantage that infrared thermal imaging can help reduce the interference of background objects is used to detect illegal article binding and hiding in the infrared image. At the same time, the embodiments of the present application can also calculate the distance between the human body and the camera based on the human body area in the infrared image, and adaptively select different algorithms for target detection according to different distances. When imaging at a long distance, the human body features are generally clear, and at this time, a trained deep learning target detection model can be used for target detection and recognition; while when imaging at a short distance, the imaging is generally blurred, and at this time, traditional image processing methods (binary method) can be used to improve the accuracy and efficiency of detection. In order to improve the accuracy of detection, the results detected on the infrared image can also be transmitted back to the segmentation results of the visible light image for verification to exclude misrecognition.

[0054] Refer to Figure 1 and Figure 2 An embodiment of the present application provides an article binding and hiding detection method, and the detection method may include the following steps.

[0055] Step 1, obtain visible light images and infrared images collected synchronously of pedestrians passing through the detection area.

[0056] In some embodiments, specifically, infrared images and visible light images of pedestrians passing through the security check area are captured in real time based on a dual-light acquisition device. This step is performed by a fixed-focus high-resolution camera and an infrared sensor to ensure clear image data is captured.

[0057] By integrating the dual - light image technology, this application effectively overcomes the limitations of traditional single - modality recognition methods and achieves higher detection accuracy.

[0058] Step 2: Perform image registration on the infrared image and the visible - light image.

[0059] In some embodiments, specifically, five groups of corresponding feature points are extracted from the visible - light image and the infrared image, normalized, and the normalized control points are calculated. The set of normalized control points is extended to homogeneous coordinates, and a matrix representing the set of normalized control points is constructed, and the set of normalized control points can be stored in this matrix.

[0060] Use the SVD (Singular Value Decomposition) method to calculate the inverse matrix of the set of normalized control points, and judge the linear independence of points (linear independence condition: at least 3 non - collinear points). If the linear independence condition is met, continue with the matrix calculation.

[0061] Through matrix operations, calculate the inverse matrix and the set of normalized control points. Utilize the relationship between the normalization matrix, the inverse matrix, and the original feature point set to obtain a 3x3 transformation matrix from the visible - light image to the infrared image. This 3x3 transformation matrix can convert the point coordinates in the visible - light image to the corresponding positions in the infrared image, thereby aligning the visible - light image with the infrared image.

[0062] Step 3: Dynamically and adaptively select a suitable target detection method according to the actual detection conditions to perform target detection on the registered infrared image.

[0063] In some embodiments, specifically, according to the actual detection conditions (such as distance and imaging effect), dynamically select the optimal image - processing method or target detection means.

[0064] For the case where the imaging distance is far and the imaging is clear (when the imaging distance is far, the imaging is generally clear), a trained deep - learning target detection model can be used for accurate target detection; while for the case where the imaging distance is near and the imaging is blurred (when the imaging distance is near, the imaging is generally blurred), traditional image - processing methods such as binarization can be used for auxiliary target detection.

[0065] In some embodiments, more specifically, the acquisition method of the trained deep - learning target detection model may include the following parts.

[0066] (1) Dataset collection

[0067] Based on the dedicated dual - light acquisition device K1280 developed by the company, the test camera is placed on a tripod, with a height of 1.3 m from the ground, and the object to be tested is 0.5 - 5 m away from the camera. It simulates illegal items that may be involved in smuggling, such as electronic products (mobile phones), jewelry, or even guns and drugs. The binding and hiding parts include the waist, legs, back, etc. Videos of illegal items bound and hidden, taking into account factors such as different heights, genders, and clothing, are collected in the form of video streams, and then the videos are converted into pictures at a certain frame rate interval, and duplicate pictures are deleted to ensure the uniqueness of each picture, obtaining an image dataset.

[0068] (2) Dataset annotation

[0069] Analyze the infrared imaging data collected on - site. Items with a relatively high imaging similarity to illegal items are divided into eight categories: backpacks, shoulder bags, hats, skirts, loose coats, water stains, sweat stains, and oil stains. Any non - illegal item presenting an imaging similar to a suspected illegal item will affect the accuracy of the recognition result. Therefore, it is necessary to segment and exclude these areas of items with relatively similar imaging on non - human parts (such as backpacks, shoulder bags, hats, skirts, loose coats, etc.) on the image, and use the threshold segmentation method to correspondingly exclude parts such as water stains, sweat stains, and oil stains. Then, label the illegal items bound and hidden. When labeling, only label the part within the human torso, and at most two boxes are used to detect one person. In this application, areas such as backpacks, shoulder bags, hats, skirts, and coats on the images of the image dataset are segmented and excluded, parts such as water stains, sweat stains, and oil stains are correspondingly excluded using the threshold segmentation method, and then label the illegal items, thus effectively reducing the problem of mis - recognition and improving the detection accuracy of illegal items.

[0070] (3) Data augmentation

[0071] It is used to balance the number of category samples and increase the generalization ability of the model.

[0072] Image enhancement: The main means include horizontal flipping, random translation, random angle transformation, random cropping, random brightness change, random contrast change, random sharpness change, and random scale change, etc.

[0073] Image registration enhancement: Affine transformation.

[0074] Enhancement specific to infrared images: Temperature range adjustment, thermal noise simulation, and dynamic blur, etc.

[0075] Target scale transformation: Target distance adjustment, target size adjustment, and dynamic target boundary enhancement, etc.

[0076] Background adjustment: Background blur, background noise adjustment, etc.

[0077] Data mixing: Image fusion, synthetic data generation, etc.

[0078] (4) Model training and test adjustment

[0079] The image dataset is divided into a training set and a test set in a ratio of 8:2, and each image must be paired with its corresponding label (including the target bounding box and segmentation mask).

[0080] After applying various data augmentation techniques to the data in the training set, the YOLOv8 model is used for training. After inputting the preprocessed images into the object detection and segmentation model, the model simultaneously generates object detection results (bounding box, class, confidence) and object segmentation results (mask for each object).

[0081] Then, the loss is calculated for iterative improvement of the model's performance. The loss calculation for object detection includes location loss (bounding box regression loss), class loss (classification loss), and confidence loss (probability loss of object existence). The loss calculation for object segmentation includes comparing the difference between the predicted segmentation mask and the ground truth mask. By analyzing the output results of each round, the comprehensive loss value between the predicted image and the annotated image (ground truth image) is calculated. Based on the comprehensive loss value, the gradient is backpropagated to optimize the model's parameters, and the iteration continues until the model parameters converge, obtaining a deep learning object detection model.

[0082] Finally, the performance of the model is evaluated on the test set, including the accuracy of object detection (mAP) and the performance of segmentation (IoU). According to the evaluation results, hyperparameters such as the learning rate and batch size are adjusted to improve the model's performance and ensure excellent performance in both object detection and image segmentation tasks. After such test evaluation and hyperparameter adjustment, a trained deep learning object detection model can be obtained.

[0083] In this application, through the dual - light image registration technology, the segmentation and detection results of the registered visible light are applied to the infrared image, and then the advantage of infrared thermal imaging in helping to reduce the interference of background objects is utilized to detect the concealed illegal items in the infrared image, enabling a comprehensive detection of concealed illegal items on the body. This application can automatically select the optimal image - processing strategy according to different imaging distances and imaging effects, and this dynamic adaptive method significantly improves the adaptability and detection accuracy for various scenarios.

[0084] Step 4: In the registered infrared image, obtain the item exposure ratio of each item area's pixel points to the corresponding human body area's pixel points respectively, and determine whether the item exposure ratio of each item area is greater than the preset item exposure ratio threshold. If it is greater, it is determined that the corresponding human body area conceals an item.

[0085] In some embodiments, specifically, in the registered infrared image, the determination of whether there is an illegal item hidden is achieved by calculating the proportion of the pixels of the suspected illegal item in the pixels of the human body area. The algorithm first detects the set of pixels of the suspected illegal item and calculates the total number of pixels of these suspected illegal item pixels in the human body area. Then, it calculates the total number of pixels of the corresponding human body area pixels. By calculating the ratio of the total number of pixels of these suspected illegal item pixels to the total number of pixels of the human body area pixels, that is, the "item exposure ratio", it is determined whether there is an illegal item hidden. If this item exposure ratio is greater than the preset threshold, it is determined that there is an illegal item hidden in the human body area; if the item exposure ratio is lower than the preset threshold, it is considered that there is no illegal item hidden. This method of threshold discrimination of the item exposure ratio effectively improves the accuracy and reliability of the detection of illegal items.

[0086] Reference Figure 3 , calculate respectively Figure 3 the proportion of the black area pixels (i.e., the item area pixels) in the three red frames in the pixels of the human body area (the pixels of the entire human body area), and obtain the corresponding item exposure ratio in each frame respectively. Then, judge the size relationship between each item exposure ratio and the preset item exposure ratio threshold one by one, so as to further determine whether there is an illegal item hidden in each frame. The three red frames indicate that the human body areas of the pedestrians carry items at the three red frames. Only when the item exposure ratio of the frame exceeds the preset item exposure ratio threshold, it is determined that there is probably an illegal item hidden in the human body area at this frame. As for whether it is definitely an illegal item, further confirmation is needed.

[0087] To improve the detection accuracy, the result detected on the infrared image can also be fed back to the segmentation result of the aforementioned visible light image for verification to exclude the situation of misidentification.

[0088] Step 5. Preferably, after determining that there is an illegal item hidden in the corresponding human body area, an alarm message is automatically sent.

[0089] In some embodiments, specifically, when a potential illegal item hiding behavior is detected, the intelligent alarm mechanism will be automatically triggered to timely remind the security inspection personnel to conduct further inspections. Combining the characteristics of the infrared image and the visible light image, through the comprehensive application of the segmentation and detection models to deeply analyze the dual-light image data, it is judged whether it meets the standard of illegal item hiding. Through this multi-modal information fusion method, the misidentification of non-illegal items can be effectively excluded, and the detection accuracy can be improved.

[0090] To ensure the effectiveness and reliability of this detection method in practical applications, a 60-day actual scenario test was conducted. During the test, the detection threshold was optimized according to different environments and personnel states to minimize the situation of misidentification. Finally, while accurately identifying the binding and hiding of prohibited items, the intelligent alarm function reduced the false alarm rate and improved the overall detection effect.

[0091] This application combines infrared image monitoring and visible light monitoring. Based on the dual-light image registration technology and the dynamic adaptive detection strategy, it can monitor the waist, abdomen and leg areas of the human body in real time (it can be known from historical data that the binding and hiding of prohibited items mainly occur in the waist and legs of the human body, and the detection of prohibited items based on deep learning can also focus on these two parts. Conditions or thresholds can be set for identification according to the parts where prohibited items appear) to ensure the effective identification of the binding and hiding behavior of prohibited items under various imaging conditions. Based on the user-defined similarity alarm threshold for prohibited items, an alarm can be automatically issued, thus improving the efficiency and accuracy of intelligent security inspection.

[0092] The detection method of this application can prevent lawbreakers from using the passing crowd as a cover and mixing among passengers to engage in smuggling activities. By using the dual-light image technology to integrate the dynamic adaptive algorithm for human body binding and hiding detection, an intelligent alarm function is also designed to assist law enforcement, which has significant practical value in intelligent security inspection applications.

[0093] Figure 4 A block diagram of a device suitable for implementing the detection method described above according to an embodiment of the present application is schematically shown. Figure 4 The device shown is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0094] As Figure 4 shown, the device 1000 described in this embodiment includes: a processor 1001, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage section 1008 into the random access memory (RAM) 1003. The processor 1001 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), and so on. The processor 1001 can also include on-board memory for caching purposes. The processor 1001 can include a single processing unit or multiple processing units for performing different actions of the detection method flow according to the embodiments of the present application.

[0095] In the RAM 1003, various programs and data required for the operation of the device 1000 are stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. The processor 1001 performs various operations of the detection method process according to the embodiments of the present application by executing programs in the ROM 1002 and / or the RAM 1003. It should be noted that the programs can also be stored in one or more memories other than the ROM 1002 and the RAM 1003. The processor 1001 can also perform various operations of the detection method process according to the embodiments of the present application by executing programs stored in the one or more memories.

[0096] According to an embodiment of the present application, the device 1000 may further include an input / output (I / O) interface 1005, and the input / output (I / O) interface 1005 is also connected to the bus 1004. The device 1000 may further include one or more of the following components connected to the I / O interface 1005: an input part 1006 including a keyboard, a mouse, etc.; an output part 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage part 1008 including a hard disk, etc.; and a communication part 1009 including a network interface card such as a LAN card, a modem, etc. The communication part 1009 performs communication processing via a network such as the Internet. A driver 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the driver 1010 as needed so that a computer program read from it can be installed into the storage part 1008 as needed.

[0097] According to an embodiment of the present application, the device 1000 may further include a dedicated dual-light acquisition device K1280 for synchronously acquiring visible light images and infrared images of pedestrians passing through the detection area. After the acquisition is completed, the relevant image data can be transmitted to the processor 1001, and the process instructions of the aforementioned item binding and hiding detection method are generated and executed to achieve precise detection of illegal item binding and hiding in the human body area. Preferably, the device of the present application can actively receive the infrared radiation of an object, realize channel non-intrusive recognition, and support hoisting without occupying floor space.

[0098] The detection method flow according to the embodiments of the present application can be implemented as a computer software program. For example, embodiments of the present application include a computer program product, which includes a computer program carried on a computer-readable storage medium. The computer program contains program codes for executing the detection method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1009, and / or installed from the removable medium 1011. When the computer program is executed by the processor 1001, the above functions defined in the system of the embodiments of the present application are executed. According to the embodiments of the present application, the above-described systems, devices, apparatuses, modules, and / or units, etc. can be implemented by computer program modules.

[0099] Embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium can be included in the device / device / system described in the above embodiments, or can exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed, the steps of the detection method according to the embodiments of the present application can be implemented.

[0100] According to the embodiments of the present application, the computer-readable storage medium can be a non-volatile computer-readable storage medium. For example, it can include but is not limited to: portable computer disks, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the embodiments of the present application, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device. For example, according to the embodiments of the present application, the computer-readable storage medium can include one or more memories other than the above-described ROM 1002 and / or RAM 1003.

[0101] It should be noted that in each embodiment of the present application, the various functional modules can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product.

[0102] The flowcharts and / or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart and / or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than that noted in the accompanying drawings. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0103] Those skilled in the art will understand that the features recited in the various embodiments and / or claims of the present application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly recited in the present application. In particular, without departing from the spirit and teachings of the present application, the technical features recited in the various embodiments and / or claims of the present application can be combined and / or combined in various ways, and all such combinations and / or combinations fall within the scope of the present application.

[0104] Although the present application has been shown and described with reference to specific exemplary embodiments thereof, those skilled in the art should understand that various changes in form and detail may be made therein without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents. Therefore, the scope of the present application should not be limited to the above embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.

Claims

1. A method for detecting hidden objects, characterized in that: include: acquiring simultaneously collected visible light images and infrared images of pedestrians passing through a detection area; Performing image registration on the infrared image and the visible light image; According to the actual detection conditions, the appropriate target detection means are dynamically and adaptively selected to perform target detection on the registered infrared image; In the registered infrared image, the object exposure ratio of each object area pixel to the corresponding human body area pixel is obtained respectively, and it is determined whether the object exposure ratio of each object area is greater than a preset object exposure ratio threshold. If so, it is determined that the corresponding human body area has hidden an object.

2. The method for detecting hidden objects as claimed in claim 1, characterized in that: The dynamically and adaptively selecting a suitable target detection means to perform target detection on the registered infrared image according to the actual detection conditions includes: If the imaging distance of the registered infrared image is greater than the first preset distance threshold, the trained deep learning target detection model is used for accurate target detection; If the imaging distance of the registered infrared image is less than the second preset distance threshold, a conventional image processing method is used to perform auxiliary target detection.

3. The method for detecting hidden objects as claimed in claim 2, characterized in that: The method of obtaining the trained deep learning target detection model includes: Collect videos of people hiding objects in multiple scenes, convert the videos into pictures at preset frame rate intervals, remove duplicates, and obtain image data sets; Segment and exclude the non-human areas in the image data set, and then label the objects; After performing data enhancement processing on the image dataset, the YOLOv8 model is used for training to obtain the trained deep learning target detection model.

4. The method for detecting hidden objects as claimed in claim 1, characterized in that: Methods for determining the object exposure ratio of the object area pixels to the corresponding human body area pixels include: In the registered infrared image, the total number of pixels in each object area is obtained respectively; Obtain the total number of human body area pixels on the corresponding human body area; The object exposure ratio of the object area=(the total number of pixels in the object area) / (the total number of pixels in the human body area); If the item exposure ratio is greater than the preset item exposure ratio threshold, it is determined that the corresponding human body area hides the item; otherwise, it is determined that the corresponding human body area does not hide the item.

5. The method for detecting hidden objects as claimed in claim 1, characterized in that: The performing image registration on the infrared image and the visible light image comprises: Extracting several groups of corresponding feature points from the infrared image and the visible light image and performing normalization processing, obtaining normalized control points, constructing a matrix representing the normalized control point set, and expanding the normalized control point set into homogeneous coordinates; The inverse matrix of the matrix is ​​obtained by using the singular value decomposition method to determine the linear independence of the points. If the linear independence condition is met, the matrix calculation is continued. Based on the correspondence between the normalized matrix, the inverse matrix and the original feature point set, matrix operations are performed on the inverse matrix and the matrix to obtain a transformation matrix from the visible light image to the infrared image, and the point coordinates in the visible light image are converted to the corresponding positions in the infrared image to achieve alignment of the visible light image with the infrared image.

6. The method for detecting hidden objects as claimed in claim 3, characterized in that: After performing data enhancement processing on the image data set, training is performed using the YOLOv8 model, and obtaining the trained deep learning target detection model includes: The image dataset after data enhancement processing is divided into a training set and a test set according to a preset ratio; Input the training set into the YOLOv8 model to generate target detection results and segmentation results; In each round, the comprehensive loss value between the predicted image and the annotated image is calculated based on the loss of target detection and the loss of target segmentation; the loss calculation of target detection includes one or more of position loss, category loss and confidence loss; the loss calculation of target segmentation includes comparing the difference between the predicted segmentation mask and the true mask; Based on the comprehensive loss value, gradient backpropagation is performed to optimize the parameters of the model, and the iteration is continued until the model parameters converge to obtain a deep learning target detection model; The test set is used to evaluate the deep learning target detection model, and relevant hyperparameters are adjusted according to the evaluation results to obtain the trained deep learning target detection model.

7. The method for detecting hidden objects as claimed in claim 1, characterized in that: Also includes: After determining that the object is hidden in the corresponding human body area, an alarm message is automatically issued.

8. The method for detecting hidden objects as claimed in claim 3, characterized in that: The non-human body area includes one or more of a backpack, a shoulder bag, a hat, a skirt and a coat.

9. A device, characterized in that: The invention comprises at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the computer program is executed by the processing unit, the processing unit is enabled to perform the steps of the detection method according to any one of claims 1 to 8.

10. The device according to claim 9, characterized in that The equipment is hoisted and installed.

Images