Millimeter wave human body image denoising method and device based on image segmentation large model

Abstract

The invention provides a millimeter wave human body image denoising method and device based on an image segmentation large model, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining a first two-dimensional pixel value matrix of an original millimeter wave human body image, and inputting the first two-dimensional pixel value matrix into the image segmentation large model, obtaining a first mask matrix of the original millimeter wave human body image; performing expansion processing on the first mask matrix to obtain a second mask matrix of the original millimeter wave human body image; and performing noise reduction processing on the second mask matrix to obtain a millimeter wave human body image after noise reduction. According to the millimeter wave human body image denoising method and device based on the image segmentation large model, the noise of the background area can be effectively suppressed, the high-quality millimeter wave human body image is further obtained, and image support is provided for subsequent hazardous article detection based on the millimeter wave human body image.

Description

A method and apparatus for denoising millimeter-wave human images based on a large-scale image segmentation model. Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and more specifically, it relates to a method and apparatus for denoising millimeter-wave human images based on a large image segmentation model. Background Technology

[0002] In recent years, with increasing public focus on safety, security inspection technologies have faced more challenges. Active millimeter-wave body scanning technology, with its advantages of low radiation, high efficiency, privacy protection, and reduced labor costs, has become a research focus in the security inspection field. However, due to factors such as the transceiver performance of millimeter-wave systems and imaging algorithms, the resolution of millimeter-wave body images is relatively low, and background noise exists. Background noise can affect the detection of suspicious items in millimeter-wave images, creating detection blind spots and reducing the detection rate of suspicious items.

[0003] Therefore, how to eliminate background noise in millimeter-wave human body images and thus improve the detection performance of suspicious items is an urgent problem to be solved in the field of millimeter-wave security inspection. Summary of the Invention

[0004] The purpose of this application is to provide a method for denoising millimeter-wave human images based on a large model of image segmentation, which can effectively suppress noise in the background region and obtain high-quality millimeter-wave human images, providing image support for subsequent detection of dangerous items based on millimeter-wave human images.

[0005] A first aspect of this application provides a millimeter-wave human image denoising method based on a large image segmentation model, comprising:

[0006] Obtain the first two-dimensional pixel value matrix of the original millimeter-wave human body image, and input the first two-dimensional pixel value matrix into the large image segmentation model to obtain the first mask matrix of the original millimeter-wave human body image.

[0007] Dilation is applied to the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image;

[0008] The second mask matrix is ​​denoised to obtain a denoised millimeter-wave human body image.

[0009] A second aspect of this application provides a millimeter-wave human image denoising device based on a large image segmentation model, comprising:

[0010] The acquisition module is used to acquire the first two-dimensional pixel value matrix of the original millimeter-wave human body image, and input the first two-dimensional pixel value matrix into the large image segmentation model to obtain the first mask matrix of the original millimeter-wave human body image.

[0011] The dilation module is used to dilate the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image.

[0012] The noise reduction module is used to perform noise reduction processing on the second mask matrix to obtain a noise-reduced millimeter-wave human body image.

[0013] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described millimeter-wave human image denoising method based on a large image segmentation model.

[0014] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described millimeter-wave human image denoising method based on a large image segmentation model.

[0015] The beneficial effects of the millimeter-wave human image denoising method based on a large image segmentation model provided in this application are as follows:

[0016] First, this embodiment obtains a first two-dimensional pixel value matrix from the original millimeter-wave human body image and inputs it into a large-scale image segmentation model to obtain a first mask matrix. This leverages the powerful image segmentation capabilities of the large-scale model to accurately locate the boundary between the human body region and the background region, providing a precise basis for subsequent noise removal. Second, this embodiment dilates the first mask matrix to obtain a second mask matrix, effectively covering noise regions near the human body edges that are easily missed, avoiding residual edge noise due to overly tight segmentation boundaries. Finally, this embodiment performs noise reduction based on the second mask matrix, specifically eliminating background noise while preserving as much detail as possible in the human body region and suspicious items. This embodiment, focusing on the targeted and accurate noise removal, effectively solves the technical problems of background noise interference and low detection rate of suspicious items in millimeter-wave human body images, significantly improving the quality of millimeter-wave human body security inspection images and the accuracy and reliability of suspicious item detection. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 is a flowchart illustrating the xxxxxx method provided in an embodiment of this application;

[0019] Figure 2 is an original millimeter-wave human body image provided in an embodiment of this application;

[0020] Figure 3 shows the segmented human body region and background region provided in an embodiment of this application;

[0021] Figure 4 shows the expanded human body region and background region provided in an embodiment of this application;

[0022] Figure 5 is a denoised millimeter-wave human body image provided in an embodiment of this application;

[0023] Figure 6 is a structural block diagram of a millimeter-wave human image denoising device based on a large image segmentation model provided in an embodiment of this application;

[0024] Figure 7 is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0027] Extensive research has been conducted to denoise millimeter-wave images. This includes employing more precise imaging algorithms and improving the performance of millimeter-wave transceiver systems. However, more precise imaging algorithms often come with greater computational burden, reducing the real-time requirements of security inspection systems; improving hardware performance implies higher economic costs. Additionally, some millimeter-wave image denoising methods based on optical image improvements, such as histogram transform, frequency domain filtering, space-time filtering, and wavelet transform, are frequently used in millimeter-wave image post-processing. However, these methods are not very effective at removing image noise, and their improvement in image quality is limited.

[0028] To address the aforementioned technical issues, this application provides a millimeter-wave human image denoising method based on a large image segmentation model. This embodiment innovatively applies the general visual model (SAM) to the specific field of millimeter-wave security inspection images. Furthermore, it proposes a security image enhancement paradigm combining "human region dilation processing" and "targeted background region denoising." Traditional image denoising is performed on the entire image, which can blur or weaken the signals of dangerous items closely attached to the human body surface (especially the edges of the human body contour). This solution constructs a security protection region containing surface attachments through morphological dilation operations and selectively denoises only the background outside this region, achieving effective suppression of background clutter while completely preserving the signal integrity of potential threats.

[0029] Specifically, referring to Figure 1, the method may include:

[0030] S101: Obtain the first two-dimensional pixel value matrix of the original millimeter-wave human body image, input the first two-dimensional pixel value matrix into the large image segmentation model, and obtain the first mask matrix of the original millimeter-wave human body image.

[0031] In this embodiment, a large-scale image segmentation model can be used for image segmentation tasks. This model is the Segment Anything Model (SAM). The SAM dataset includes 11 million images and 1.1 billion segmentation masks. Because it is a large model trained on massive amounts of data, it has extremely strong generalization ability, so it can achieve millimeter-wave image segmentation without additional training. SAM supports integration with other systems, has scalable output, strong generalization ability, and the processed images lay the foundation for various downstream tasks, finding wide application in fields such as object detection, image enhancement, and medical imaging. The mask matrix output by SAM can be used as input for other tasks, providing conditions for processing specific targets or regions in images. Therefore, choosing SAM makes this embodiment simple to use and saves manpower and time costs.

[0032] Specifically, obtaining the first two-dimensional pixel value matrix of the original millimeter-wave human body image includes: obtaining the original millimeter-wave human body image; converting the original millimeter-wave human body image into a two-dimensional pixel value matrix, wherein the value of each element in the two-dimensional pixel value matrix represents the pixel value at the corresponding position in the original millimeter-wave human body image.

[0033] In this embodiment, all images that need to be segmented are first organized and placed in a folder. Then, the OpenCV library in Python is used to implement the function of batch reading files. Since the original millimeter-wave human body image is a grayscale image, a two-dimensional pixel value matrix is ​​obtained after reading the image.

[0034] For example, the raw millimeter-wave human image can be read in grayscale mode using cv2.IMREAD_GRAYSCALE from the OpenCV library in Python, obtaining the first two-dimensional pixel value matrix D1. Each element represents the grayscale value of the corresponding pixel, ranging from 0 to 255. Let the original millimeter-wave human image size be rows × cols; then the two-dimensional pixel matrix D1 is a two-dimensional matrix of size rows × cols.

[0035] Furthermore, the first two-dimensional pixel value matrix is ​​input into the large image segmentation model to obtain the first mask matrix of the original millimeter-wave human body image, including: performing image segmentation on the first two-dimensional pixel value matrix to obtain the human body region and the background region; and generating the first mask matrix of the original millimeter-wave human body image based on the human body region and the background region. The elements in the first mask matrix are represented by the following formula.

[0036]

[0037] Where (i, j) represents the position of each element in the first mask matrix, rows represents the number of rows in the first mask matrix, and cols represents the number of columns in the first mask matrix.

[0038] In this embodiment, after completing the batch reading of millimeter-wave human images and obtaining the pixel value matrix (first two-dimensional pixel value matrix), the next step is to use SAM to extract the human body region and background region in the image. First, a suitable SAM pre-trained model is selected. The SAM pre-trained model is trained on massive amounts of image data of different types and has powerful feature recognition and region segmentation capabilities. Therefore, it can complete the millimeter-wave human image segmentation task without additional training. Figure 2 shows the original millimeter-wave human image provided in an embodiment of this application. Since there is only one human target in an image, the human body occupies the core position in the image, and there is no mutual occlusion between multiple targets. Therefore, after SAM segmentation, only one mask matrix M1 will be obtained. M1 is a binary matrix, where the position of 1 in the matrix represents the human body region, and the position of 0 in the matrix represents the background region. The schematic diagram after segmentation is shown in Figure 3, where the white part represents the human body region and the black part is the background region.

[0039] S102: Dilate the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image.

[0040] In this embodiment, considering the complexity and application requirements of actual security inspection scenarios, when using millimeter-wave human body security inspection technology to screen dangerous items, dangerous items carried by the human body may take various forms, with some potentially hanging on the outside of the body. In this case, improper image noise reduction processing may blur or eliminate the imaging information of dangerous items hanging on the outside of the body while filtering out noise, causing significant problems for subsequent work. Therefore, to ensure effective removal of background noise while preserving complete imaging information of dangerous items on the outside of the body, dilation processing of the human body area is required. Based on the original mask, a dilation algorithm is used to appropriately extend the human body area outward.

[0041] Specifically, the first mask matrix is ​​dilated to obtain the second mask matrix of the original millimeter-wave human body image. This includes: using the elements corresponding to the human body region in the first mask matrix as the origin, a cross-dilation method is used to dilate the second mask matrix.

[0042] For example, the expansion of the human body area can be performed using rectangular expansion, circular expansion, or cross expansion. In this embodiment, the cross expansion method is chosen because it expands the human body area mainly along the axial direction (up, down, left, and right), which can maintain the separation of these critical gaps to the greatest extent, thereby more accurately defining the protected area of ​​"human body and its surface", ensuring the integrity of the threat signal, and conforming to the distribution characteristics of human body and dangerous goods.

[0043] In one embodiment of this application, a second mask matrix is ​​obtained by using the element corresponding to the human body region in the first mask matrix as the origin and performing a cross-dilation method. This includes: determining the number of pixels extending into the background region corresponding to each pixel in the human body region, where each pixel in the human body region corresponds to an element in the first mask matrix; using each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending into the background region corresponding to each pixel in the human body region, and the first mask matrix, performing a cross-dilation method to obtain the second mask matrix.

[0044] Specifically, experiments have shown that, for example, on an image with a resolution of 440×840 pixels, it is reasonable to expand outward by 30 to 50 pixels, meaning that the number of pixels that each pixel in the human body region extends into the background region can be 30-50.

[0045] Furthermore, taking each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending from each pixel in the human body region to the background region, and the first mask matrix, a cross-dilation method is used to dilate the mask matrix to obtain a second mask matrix. This includes: taking each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending from each pixel in the human body region to the background region, and using a cross-dilation method, determining the cross-shaped structural element corresponding to each element; and based on the cross-shaped structural element corresponding to each element and the first mask matrix, determining the second mask matrix using the following formula.

[0046]

[0047] in, This is the result of N being symmetric about the origin. yes The result after translation by z, where z is the dimension to be translated;

[0048] N={(p, q)||p|≤n, |q|≤n, (p=0∨q=0)}

[0049] Where N is the cross structure matrix corresponding to each element, the size of the cross structure matrix is ​​(2n+1)×(2n+1), n ​​represents the number of pixels that extend outward from the cross with the element as the origin in the first mask matrix, the coordinates of the center element of the cross structure matrix are (0, 0), and (p, q) represent the element positions of the cross structure matrix.

[0050] For example, N is a 3×3 cross-structure matrix. Transforming the coordinates (p, q) of each element in N into (-p, -q) results in a new structural element. z is a translation vector, corresponding to the pixel position offset in the image. The entire structure is moved along the direction corresponding to z (e.g., horizontally or vertically), resulting in a new structural element. The second mask matrix M2 can be interpreted as: finding all "translated" masks. "The sets of positions z that overlap with the first mask matrix M1 are the expanded second mask matrix M2."

[0051] For example, when n=1, N is a 3×3 cross-shaped structuring element, which can be represented as

[0052]

[0053] From the matrix above, we can see that N represents the row where p = 0 or the column where q = 0, where all elements are 1, and all elements outside that row and column are 0. The first mask matrix is ​​M1. After cross-expansion, we obtain a new mask matrix (the second mask matrix) M2. The expanded second mask matrix M2 can be simplified as follows:

[0054]

[0055] Figure 4 shows the expanded human body area and background area provided in an embodiment of this application. The white part represents the expanded human body area, and the black part represents the background area.

[0056] S103: Perform noise reduction processing on the second mask matrix to obtain a noise-reduced millimeter-wave human body image.

[0057] Specifically, the second mask matrix is ​​denoised to obtain a denoised millimeter-wave human body image, including: determining elements that meet preset conditions based on the first and second mask matrices; setting the values ​​of the elements that meet the preset conditions to 0 in the second mask matrix to obtain an adjusted second mask matrix; determining a second two-dimensional pixel matrix based on the adjusted second mask matrix; and converting the second two-dimensional pixel value matrix into an actual image based on an image processing library to obtain a denoised millimeter-wave human body image; wherein the preset condition is that elements at the same position are 0 in the first mask matrix and 1 in the second mask matrix.

[0058] In this embodiment, in the second mask matrix M2, the portion with a value of 1 represents the dilated human body region, and the portion with a value of 0 represents the background region. By setting the values ​​of elements that meet preset conditions to 0, background noise is eliminated, and the denoised image pixel value matrix (second two-dimensional pixel value matrix) D2 is obtained. The second two-dimensional pixel matrix is ​​represented as follows:

[0059]

[0060] Specifically, the second two-dimensional pixel value matrix D2 can be interpreted as follows: the element with a value of 1 in the second mask matrix D2 has the same pixel value as the element at the same position in the first two-dimensional pixel value matrix D1, and the element with a value of 0 in the second mask matrix D2 has a value of 0 in the second two-dimensional pixel value matrix D2.

[0061] Figure 5 shows the denoised millimeter-wave human body image represented by the second two-dimensional pixel value matrix D2, which is a denoised millimeter-wave human body image provided in an embodiment of this application.

[0062] Thus, this embodiment separates the human body region from the background region in a millimeter-wave human body image using SAM, and performs noise suppression processing on the background region without affecting the imaging of the human body region, thereby improving the quality of the millimeter-wave human body image.

[0063] As can be seen from the above, firstly, this embodiment obtains the first two-dimensional pixel value matrix of the original millimeter-wave human body image and inputs it into the image segmentation large model to obtain the first mask matrix. This leverages the powerful image segmentation capabilities of the large image segmentation model to accurately locate the boundary between the human body region and the background region, providing a precise basis for subsequent noise removal. Secondly, this embodiment performs dilation processing on the first mask matrix to obtain the second mask matrix, effectively covering noise regions near the human body edge that are easily missed, avoiding the problem of residual edge noise due to overly tight segmentation boundaries. Finally, this embodiment performs noise reduction processing based on the second mask matrix, specifically eliminating interference noise in the background region while preserving the detailed information of the human body region and suspicious items to the greatest extent. This embodiment, starting from the targeted and accurate nature of noise removal, effectively solves the technical problems of background noise interference and low detection rate of suspicious items in millimeter-wave human body images, significantly improving the quality of millimeter-wave human body security inspection images and the accuracy and reliability of suspicious item detection. Furthermore, in this embodiment, the mask matrices of the human body region and the background region can be obtained separately through SAM, thus obtaining the coordinates of the human body region and the background region in the image. Compared to traditional millimeter-wave image denoising methods, this embodiment can suppress background noise without affecting the human body area, achieving precise noise reduction. Furthermore, this embodiment also enables the construction of an automated data augmentation and preprocessing workflow for millimeter-wave security threat detection. This workflow can generate standard training samples with controllable background noise and lossless threat target signals in batches, effectively improving the generalization performance and recognition accuracy of downstream hazardous materials detection models.

[0064] Corresponding to the millimeter-wave human image denoising method based on a large image segmentation model in the above embodiments, Figure 6 is a structural block diagram of a millimeter-wave human image denoising device based on a large image segmentation model provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiments of this application are shown. Referring to Figure 6, the millimeter-wave human image denoising device 20 based on a large image segmentation model includes: an acquisition module 21, a dilation module 22, and a noise reduction module 23.

[0065] Among them, the acquisition module 21 is used to acquire the first two-dimensional pixel value matrix of the original millimeter-wave human body image, input the first two-dimensional pixel value matrix into the image segmentation large model, and obtain the first mask matrix of the original millimeter-wave human body image.

[0066] Dilation module 22 is used to dilate the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image;

[0067] The noise reduction module 23 is used to perform noise reduction processing on the second mask matrix to obtain a noise-reduced millimeter-wave human body image.

[0068] In one embodiment of this application, when acquiring the first two-dimensional pixel value matrix of the original millimeter-wave human body image, the acquisition module 21 is specifically used for:

[0069] Acquire raw millimeter-wave human images;

[0070] The original millimeter-wave human body image is converted into a two-dimensional pixel value matrix, where the value of each element in the two-dimensional pixel value matrix represents the pixel value at the corresponding position in the original millimeter-wave human body image.

[0071] In one embodiment of this application, when the acquisition module 21 inputs the first two-dimensional pixel value matrix into the image segmentation large model to obtain the first mask matrix of the original millimeter-wave human body image, it is specifically used for:

[0072] Image segmentation is performed on the first two-dimensional pixel value matrix to obtain the human body region and the background region;

[0073] Based on the human body region and the background region, a first mask matrix of the original millimeter-wave human body image is generated. The elements in the first mask matrix are characterized by the following formula.

[0074]

[0075] Where (i, j) represents the position of each element in the first mask matrix, rows represents the number of rows in the first mask matrix, and cols represents the number of columns in the first mask matrix.

[0076] In one embodiment of this application, when the dilation module 22 performs dilation processing on the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image, it is specifically used for:

[0077] Using the elements corresponding to the human body region in the first mask matrix as the origin, the second mask matrix is ​​obtained by performing a cross-expansion process.

[0078] In one embodiment of this application, when the dilation module 22 performs dilation using a cross-dilation method with the elements corresponding to the human body region in the first mask matrix as the origin to obtain the second mask matrix, it is specifically used for:

[0079] Determine the number of pixels that extend into the background region corresponding to each pixel in the human body region. Each pixel in the human body region corresponds to an element in the first mask matrix.

[0080] Using each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending into the background region corresponding to each pixel in the human body region, and the first mask matrix, a cross-dilation method is used to perform dilation processing to obtain the second mask matrix.

[0081] In one embodiment of this application, when the dilation module 22 performs dilation processing using a cross-dilation method with each element corresponding to the human body region in the first mask matrix as the origin, based on the number of pixels extending from each pixel in the human body region to the background region, and the first mask matrix, to obtain the second mask matrix, it is specifically used for:

[0082] Taking each element corresponding to the human body region in the first mask matrix as the origin, based on the number of pixels extending to the background region corresponding to each pixel in the human body region, and using the cross dilation method, the cross structure element corresponding to each element is determined.

[0083] Based on the cross-shaped structural element corresponding to each element and the first mask matrix, the second mask matrix is ​​determined by the following formula;

[0084]

[0085] in, This is the result of N being symmetric about the origin. yes The result after translation by z, where z is the dimension to be translated;

[0086] N=[(p, q)||p|≤n, |q|≤n, (p=0∨q=0)}

[0087] Where N is the cross structure matrix corresponding to each element, the size of the cross structure matrix is ​​(2n+1)×(2n+1), n ​​represents the number of pixels that extend outward from the cross with the element as the origin in the first mask matrix, the coordinates of the center element of the cross structure matrix are (0, 0), and (p, q) represent the element positions of the cross structure matrix.

[0088] In one embodiment of this application, when the noise reduction module 23 performs noise reduction processing on the second mask matrix to obtain a noise-reduced millimeter-wave human body image, it is specifically used for:

[0089] Based on the first mask matrix and the second mask matrix, determine the elements that satisfy the preset conditions;

[0090] Set the values ​​of elements that meet the preset conditions to 0 in the second mask matrix to obtain the adjusted second mask matrix, and determine the second two-dimensional pixel matrix based on the adjusted second mask matrix;

[0091] Based on the image processing library, the second two-dimensional pixel value matrix is ​​converted into an actual image to obtain a denoised millimeter-wave human body image;

[0092] The preset condition is that elements at the same position are 0 in the first mask matrix and 1 in the second mask matrix.

[0093] Referring to Figure 7, which is a schematic block diagram of an electronic device provided in an embodiment of this application, the electronic device 300 in this embodiment, as shown in Figure 7, may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 are used to store computer programs, which include program instructions. The processors 301 are used to execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to execute the functions of the modules in the above-described device embodiments, such as the functions of the acquisition module 21, the expansion module 22, and the noise reduction module 23 shown in Figure 6.

[0094] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), but it may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0095] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0096] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory.

[0097] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the millimeter-wave human image denoising method based on large image segmentation model provided in the embodiments of this application, or they can execute the implementation method of the electronic device described in the embodiments of this application, which will not be repeated here.

[0098] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0099] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0100] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0101] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0102] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces or units, or they may be electrical, mechanical, or other forms of connection.

[0103] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0104] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0105] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A millimeter-wave human image denoising method based on a large image segmentation model, characterized in that, include: Obtain the first two-dimensional pixel value matrix of the original millimeter-wave human body image, and input the first two-dimensional pixel value matrix into the image segmentation large model to obtain the first mask matrix of the original millimeter-wave human body image; Dilation is applied to the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image; The second mask matrix is ​​denoised to obtain a denoised millimeter-wave human body image.

2. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 1, characterized in that, The step of obtaining the first two-dimensional pixel value matrix of the original millimeter-wave human body image includes: obtaining the original millimeter-wave human body image; converting the original millimeter-wave human body image into a two-dimensional pixel value matrix, wherein the value of each element in the two-dimensional pixel value matrix represents the pixel value at the corresponding position of the original millimeter-wave human body image.

3. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 1, characterized in that, The step of inputting the first two-dimensional pixel value matrix into the image segmentation large model to obtain the first mask matrix of the original millimeter-wave human body image includes: performing image segmentation on the first two-dimensional pixel value matrix to obtain human body region and background region; generating the first mask matrix of the original millimeter-wave human body image based on the human body region and background region, wherein the elements in the first mask matrix are characterized by the following formula. Where (i, j) represents the position of each element in the first mask matrix, rows represents the number of rows in the first mask matrix, and cols represents the number of columns in the first mask matrix.

4. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 1, characterized in that, The step of performing dilation processing on the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image includes: using the element corresponding to the human body region in the first mask matrix as the origin, performing dilation processing using a cross dilation method to obtain the second mask matrix.

5. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 4, characterized in that, The step of using the element corresponding to the human body region in the first mask matrix as the origin and performing dilation using a cross-dilation method to obtain the second mask matrix includes: determining the number of pixels extending into the background region corresponding to each pixel in the human body region, wherein each pixel in the human body region corresponds to an element in the first mask matrix; and using each element corresponding to the human body region in the first mask matrix as the origin, based on the number of pixels extending into the background region corresponding to each pixel in the human body region, and the first mask matrix, performing dilation using a cross-dilation method to obtain the second mask matrix.

6. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 5, characterized in that, The step of using each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending from each pixel in the human body region to the background region, and the first mask matrix, and performing dilation using a cross-dilation method to obtain the second mask matrix includes: using each element corresponding to the human body region in the first mask matrix as the origin, and based on the number of pixels extending from each pixel in the human body region to the background region, and performing dilation using a cross-dilation method to determine the cross-shaped structural element corresponding to each element; and based on the cross-shaped structural element corresponding to each element and the first mask matrix, determining the second mask matrix using the following formula. in, This is the result of N being symmetric about the origin. yes The result after translation z, where z is the size to be translated; N = {(p, q) ||p|≤n, |q|≤n, (p = 0 ∨ q = 0)} where N is the cross structure matrix corresponding to each element, the size of the cross structure matrix is ​​(2n+1) × (2n+1), n ​​represents the number of pixels that extend outward from the cross with the element as the origin in the first mask matrix, the coordinates of the center element of the cross structure matrix are (0, 0), and (p, q) represent the element position of the cross structure matrix.

7. The millimeter-wave human image denoising method based on a large image segmentation model as described in claim 3, characterized in that, The step of denoising the second mask matrix to obtain a denoised millimeter-wave human body image includes: determining elements that satisfy preset conditions based on the first mask matrix and the second mask matrix; setting the values ​​of the elements that satisfy the preset conditions to 0 in the second mask matrix to obtain an adjusted second mask matrix; determining a second two-dimensional pixel matrix based on the adjusted second mask matrix; and converting the second two-dimensional pixel value matrix into an actual image based on an image processing library to obtain the denoised millimeter-wave human body image; wherein, the preset conditions are that elements at the same position are 0 in the first mask matrix and 1 in the second mask matrix.

8. A millimeter-wave human image denoising device based on a large image segmentation model, characterized in that, include: The acquisition module is used to acquire the first two-dimensional pixel value matrix of the original millimeter-wave human body image, and input the first two-dimensional pixel value matrix into the image segmentation large model to obtain the first mask matrix of the original millimeter-wave human body image. The dilation module is used to dilate the first mask matrix to obtain the second mask matrix of the original millimeter-wave human body image; The noise reduction module is used to perform noise reduction processing on the second mask matrix to obtain a noise-reduced millimeter-wave human body image.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images