An article image data acquisition method and system for X-ray security check

Abstract

The application relates to the technical field of image processing, in particular to an article image data acquisition method and system for X-ray security check. The method comprises the following steps: constructing an observation window for any pixel point, obtaining a consistency level of structural consistency corresponding to the pixel point in the observation window, taking a difference value between a maximum value of the structural consistency in the observation window and the consistency level, and taking an inverse tangent function processing result of the difference value as a structural anomaly significant index; optimizing an initial contrast limit factor in a CLAHE algorithm based on the structural anomaly significant index, obtaining an adaptive contrast limit factor, wherein the adaptive contrast limit factor is inversely proportional to the structural anomaly significant index of the pixel point; and enhancing a security check article image by using the CLAHE algorithm based on the adaptive contrast limit factor. The application has the effect of improving the subsequent recognition accuracy of the security check article.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method and system for acquiring image data of items for X-ray security inspection. Background Technology

[0002] X-ray security inspection systems are core technological equipment for ensuring public safety in aviation, railways, customs, and other important locations. During security checks, the system emits X-rays that penetrate baggage, which are then received by detectors to generate two-dimensional grayscale or pseudo-color images. Security personnel need to quickly and accurately determine the presence of contraband or dangerous items based on these images. However, raw X-ray images often suffer from low contrast, blurred details, and noise interference, especially when there are many items stacked tightly together, making it difficult to discern the outlines and internal structures of target items. To facilitate the differentiation of items in X-ray images, image enhancement processing is usually required. Adaptive Histogram Equalization (CLAHE) is a common image enhancement algorithm. First, the original image is divided into multiple small regions, and the local histogram corresponding to each small region is calculated, reflecting the distribution of pixel values ​​within that region. Contrast limiting is applied to each local histogram. The contrast limiting is controlled by setting a threshold (Clip Limit), which restricts the cumulative frequency of values ​​exceeding the threshold to a certain range. Local histogram equalization is then applied to make the pixel value distribution in each small region as uniform as possible, thereby improving local contrast. Output enhanced image: The final enhanced image is obtained.

[0003] However, X-ray security images have unique characteristics. They include areas of significant overlap due to the haphazard placement of items in luggage, causing information from objects of different materials and thicknesses to intertwine, creating extremely complex local content and background. They also include areas with relatively simple edge structures. The contrast limiting factor in the histogram equalization algorithm is preset based on experience by those skilled in the art. This method is a global operation applied to the entire image, uniformly enhancing the entire image. Therefore, when using traditional image enhancement algorithms to directly enhance X-ray security images, the uniform enhancement strategy may lead to the loss of local detail information or amplification of noise, thus affecting the accuracy of subsequent recognition. Summary of the Invention

[0004] To reduce the impact of artifacts in security inspection images on the accuracy of subsequent identification, this application provides a method and system for acquiring image data of items for X-ray security inspection.

[0005] Firstly, this application provides a method for acquiring image data of items for X-ray security inspection, employing the following technical solution:

[0006] A method for acquiring image data of items for X-ray security inspection is provided, which acquires preprocessed images of items to be inspected; for any pixel, the image of the item to be inspected is convolved in different directions using convolution kernels of different scales based on image frequency domain analysis technology, and the energy amplitude and phase of the pixel in different directions are obtained. Based on the energy amplitude and phase of the convolution kernels of each scale in multiple directions, the structural consistency reflecting the texture complexity of the local area where the pixel is located is calculated.

[0007] For any pixel, an observation window is constructed, and the consistency level of the structural consistency corresponding to the pixel in the observation window is obtained. The difference between the maximum value of structural consistency in the observation window and the consistency level is taken as the difference value, and the result of the arctangent function after processing the difference value is taken as the structural anomaly significance index.

[0008] The initial contrast constraint factor in the CLAHE algorithm is optimized based on the structural anomaly saliency index to obtain an adaptive contrast constraint factor, which is inversely proportional to the structural anomaly saliency index of the pixel. The CLAHE algorithm is then used to enhance the security inspection item image based on the adaptive contrast constraint factor.

[0009] By extracting energy amplitude and phase from images at multiple scales and directions through frequency domain convolution, structural consistency can be calculated, thus reflecting the response characteristics of pixels in texture structure. In images of security items, areas containing concealed objects often exhibit a high degree of structural consistency among individual pixels. Therefore, a structural anomaly saliency index can be constructed based on this feature. The initial contrast constraint factor in the traditional CLAHE algorithm can be adjusted based on this index to obtain a dynamic adaptive contrast constraint factor. This factor enhances different areas of the image to varying degrees, reducing artifacts caused by over-enhancement in certain regions of the security image and facilitating subsequent detection of security items.

[0010] Optionally, Log-Gabor wavelets can be used to convolve the images of items being inspected.

[0011] Traditional Log-Gabor wavelets are particularly suitable for structure-aware modeling of multiple materials in X-ray images due to their higher frequency resolution and zero DC component in the frequency domain.

[0012] Optionally, for any convolution direction of any scale convolution kernel, obtain the complex response value of each pixel after convolution processing of the security inspection item image; take the real part of the complex response value as the energy amplitude of the pixel in the convolution direction of the scale convolution kernel, and take the imaginary part of the complex response value as the phase of the pixel in the convolution direction of the scale convolution kernel.

[0013] By distinguishing between the real and imaginary parts in calculating energy amplitude and phase, pixel response characteristics can be more accurately reflected while maintaining spatial positioning capabilities, laying a mathematical foundation for subsequent calculations of structural consistency and structural anomaly significance index.

[0014] Optionally, the step of calculating the structural consistency reflecting the texture complexity of the local region where the pixel is located based on the energy amplitude and phase of the convolution kernel at each scale in multiple directions includes: for any pixel, calculating the same-scale response amplitude based on the energy amplitude and phase of the pixel in different convolution directions at the same scale, taking the sum of the same-scale response amplitudes at multiple scales as the total response amplitude, constructing a complex response matrix based on the complex response values ​​of the pixel at different scale convolution kernels and different convolution directions; and taking the ratio of the total response amplitude to the Shannon entropy of the modulus of the complex response values ​​in the complex response matrix as the structural consistency.

[0015] In images of items inspected at security checkpoints, if a region contains a single item, such as the straight edge of a metal plate, then the energy amplitudes of the pixels in that region are similar. During summation, these energy amplitudes do not cancel each other out, resulting in a larger response amplitude at the same scale, and consequently, a larger overall response amplitude. However, if the real and imaginary parts of the complex response matrix are no longer consistent, it indicates that the texture of the current region is complex. Consequently, during the calculation of the overall response amplitude, the different real and imaginary parts cancel each other out, leading to a smaller final calculated overall response amplitude. Simultaneously, due to the more dispersed energy, the Shannon entropy of the modulus of the complex response values ​​in the corresponding complex response matrix is ​​also larger, resulting in less structural consistency.

[0016] Optionally, the steps for calculating the same-scale response amplitude include: for any scale convolution kernel, the sum of the energy amplitudes in different convolution directions is taken as the total energy; the sum of the phases of the convolution kernel at that scale in different convolution directions is taken as the total phase; and the sum of the squares of the total energy and the total phase is taken as the same-scale response amplitude.

[0017] The energy amplitude is calculated as the sum of squares of the phase, so that the response amplitude not only reflects the intensity of the convolution energy, but also incorporates the directional characteristics of the phase structure.

[0018] Optionally, convolutional kernels of four scales are used to obtain the complex response values ​​of pixels in four convolutional directions; where the four scales are: , , , The four convolution directions are 0°, 45°, 90°, and 135°.

[0019] Optionally, for any patch formed after segmenting the security inspection item image in the CLAHE algorithm, the ratio of the initial contrast limiting factor to the mean of the structural anomaly significance index of all pixels in the patch is used as the adaptive contrast limiting factor.

[0020] Over-enhancing complex regions in security inspection images can lead to artifacts. Therefore, a more conservative enhancement is applied to complex regions, namely, the adaptive contrast limiting factor is negatively correlated with the structural heterosalience index.

[0021] Optionally, the steps for constructing the observation window include: for any pixel, constructing a square region with a preset side length centered on that pixel, and using that square region as the observation window.

[0022] Optionally, the mean of the structural consistency of each pixel in the observation window can be used as the consistency level of the corresponding pixel in the observation window.

[0023] Secondly, this application provides an image data acquisition system for X-ray security inspection, which adopts the following technical solution:

[0024] An image data acquisition system for X-ray security inspection includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an image data acquisition method for X-ray security inspection as described above is implemented.

[0025] The beneficial effect is that the above-mentioned method for acquiring image data of items for X-ray security inspection generates a computer program and stores it in the memory so that it can be loaded and executed by the processor. Thus, a system can be made based on the memory and the processor, which is convenient to use.

[0026] This application has the following technical effects:

[0027] In this application, frequency domain analysis is performed on images of security-inspected items to obtain the structural consistency of each pixel. Based on the structural consistency, a structural anomaly salience index reflecting the degree of texture anomaly in the current region is obtained. A dynamic adaptive contrast limiting factor is constructed based on the structural anomaly salience index. Based on the adaptive limiting factor, the security-inspected item image is dynamically enhanced to reduce the occurrence of artifacts caused by excessive enhancement of some regions in the security-inspected item image, which facilitates the subsequent identification and detection of security-inspected items. Attached Figure Description

[0028] Figure 1 This is a flowchart of a method for acquiring image data of items for X-ray security inspection, according to an embodiment of this application. Detailed Implementation

[0029] This application discloses a method for acquiring image data of items for X-ray security inspection. It utilizes image frequency domain analysis technology to perform convolution in different directions of the image, obtaining the energy amplitude and phase of each pixel after passing through convolution kernels of different scales and convolution directions. Then, it constructs structural consistency based on the energy amplitude and phase of each pixel to reflect the texture complexity of the region where the pixel is located. Subsequently, it constructs an observation window based on the pixels, and determines whether there are maxima of structural consistency within the observation window based on the structural consistency of the pixels, constructing a structural anomaly significance index. Finally, it optimizes the initial contrast constraint factor in the CLAHE algorithm based on the structural anomaly index to obtain a dynamic adaptive contrast constraint factor. This dynamic adaptive contrast constraint factor enhances the security inspection item image, reducing artifacts caused by over-enhancement in some areas, thus facilitating subsequent recognition and detection.

[0030] Reference Figure 1 A method for acquiring image data of items for X-ray security inspection includes steps S1-S4.

[0031] S1: Obtain pre-processed images of items to be inspected.

[0032] The X-ray security scanner captures images of the items to be inspected. The resulting images are grayscale images, where the grayscale value of each pixel reflects the degree to which the item absorbs X-rays. A higher grayscale value indicates greater X-ray absorption, such as for metals or other high-density items; a lower grayscale value indicates less X-ray absorption, such as for low-density items like plastics or liquids.

[0033] During the acquisition of X-ray security inspection images, noise may be present in the acquired images due to environmental interference and other factors. Therefore, this application uses the Wiener filtering denoising algorithm to denoise the images and uses the filtered images as the security inspection images. The denoising process of the Wiener filtering algorithm is a well-known technique and will not be described in detail here.

[0034] S2: For any pixel, the image of the security inspection item is convolved in different directions using convolution kernels of different scales based on image frequency domain analysis technology. The energy amplitude and phase of the pixel in different directions are obtained. The structural consistency reflecting the texture complexity of the local area where the pixel is located is calculated based on the energy amplitude and phase of the convolution kernels of each scale in multiple directions.

[0035] In X-ray security inspection images, items of the same material absorb X-rays to a similar degree, resulting in similar grayscale values. The edges and textures of these items exhibit a high degree of local phase consistency; that is, the phases of the Fourier components of different frequencies are highly aligned, forming a clear structural texture. However, when multiple items overlap, such as clothing, wires, or metal sheets, this phase consistency is disrupted. In the X-ray security inspection image, this area no longer appears as a single edge or corner, but rather as a chaotic superposition of multiple structural features, leading to a disordered local phase structure.

[0036] Based on the above characteristics, this embodiment calculates the energy amplitude and phase of the pixel using image frequency domain analysis technology, and constructs the structural consistency corresponding to the pixel based on this.

[0037] Specifically: In this embodiment, Log-Gabor wavelets are used to convolve the security inspection item image to obtain the complex response value of each pixel in the security inspection item image; for any convolution direction of any scale convolution kernel, the real part of the complex response value is used as the energy amplitude of the pixel in the convolution direction of the scale convolution kernel, and the imaginary part of the complex response value is used as the phase of the pixel in the convolution direction of the scale convolution kernel.

[0038] In this embodiment, the scale of the convolution kernel is set to four, namely... , , , Different convolution kernels are used to convolve the security inspection item image along different directions. In this embodiment, the convolution kernel has four convolution directions: 0°, 45°, 90°, and 135°. In other embodiments, more convolution kernels of different scales can be set, or convolution can be performed in fewer directions. The specific settings can be based on the actual application environment. For example, in scenarios with high requirements for computing speed, the number of convolution kernel scales and the number of convolution kernel directions can be appropriately reduced to improve computing efficiency.

[0039] After convolving the security inspection image with Log-Gabor wavelets, each pixel corresponds to 16 complex response values, which are the results of convolving the security inspection image with four scale kernels in four directions.

[0040] Subsequently, structural consistency reflecting the texture complexity of the local region where the pixel is located can be calculated based on the energy amplitude and phase of the convolution kernel at each scale in multiple directions. The steps include: for any pixel, calculating the same-scale response amplitude based on the energy amplitude and phase of the pixel in different convolution directions at the same scale; taking the sum of the same-scale response amplitudes at multiple scales as the total response amplitude; constructing a complex response matrix based on the complex response values ​​of the pixel at different scale convolution kernels and different convolution directions; and taking the ratio of the total response amplitude to the Shannon entropy of the modulus of the complex response values ​​in the complex response matrix as the structural consistency.

[0041] Based on the above analysis, the 16 complex response values ​​corresponding to each pixel constitute a complex response matrix, which can be represented as: ; Represents the complex response matrix of a pixel; This represents the complex response value of a pixel in the first convolution direction of the first convolution kernel at the first scale; This represents the complex response value of a pixel in the first convolution direction of the convolution kernel at the fourth scale.

[0042] The response amplitude at the same scale is calculated based on the energy amplitude and phase of pixels in different convolution directions at the same scale, and the sum of the response amplitudes at multiple scales is taken as the total response amplitude.

[0043] In one embodiment, for any scale convolution kernel, the sum of the energy amplitudes in different convolution directions is taken as the total energy; the sum of the phases of the convolution kernel at that scale in different convolution directions is taken as the total phase; and the sum of the squares of the total energy and the total phase is taken as the response amplitude at the same scale.

[0044] In a complex response matrix, data in the same row represent the complex response values ​​of the convolution kernel at the same scale under different convolution directions. Therefore, the total energy can also be understood as the sum of the real parts of the complex response values ​​in the same row of the complex response matrix, and the total phase is similar to the total energy.

[0045] In another embodiment, the same-scale response amplitude can also be calculated using the following formula: In the formula, This represents the sum of squares of pixels at the same scale. This represents the sum of energy amplitudes corresponding to multiple convolution directions under the same scale convolution kernel, and can also be understood as the sum of the energy amplitudes of the complex response matrix at the th digit. The sum of the real coefficients of the complex response value; This represents the sum of phases corresponding to multiple convolution directions under the same scale convolution kernel, and can also be understood as the sum of phases of the complex response matrix. The sum of the imaginary coefficients of the complex response value.

[0046] In this embodiment, the square root operation is performed based on the sum of squares, and the range of values ​​for the final calculation result is scaled to facilitate subsequent calculations.

[0047] The same-scale response amplitude represents the result of processing the security inspection item image by a single-scale convolution kernel. Therefore, to further reflect the texture structure on the image, multiple same-scale response amplitudes corresponding to different scales are added together to obtain the total response amplitude. Finally, the Shannon entropy of the modulus of the complex response values ​​in the complex response matrix is ​​obtained, and the ratio of the total response amplitude to the Shannon entropy is used as the structural consistency.

[0048] To prevent the denominator from being zero, a hyperparameter can be set in the denominator. In this embodiment, it is set to 1. The final formula for calculating structural consistency can be expressed as:

[0049] In the formula, Indicates the structural consistency of pixels. This represents the total response amplitude corresponding to a pixel; The Shannon entropy represents the modulus of all complex response values ​​in the complex response matrix; 1 is a hyperparameter. For example, suppose the complex response value corresponding to a certain pixel is... The energy amplitude of this pixel is 120, and the phase is 40°. The modulus of this complex response value is... .

[0050] In images of items inspected at security checkpoints, when a pixel is located on an object of a single material with a clear structure, such as the straight edge of a metal plate, the edge features of this pixel are relatively stable and consistent when observed at multiple scales from small to large. In this case, the complex response values ​​at different scales remain the same or close, corresponding to a larger sum of the real and imaginary parts of the complex response values ​​in each row of the numerator, resulting in a larger calculated value for the numerator. At the same time, since the energy is concentrated at the edge, only some complex response values ​​have a larger modulus, resulting in a smaller calculated Shannon entropy, thus a larger calculated multi-scale structural consistency index.

[0051] Conversely, when pixels are located in areas where multiple materials overlap significantly, such as tangled wires or clothing folds, small, fragmented, and chaotic edges can be observed at small scales, while at large scales, a chaotic superposition of multiple textures can be observed. That is, for different scales, the phase direction and energy amplitude of the pixels are inconsistent. This change will cause the complex response values ​​at different scales to no longer be consistent, or even opposite. Therefore, the sum of the real and imaginary parts of the complex response values ​​in each row of the molecule will be smaller due to mutual cancellation, resulting in a smaller value calculated for the molecule. At the same time, because the energy is more dispersed, the calculated Shannon entropy is larger, and therefore the calculated multi-scale structural consistency index is smaller.

[0052] S3: For any pixel, construct an observation window, obtain the consistency level of the structural consistency corresponding to the pixel in the observation window, take the difference between the maximum value of structural consistency in the observation window and the consistency level as the difference value, and take the result of the arctangent function after processing the difference value as the structural anomaly significance index.

[0053] In X-ray security inspection images, special attention should be paid to localized anomalies that differ significantly from the surrounding environment, such as contraband hidden in clothing or wrapped in food. These target items exhibit the following characteristics in X-ray security inspection images: within a complex region with a relatively uniform distribution of structural consistency index, there are one or more anomalous peak points.

[0054] First, for any given pixel, construct an observation window corresponding to that pixel, and search for pixels with the aforementioned anomalous characteristics by analyzing the structural consistency of the pixels in the observation window.

[0055] In this embodiment, a square region with a preset side length is constructed centered on the pixel, and this square region serves as the observation window. In this embodiment, the preset length is 9, meaning the side length of the observation window is 9. This length is set by those skilled in the art based on experience, and can also be set to other lengths in other embodiments.

[0056] After the observation window is constructed, the consistency level of structural consistency for all pixels within the window is calculated. In this embodiment, the mean of the structural consistency of all pixels in the observation window is used as the consistency level. The maximum value of structural consistency in the observation window is obtained, and the consistency level is subtracted from the maximum value to obtain the difference value. The larger the difference value, the more likely there are individual pixels in the observation window with significantly different structural consistency from the rest, indicating that the area where that pixel is located may belong to a region with complex structure and texture, possibly containing hidden objects. Finally, the arctangent function is used to amplify the difference value to obtain the structural anomaly significance index, improving the model's sensitivity to hidden objects.

[0057] S4: Optimize the initial contrast constraint factor in the CLAHE algorithm based on the structural anomaly significance index to obtain an adaptive contrast constraint factor, where the adaptive contrast constraint factor is inversely proportional to the structural anomaly significance index of the pixel; use the CLAHE algorithm to enhance the security inspection item image based on the adaptive contrast constraint factor.

[0058] The structural anomaly salience index represents the probability that a pixel belongs to a hidden object. Therefore, the contrast constraint factor in the traditional CLAHE algorithm can be dynamically adjusted based on this index to obtain a dynamic adaptive contrast constraint factor. During enhancement, over-enhancing structurally complex regions can produce artifacts. Therefore, in actual enhancement, regions with a high structural anomaly salience index are enhanced more conservatively; that is, the structural anomaly salience index is inversely proportional to the adaptive contrast constraint factor.

[0059] In one embodiment, during the CLAHE algorithm enhancement process, the image of the security inspection item is segmented to form multiple patches. For any patch, the ratio of the initial contrast constraint factor to the mean of the structural anomaly significance index of all pixels in the patch can be used as the adaptive contrast constraint factor.

[0060] In another embodiment, the formula for calculating the adaptive contrast limiting factor can be expressed as: In the formula, Indicates the adaptive contrast limiting factor; Indicates the initial contrast limiting factor; This represents the mean of the structural anomaly significance index of all pixels in the patch formed by segmenting the image of security-inspected items during the CLAHE algorithm enhancement process.

[0061] If the image contains complex structural information, such as clothing folds, contraband hidden in clothing or other items, then... When the value is large, a conservative enhancement strategy is needed to avoid over-enhancement and artifacts, resulting in a smaller calculated adaptive contrast limiting factor. If the structural texture information in the image patch is relatively simple, i.e. Since the value is relatively small, a stronger enhancement amplitude can be used to make the details in the image more obvious, resulting in a larger calculated adaptive contrast limiting factor.

[0062] The image of the items being inspected is used as input to the CLAHE algorithm. The adaptive contrast limiting factor for each patch, calculated based on the steps described above, replaces the initial contrast limiting factor in the traditional CLAHE algorithm. The output of the CLAHE algorithm is the enhanced X-ray image of the items being inspected. This enhanced image serves as the image acquisition result, providing security personnel with a high-quality, easily interpretable visual image, improving the efficiency and accuracy of their identification of contraband in X-ray images. The enhancement process of the CLAHE algorithm is a well-known technique and will not be elaborated upon here.

[0063] This application also discloses an image data acquisition system for X-ray security inspection, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an image data acquisition method for X-ray security inspection according to this application is implemented.

[0064] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0065] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for acquiring image data of items for X-ray security inspection, characterized in that, include: Acquire preprocessed images of security items; for any pixel, use convolution kernels of different scales in different directions based on image frequency domain analysis technology to convolve the security item image and obtain the energy amplitude and phase of the pixel in different directions; The structural consistency calculation, reflecting the textural complexity of the local region where a pixel is located, is based on the energy amplitude and phase values ​​of the convolution kernels at various scales across multiple directions. This includes: for any pixel, calculating the same-scale response amplitude based on the energy amplitude and phase values ​​of the pixel at different convolution directions at the same scale; using the sum of the same-scale response amplitudes across multiple scales as the total response amplitude; constructing a complex response matrix based on the complex response values ​​of the pixel at different scales and convolution directions; and using the ratio of the total response amplitude to the Shannon entropy of the modulus of the complex response values ​​in the complex response matrix as the structural consistency. The calculation of the same-scale response amplitude includes: for any scale convolution kernel, using the sum of the energy amplitudes at different convolution directions as the total energy; using the sum of the phases of the convolution kernel at that scale at different convolution directions as the total phase; and using the sum of the squares of the total energy and the total phase as the same-scale response amplitude. For any pixel, an observation window is constructed, and the consistency level of the structural consistency corresponding to the pixel in the observation window is obtained. The difference between the maximum value of structural consistency in the observation window and the consistency level is taken as the difference value, and the result of the arctangent function after processing the difference value is taken as the structural anomaly significance index. The initial contrast constraint factor in the CLAHE algorithm is optimized based on the structural anomaly saliency index to obtain an adaptive contrast constraint factor, which is inversely proportional to the structural anomaly saliency index of the pixel. The CLAHE algorithm is then used to enhance the security inspection item image based on the adaptive contrast constraint factor.

2. The method for acquiring image data of items for X-ray security inspection according to claim 1, characterized in that, The image of the security inspection item is convolved using Log-Gabor wavelets.

3. The method for acquiring image data of items for X-ray security inspection according to claim 1, characterized in that, For any convolution direction of any scale convolution kernel, obtain the complex response value of each pixel after convolution processing of the security inspection item image; take the real part of the complex response value as the energy amplitude of the pixel in the convolution direction of the scale convolution kernel, and take the imaginary part of the complex response value as the phase of the pixel in the convolution direction of the scale convolution kernel.

4. A method for acquiring image data of items for X-ray security inspection according to claim 3, characterized in that, The complex response values ​​of a pixel are obtained using convolutional kernels of four scales; where the four scales are: , , , The four convolution directions are 0°, 45°, 90°, and 135°.

5. A method for acquiring image data of items for X-ray security inspection according to claim 1, characterized in that, For any patch formed after segmenting the security inspection item image in the CLAHE algorithm, the ratio of the initial contrast constraint factor to the mean of the structural anomaly significance index of all pixels in the patch is used as the adaptive contrast constraint factor.

6. A method for acquiring image data of items for X-ray security inspection according to claim 1, characterized in that, The steps for constructing the observation window include: for any pixel, constructing a square region with a preset side length centered on that pixel, and using this square region as the observation window.

7. A method for acquiring image data of items for X-ray security inspection according to claim 1, characterized in that, The mean value of the structural consistency of each pixel in the observation window is taken as the consistency level of the corresponding pixel in the observation window.

8. A system for acquiring image data of items for X-ray security inspection, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for acquiring image data of items for X-ray security inspection according to any one of claims 1-7.

Images