Method for Evaluating Image Quality of Muon Imaging Reconstruction

Abstract

The present application relates to a method for evaluating the quality of a reconstructed image in muon imaging, which is used to directly compare the reconstructed image with a continuous target model for evaluating the quality of the reconstructed image. The method includes: setting a target model, the target model having a determined geometric shape and density distribution; selecting an actual target structure in the target model, the actual target structure being the target area for evaluating the quality of the reconstructed image; setting one or more detectors for measuring the muon tracks of cosmic ray muons penetrating the target model; reconstructing a three-dimensional density image of the target model based on the muon tracks; segmenting the three-dimensional density image to extract the reconstructed target structure; calculating the intersection over union of the actual target structure and the reconstructed target structure to characterize the quality of the reconstructed image. The technical solution of the present application can directly compare the reconstructed image with a continuous target model for evaluating the quality of the reconstructed image, without discretizing the target model and without causing signal loss.

Description

Technical Field

[0001] This application relates to the field of image recognition, and particularly to a method for evaluating the quality of reconstructed images in muon imaging. Background Art

[0002] Cosmic ray muon imaging (referred to as "muon imaging" for short) uses naturally occurring cosmic ray muons to achieve imaging of the internal structure of an object to be measured. Muon imaging can be divided into two categories: an imaging method that images the internal mass density structure of an object to be measured based on the attenuation of the muon flux after penetrating the object to be measured is called muon absorption imaging; an imaging method that identifies an object to be measured with a high atomic number based on the change in the muon direction before and after penetrating the object to be measured is called muon scattering imaging. Muon imaging has the following unique advantages: high energy (the average energy of cosmic ray muons at sea level is about 3 - 4 GeV), strong penetration, and naturally occurring without the need for protection, making it an important supplement to traditional non-destructive detection techniques.

[0003] The final output of a radiation imaging system is a reconstructed image, and what the industry is most concerned about is also the quality of the reconstructed image. The evaluation of the quality of the reconstructed image requires a set of recognized and reliable evaluation methods. Using an object to be measured with completely known material and structure in a radiation imaging system is called a target model. By comparing the similarity between the reconstructed image and the target model, the quality of the reconstructed image can be evaluated. The higher the similarity, the better the quality of the reconstructed image is considered.

[0004] There are many such evaluation methods in X-ray imaging, such as mean squared error (MSE), peak signal-to-noise ratio (PSNR), structural similarity (SSIM), etc. However, the physical mechanisms and target types of X-ray imaging and muon imaging are very different, and the methods suitable for evaluating the quality of X-ray imaging reconstruction images may not be suitable for evaluating the quality of muon imaging reconstruction images. Among them, the imaging principles of muon scattering imaging and X-ray imaging are very different. Muon scattering imaging is based on the change in direction of the ray before and after it penetrates the object to be measured, while X-ray imaging is based on the intensity attenuation of the ray after it penetrates the object to be measured. Although both muon absorption imaging and X-ray imaging are based on the intensity attenuation (flux attenuation) of the radiation after it penetrates the object to be measured, the light intensity in X-ray imaging decays exponentially with the thickness of a single medium, while the attenuation of the muon flux is related to the energy loss of the muon and the shape of the natural cosmic ray muon energy spectrum. In general, it does not decay exponentially with the thickness of a single medium; X-ray imaging uses an artificial radiation source with an adjustable energy spectrum, while muon absorption imaging uses a high-energy natural cosmic ray muon source with an unadjustable energy spectrum; the counting rate of natural cosmic ray muons is low, which leads to the problems of statistical fluctuations and visual angle sparsity, which usually appear simultaneously in muon absorption imaging; since the energy of natural cosmic ray muons is much higher than that of the X-rays used in X-ray imaging, the targets to be measured in muon absorption imaging are usually much larger than those to be measured in X-ray imaging, and therefore the imaging effects pursued by the two are also different. X-ray imaging pursues the accurate reconstruction of fine structures, while muon absorption imaging pursues more the identification of abnormal structures in large-scale objects.

[0005] Therefore, it is urgent to propose a method for evaluating the quality of muon imaging reconstruction images that is different from the X-ray imaging reconstruction image quality evaluation method, so as to verify or evaluate the image quality by identifying the muon imaging reconstruction image and processing the image features in the feature space. Summary of the invention

[0006] In order to at least partially solve one or more technical problems mentioned in the background technology, the present application provides a method for evaluating the quality of reconstructed images of muon imaging. The similarity between the target structure in the reconstructed image and the target structure in the target model is evaluated by using the intersection-over-union method extended to continuous sets to evaluate the quality of the reconstructed image of muon imaging.

[0007] The present application provides a method for evaluating the quality of a reconstructed image in muon imaging, which is used to directly compare the reconstructed image with a continuous target model for evaluating the quality of the reconstructed image. The method includes: setting a target model, where the target model has a determined geometric shape and density distribution; selecting an actual target structure in the target model, and the actual target structure is the target area for evaluating the quality of the reconstructed image; setting one or more detectors for measuring the muon tracks of cosmic ray muons penetrating the target model; reconstructing a three-dimensional density image of the target model based on the muon tracks; segmenting the three-dimensional density image to extract the reconstructed target structure; calculating the intersection over union of the actual target structure and the reconstructed target structure to characterize the quality of the reconstructed image.

[0008] In some embodiments, the intersection over union of the actual target structure and the reconstructed target structure is

[0009]

[0010] where the vector represents the proportion of the overlapping part of each voxel with the actual target structure, and the vector represents the proportion of the overlapping part of each voxel with the reconstructed target structure, is the voxel number, n is the total number of voxels.

[0011] In some embodiments, the element of the vector is the proportion of the overlapping part of the th voxel with the actual target structure, and has three possible values:

[0012] = 0, the th voxel is outside the actual target structure;

[0013] = 1, the th voxel is inside the actual target structure;

[0014] = q q q q q <1, the

[0015] In some embodiments, the element of the vector is the proportion of the overlapping part of the th voxel with the reconstructed target structure, and has two possible values:

[0016] = 0, the voxel is outside the reconstructed target structure;

[0017] = 1, the voxel is inside the reconstructed target structure.

[0018] In some embodiments, , the larger it is, the higher the quality of the reconstructed image.

[0019] In some embodiments, the density of the actual target structure is different from the density of the part outside the actual target structure in the target model.

[0020] In some embodiments, the step of segmenting the three-dimensional density image is based on the density value of the voxel.

[0021] In some embodiments, the step of segmenting the three-dimensional density image includes: setting a density value range, determining whether the density value of each voxel in the three-dimensional density image belongs to the density value range, and if it belongs, the voxel belongs to the reconstructed target structure, otherwise the voxel does not belong to the reconstructed target structure.

[0022] In some embodiments, for muon absorption imaging, the steps of reconstructing the three-dimensional density image of the target model based on the muon track include: calculating the muon flux distribution based on the muon track; calculating the minimum energy required for muons incident in different ( ) directions to penetrate the target model, where is the zenith angle, is the azimuth angle; solving the mass thickness of muons incident in different ( ) directions penetrating the target model according to the minimum energy; dividing the target model or the region of the target model containing the actual target structure into a plurality of voxels, and solving the mass density value of each voxel according to the mass thickness.

[0023] In some embodiments, for muon scattering imaging, the steps of reconstructing the three-dimensional density image of the target model based on the muon track include: calculating the angle between the muon track before penetrating the target model and the muon track after penetrating the target model; calculating the square of the angle based on the angle ; dividing the target model or the region of the target model containing the actual target structure into a plurality of voxels, and solving the scattering density of each voxel according to the square of the angle value.

[0024] The technical solution of the present application is applicable to the characteristics of the quality evaluation of muon imaging reconstruction images. It can directly compare the reconstructed images with the continuous target model for the quality evaluation of the reconstructed images without discretizing the target model, and no signal loss will be caused. Through the muon imaging reconstruction image quality evaluation method of the present application, it is possible to focus on evaluating the imaging effect of the target structure in the target model and avoid the influence of other unimportant structures in the target model on the evaluation index of the imaging result. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By referring to the accompanying drawings and reading the following detailed description, the above and other objects, features, and advantages of the exemplary embodiments of the present application will become readily understood. In the drawings, several embodiments of the present application are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 A schematic flowchart showing the muon imaging reconstruction image quality evaluation method according to an embodiment of the present application;

[0027] Figure 2 A schematic diagram showing a cosmic ray muon imaging system according to an embodiment of the present application;

[0028] Figure 3 A schematic diagram showing the reconstruction image quality evaluation process according to an embodiment of the present application;

[0029] Figure 4 A schematic diagram showing the calculation of the intersection over union of the target structure according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0031] It should be understood that the terms "first", "second", "third", and "fourth", etc. in the claims, specifications, and drawings of the present application are used to distinguish different objects, rather than to describe a specific order. The terms "including" and "comprising" used in the specifications and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0032] It should also be understood that the terms used in the specification of this application are merely for the purpose of describing specific embodiments and are not intended to limit this application. As used in the specification and claims of this application, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should be further understood that the term "and / or" used in the specification and claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0033] As used in this specification and the claims, the term "if" can be interpreted as "when", "once", "in response to determining", or "in response to detecting" depending on the context. Similarly, the phrase "if determined" or "if [the described condition or event] is detected" can be interpreted as meaning "once determined", "in response to determining", "once [the described condition or event] is detected", or "in response to detecting [the described condition or event]" depending on the context.

[0034] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0035] This application conducts research on the method for evaluating the quality of reconstructed images in muon imaging, and evaluates the similarity of the geometric structures between the target structures in the reconstructed images and the target structures in the target model in muon imaging based on the Jaccard similarity index. However, the traditional Jaccard similarity index method is only applicable to the case where the target model and the reconstructed image have the same image resolution. In actual situations, the object to be measured is often a complex continuous set. Therefore, the target model of the complex continuous set is closer to the real physics. If the traditional Jaccard similarity index method is used to evaluate the reconstructed image of the target model of the complex continuous set, the target model needs to be discretized according to the reconstructed image resolution, which will cause signal loss and lead to distorted evaluation results. The technical solution of this application is applicable to the characteristics of evaluating the quality of reconstructed images in muon imaging, and can directly compare the reconstructed image with the continuous target model for evaluating the quality of the reconstructed image without discretizing the target model and without causing signal loss.

[0036] Figure 1 A schematic flowchart showing the method for evaluating the quality of reconstructed images in muon imaging according to an embodiment of this application is shown.

[0037] In step S102, a target model is set, and the target model has a determined geometric shape and density distribution.

[0038] In step S104, an actual target structure is selected in the target model, and the actual target structure is the target area for evaluating the quality of the reconstructed image.

[0039] Preferably, the density of the actual target structure is different from the density of the part outside the actual target structure in the target model.

[0040] In step S106, one or more detectors are set up to measure the muon tracks of cosmic ray muons penetrating the target model.

[0041] In step S108, a three-dimensional density image of the target model is reconstructed based on the muon flux distribution.

[0042] Optionally, for muon absorption imaging, step S108 includes: calculating the muon flux distribution based on the muon tracks; calculating the minimum energy required for muons incident in different ( ) directions to penetrate the target model, where is the zenith angle, is the azimuth angle; solving the mass thickness of the target model penetrated by muons incident in different ( ) directions according to the minimum energy; dividing the target model or the region of the target model containing the actual target structure into a plurality of voxels, and solving the mass density value of each voxel. A voxel is short for Volume Pixel, which is the smallest unit for the segmentation of digital data in three-dimensional space. Voxels are used in fields such as three-dimensional imaging and detection imaging, and conceptually are similar to the smallest unit pixel in two-dimensional space.

[0043] Optionally, for muon scattering imaging, step S108 includes: the steps of reconstructing a three-dimensional density image of the target model based on the muon tracks include: calculating the angle between the muon tracks before and after penetrating the target model ; calculating the square of the angle based on the angle ; dividing the target model or the region of the target model containing the actual target structure into a plurality of voxels, and solving the scattering density of each voxel according to the square of the angle value.

[0044] In step S110, the three-dimensional density image is segmented to extract the reconstructed target structure.

[0045] Preferably, the three-dimensional density image is segmented based on the density values of the voxels. The specific segmentation steps include, for example: setting a density value range, and determining whether the density value of each voxel in the three-dimensional density image belongs to the density value range. If it belongs, the voxel belongs to the reconstructed target structure; otherwise, the voxel does not belong to the reconstructed target structure. The result of image segmentation is related to the selection of the density value range. One optimization method is to select different density value ranges within a reasonable range, calculate the corresponding image segmentation results, calculate the image quality evaluation index for each image segmentation result, and finally take the image segmentation result corresponding to the highest image quality evaluation index value as the final image segmentation result.

[0046] In step S112, the intersection over union of the actual target structure and the reconstructed target structure is calculated to characterize the quality of the reconstructed image.

[0047] Preferably, the intersection over union of the actual target structure and the reconstructed target structure is

[0048]

[0049] where the vector represents the proportion of the overlapping part of each voxel with the actual target structure, and the vector represents the proportion of the overlapping part of each voxel with the reconstructed target structure, is the voxel number, n is the total number of voxels.

[0050] Optionally, the element of the vector is the proportion of the overlapping part of the th voxel with the actual target structure, which has three possible values:

[0051] = 0, the th voxel is outside the actual target structure;

[0052] = 1, the th voxel is inside the actual target structure;

[0053] = q , the th voxel partially overlaps with the actual target structure, and the proportion of the overlapping part in the th voxel is q , 0 < q < 1.

[0054] Optionally, the element of the vector is the The proportion of the overlapping part of each voxel and the reconstructed target structure There are two possible values:

[0055] = 0, the voxel is outside the reconstructed target structure;

[0056] = 1, the voxel is inside the reconstructed target structure.

[0057] Calculated according to the above , the larger it is, the higher the quality of the reconstructed image.

[0058] Those skilled in the art clearly understand that although Figure 1 the step sequence numbers are shown, their sequence is not limited to the illustrated order. For example, step S104 can be carried out after step S106 or after step S108, and so on.

[0059] Figure 2 shows a schematic diagram of a cosmic ray muon imaging system according to an embodiment of the present application.

[0060] Cosmic ray muon imaging is divided into two categories: (a) cosmic ray muon absorption imaging and (b) cosmic ray muon scattering imaging. Cosmic ray muon scattering imaging utilizes the physical principle that muons change direction due to multiple Coulomb scatterings when penetrating matter. Using natural cosmic rays as the radiation source, a position-sensitive detector is used to measure the muon tracks before and after penetrating the object to be measured. According to the muon tracks before and after penetrating the object to be measured, the scattering angle of the muons in the object to be measured and the tracks in the object to be measured are determined. Then, combined with the Coulomb scattering physical law of muons in matter, the scattering angle and the tracks in the object to be measured are converted into the scattering density at different positions inside the object to be measured, realizing the imaging of the scattering density inside the object to be measured. Cosmic ray muon absorption imaging utilizes the physical principle that muons lose energy and cause flux attenuation due to interactions when penetrating matter. Using natural cosmic rays as the radiation source, a position-sensitive detector is used to measure the muon flux distribution after penetrating the object to be measured. By comparing with the muon flux in the open sky, the degree of flux attenuation caused by the object to be measured is calculated. Then, combined with the energy loss law of muons in matter, the measured muon flux is converted into the mass thickness of muons penetrating the object to be measured in different directions reaching the detector. Further combined with an image reconstruction algorithm, the mass thickness can be converted into the three-dimensional mass density distribution of the object to be measured, realizing the imaging of the density structure inside the object to be measured. The cosmic ray muon imaging system includes a cosmic ray muon source, a detector, an image reconstruction algorithm, etc. The imaging system here is not limited to the imaging system in the experiment, nor can it be the imaging system in the simulation.

[0061] Referring to the foregoingFigure 1 As described in step S106, the detector in the cosmic ray muon imaging system measures the track of the incident muons. The track of the muons penetrating the detector is approximately a straight line, which can be described by the direction ([[]], ( ) ( is the zenith angle, is the azimuth angle) and the point (x, y, z) on the track. The measured muon track is converted into the input quantity for image reconstruction:

[0062] For muon absorption imaging: the input quantity is the mass thickness of the muons penetrating the object to be measured incident in different ( ) directions .

[0063] Count the muons in different ( ) directions, calculate the muon flux I ([[]]) :

[0064]

[0065] where is the effective area of the detector for the muons incident in the ([[]]) direction, is the solid angle size of the corresponding incident direction, is the measurement duration for measuring muons with the measured quantity of .

[0066] According to the theoretical muon differential energy spectrum formula calculate the minimum energy required for the muons incident in different ( ) directions to penetrate the object to be measured:

[0067]

[0068] According to the relationship between the mass thickness of the muons penetrating the object to be measured and the minimum energy required for the muons to penetrate the object to be measured: , solve the mass thickness of the muons incident in different ( ) directions penetrating the object to be measured.

[0069] For muon scattering imaging: the input quantity is the square of the angle between the muon tracks before and after penetrating the object to be measured.

[0070] Calculate the angle between the muon tracks before and after penetrating the object to be measured. According to Moliere theory, the planar scattering angle of muon small-angle scattering approximately follows a Gaussian distribution with an expected value of 0 and a variance of ,

[0071]

[0072] where represents the thickness of the object to be measured in units of radiation length ; represents the momentum of the muon; represents the velocity of the muon. The square of the measured angle can be approximated .

[0073] Furthermore, referring to the steps described in step S108 in the previous text Figure 1 , a three-dimensional density image is reconstructed. It may specifically include the following steps:

[0074] Define the region of interest (ROI). This region completely covers the target structure within the target model that is of most concern. Generally, it is considered that the input quantity for reconstructing the muon imaging image based on muon data is only related to this region, or the density distribution outside this region is completely known. Image reconstruction of the object to be measured is performed within the ROI. The ROI is divided into n discrete voxels, and each voxel has a single density value. For absorption imaging, the density value is the mass density (defined as mass divided by volume); for scattering imaging, the density value is the scattering density (defined as / L). The reconstructed image is to solve for the density value of each voxel. According to the physical mechanism of muon imaging, a system of equations can be established between the input quantity for image reconstruction and the density value.

[0075] For absorption imaging, use the vector to represent the density of different voxels, and the vector to represent the mass thickness of muons incident in different ( ) directions penetrating the object to be measured. There is

[0076]

[0077] For scattering imaging, use the vector to represent the scattering density of different voxels, and the vector to represent the square of the angle between the muon tracks measured before and after penetrating the object to be measured. There is

[0078]

[0079] Use a three-dimensional image reconstruction algorithm to solve this system of equations, and the distribution of the obtained density values is the reconstructed image.

[0080] Figure 3 A schematic diagram showing the evaluation process of the reconstructed image quality according to an embodiment of the present application.

[0081] The output of the cosmic ray muon imaging system is a reconstructed image. The higher the quality of the reconstructed image, the more accurately it can reflect the internal information of the object to be measured. To evaluate the quality of the reconstructed image, it is necessary to compare the reconstructed image with the object to be measured that is completely known. The higher the similarity between the reconstructed image and the object to be measured, the better the quality of the reconstructed image. The object to be measured that is completely known for evaluating the quality of the reconstructed image is called the target model. Similarly, the target model can be an entity used in a muon imaging experiment or a digital model constructed in a muon imaging simulation. For muon imaging, there are significant differences in the density values (mass density or scattering density ) of a small part of the target model and the density values (mass density or scattering density ) of other regions. This part of the region is usually also the most concerned region in muon imaging, called the target structure (i.e., the actual target structure) in the target model, such as Figure 2 . What muon imaging is most concerned about is the imaging effect of the imaging system on the geometric position of the target structure, that is, it is necessary to evaluate the geometric similarity between the target structure in the reconstructed image and the target structure in the target model.

[0082] As Figure 3 shown, the method for evaluating the quality of the muon imaging reconstructed image of the present application includes two parts: image segmentation of the target structure in the reconstructed image and calculation of the intersection over union of the target model and the target structure in the reconstructed image.

[0083] For the target model, its structure must be completely known, so the geometric position of the target structure in the target model is obvious. For the reconstructed image, if we want to accurately evaluate the reconstruction effect of the target structure, we first need to perform image segmentation on the reconstructed image to extract the reconstructed target structure from it.

[0084] Referring to the description in step S110 above Figure 1 in the application scenario of muon imaging, the significant feature of the target structure is that it has a mass density value (muon absorption) or scattering density different from most regions in the target model Value (muon scattering), so for the reconstructed image, image segmentation can be performed according to the density value of the voxel. A feasible image segmentation method is to set a density value interval corresponding to the target structure, and determine whether the density value of each voxel in the reconstructed image belongs to this interval. If it belongs, the voxel belongs to the target structure; otherwise, the voxel does not belong to the target structure. By traversing all voxels, image segmentation of the target structure in the reconstructed image can be achieved. The result of image segmentation is related to the selection of the density value interval. An optimization method is to select different density value intervals within a reasonable range, calculate the corresponding image segmentation results, calculate the image quality evaluation index for each image segmentation result, and finally take the image segmentation result corresponding to the highest image quality evaluation index value as the final image segmentation result.

[0085] Figure 4 Shows a schematic diagram for calculating the intersection over union of the target structure according to an embodiment of the present application.

[0086] Referring to the previous text Figure 1 As described in step S112 above, use the intersection over union method extended to the continuous set to evaluate the quality of the reconstructed image of the target structure, and define the vector representing the proportion of the overlapping part of each voxel with the target structure in the target model, The element of is the proportion of the overlapping part of the th voxel with the target structure in the target model,

[0087] = 0, the th voxel is outside the actual target structure;

[0088] = 1, the th voxel is inside the actual target structure;

[0089] = q , the th voxel partially overlaps with the actual target structure, and the proportion of the overlapping part in the th voxel is q , 0 < q < 1.

[0090] Define the vector representing the proportion of the overlapping part of each voxel with the target structure in the reconstructed image, The element of is the proportion of the overlapping part of the th voxel with the target structure in the reconstructed image,

[0091] = 0, the voxel is outside the reconstructed target structure;

[0092] = 1, the voxel is inside the reconstructed target structure.

[0093] At this time, the intersection over union (IoU) of the target model and the reconstructed image can be defined as

[0094]

[0095] The traditional IoU method considers that the target structure in the target model is a set composed of several voxels P , and at the same time, the target structure in the reconstructed image is a set composed of several voxels R , and the IoU J = , where is the number of elements in the intersection of and is the number of elements in the union of and J . Therefore, the traditional IoU method requires the target model and the reconstructed image to have the same image resolution. However, when the target model is not composed of voxel in the shape of small squares (the density at any point inside a single small square is the same), but in the form of a continuous set, specifically, it is composed of an infinite number of points, and there is a density value for any given three-dimensional coordinates (x, y, z), and any different coordinates may correspond to any different density values, the traditional IoU method cannot be used to calculate

[0096] Figure 4 Schematically, the target model, the reconstructed image, the actual target structure in the target model, and the reconstructed target structure in the reconstructed image are shown in two-dimensional form respectively. In actual calculations, it is in three-dimensional form. The IoU of the target model and the reconstructed image calculated according to the foregoing method , the larger it is, the higher the geometric similarity between the reconstructed target structure and the target structure in the target model, and the better the reconstruction effect.

[0097] Although the technical solution of this application is directed to three-dimensional images, those skilled in the art can understand that the technical solution of this application can also be applied to two-dimensional images, regarded as three-dimensional images with one dimension being 1, where voxels are simplified to pixels. For muon absorption imaging, the pixel value is the mass thickness, and for muon scattering imaging, the pixel value is the variance of the scattering angle distribution.

[0098] The above has introduced the embodiments of this application in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, based on the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A method for evaluating the quality of a muon imaging reconstruction image, characterized in that: The method is used to directly compare the reconstructed image with the continuous target model to evaluate the quality of the reconstructed image, and includes: Setting the target model, wherein the target model has a determined geometric shape and density distribution; Selecting an actual target structure in the target model, wherein the actual target structure is a target area for reconstructing image quality evaluation; One or more detectors are provided to measure the muon tracks of cosmic ray muons penetrating the target model; reconstructing a three-dimensional density image of the target model based on the muon track; Segmenting the three-dimensional density image to extract a reconstructed target structure; The intersection-over-union ratio between the actual target structure and the reconstructed target structure is calculated to characterize the quality of the reconstructed image. for Among them, the vector Represents the proportion of each voxel overlapping the actual target structure, the vector Represents the proportion of each voxel overlapping with the reconstructed target structure, is the voxel number, n is the total amount of voxels, the vector Elements For the The proportion of the overlap between the voxel and the actual target structure, the vector Elements For the The percentage of overlap between a voxel and the reconstructed target structure.

2. The method for evaluating the image quality of muon imaging reconstruction according to claim 1, characterized in that: Said There are three possible values: =0,th The voxels are outside the actual target structure; =1, No. The voxels are within the actual target structure; = q , No. The voxel partially overlaps with the actual target structure, and the overlapping part is The proportion of individual pixels is q , 0< q <1.

3. The method for evaluating the image quality of muon imaging reconstruction according to claim 1, characterized in that: Said There are two possible values: =0,th The voxel is outside the reconstruction target structure; =1, No. The voxels are within the reconstructed target structure.

4. The method for evaluating the image quality of muon imaging reconstruction according to claim 1, characterized in that: , The larger the value, the higher the quality of the reconstructed image.

5. The method for evaluating the image quality of muon imaging reconstruction according to claim 1, characterized in that: The density of the actual target structure is different from the density of a portion of the target model outside the actual target structure.

6. The method for evaluating the quality of muon imaging reconstruction images according to claim 5, characterized in that: The step of segmenting the three-dimensional density image is based on density values ​​of voxels.

7. The method for evaluating the image quality of muon imaging reconstruction according to claim 6, characterized in that: The step of segmenting the three-dimensional density image includes: setting a density value interval, determining whether the density value of each voxel in the three-dimensional density image belongs to the density value interval, if so, the voxel belongs to the reconstructed target structure, otherwise the voxel does not belong to the reconstructed target structure.

8. The method for evaluating the image quality of muon imaging reconstruction according to claim 1, characterized in that: For muon absorption imaging, the step of reconstructing the three-dimensional density image of the target model based on the muon track includes: Calculating a muon flux distribution based on the muon track; Based on the muon flux distribution, different ( ) direction incident muons need to penetrate the target model, where is the zenith angle, is the azimuth; According to the minimum energy solution, different ( ) direction, and penetrates the mass thickness of the target model; The target model or the region of the target model containing the actual target structure is divided into a plurality of voxels, and the mass density of each voxel is solved according to the mass thickness. value.

9. The method for evaluating the quality of muon imaging reconstruction images according to claim 1, characterized in that: For muon scattering imaging, the step of reconstructing the three-dimensional density image of the target model based on the muon track includes: Calculate the angle between the muon tracks before and after penetrating the target model ; Based on the angle Calculate the square of the angle ; The target model or the region of the target model containing the actual target structure is divided into a plurality of voxels, and the Solve for the scattering density of each voxel value.

Images