Method and device for detecting fission track under microscope
By using the Mask R-CNN model and ellipse fitting technology, fission tracks are automatically identified, solving the problems of low efficiency and high subjectivity in traditional methods, and achieving efficient and accurate fission track analysis.
Patent Information
- Application Number
- CN202510974349.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-15
- Publication Date
- 2025-10-21
AI Technical Summary
Traditional fission track analysis mainly relies on manual identification and measurement, which is inefficient, highly subjective, and difficult to achieve efficient and accurate identification.
The Mask R-CNN model is used for fission track localization. Combined with ellipse fitting and overlap coefficient calculation, fission tracks are automatically identified and visualized.
It achieves efficient and accurate identification of fission tracks, solves the problems of low efficiency and strong subjectivity in traditional methods, and provides more reliable data support.
Smart Images
Figure CN120823600A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of geochronology, and in particular to a method and device for detecting fission tracks under a microscope. Background Art
[0002] Fission track analysis is an important tool in geochronology and thermochronology, widely used in basin evolution reconstruction, timing constraints on tectonic activity, and oil and gas resource evaluation. The length, density, and morphological characteristics of fission tracks effectively reflect the thermal history of rocks, providing critical time-temperature constraints for geological research. However, traditional fission track analysis relies primarily on manual identification and measurement, which is limited by low efficiency and high subjectivity. Therefore, there is an urgent need to achieve efficient and accurate identification of fission tracks. Summary of the Invention
[0003] The present invention provides a method and device for detecting fission tracks under a microscope, which solves the technical problem that traditional fission track analysis mainly relies on manual identification and measurement, has limitations such as low efficiency and strong subjectivity, and urgently needs to achieve efficient and accurate identification of fission tracks.
[0004] The first aspect of the present invention provides a method for detecting fission tracks under a microscope, comprising:
[0005] Acquire the image of the fission track to be measured under a microscope;
[0006] Performing fission track positioning on the fission track image to be tested using the trained Mask R-CNN model, and outputting masks and boundary coordinates to which multiple fission tracks belong;
[0007] performing ellipse fitting on the contours of the associated fission tracks using the boundary coordinates to determine the major axis and minor axis of the ellipse of each fission track;
[0008] Calculating a track overlap coefficient based on the correspondence between the major axis of the ellipse and the minor axis of the ellipse of each fission track;
[0009] The color of the associated mask is adjusted according to the comparison result of each track overlap coefficient and the overlap threshold, and each mask is mapped to the fission track image to be measured, and the marked fission track image is output.
[0010] Optionally, the calculating the track overlap coefficient based on the correspondence between the major axis of the ellipse and the minor axis of the ellipse of each fission track includes:
[0011] Performing a ratio operation on the minor axis of each ellipse and the major axis of the associated ellipse to determine the corresponding radial-axial ratio;
[0012] Calling the cv2.contourArea function to calculate the corresponding fission track contour area based on the boundary coordinates of each fission track;
[0013] Calculating the area of an ellipse based on the minor axis of each ellipse and the associated major axis of the ellipse to determine the corresponding fission track fitting area;
[0014] Performing a ratio operation on the contour area of each fission track and the corresponding fission track fitting area to generate a corresponding convexity;
[0015] The elliptical minor axis, radial-axial ratio and convexity of each fission track are respectively calculated comprehensively to generate the corresponding track overlap coefficient.
[0016] Optionally, the calculation process of the track overlap coefficient includes:
[0017] ;
[0018] in, , ;
[0019] Where, is the track overlap coefficient, is the reference threshold of fission track, is the major axis of the ellipse, is the minor axis of the ellipse, is the radial-axial ratio, is the convexity, To adjust the exponential coefficient of the convexity influence weight, is the fission track contour area, is the base of natural logarithms.
[0020] Optionally, the training process of the Mask R-CNN model includes:
[0021] Using images marked with clear confining tracks, fuzzy confining tracks, clear radius tracks, and / or fuzzy radius tracks as training fission track images;
[0022] The training fission track images are used to train the Mask R-CNN model to be trained, and the model parameters are iteratively optimized with the goal of minimizing the overall loss value. The number of identified fission tracks is counted for model verification, and the trained Mask R-CNN model is output.
[0023] Optionally, the overall loss function for calculating the overall loss value includes RPN classification loss, RPN bounding box regression loss, MRCNN classification loss, MRCNN regression loss and mask loss;
[0024] The RPN bounding box regression loss and the MRCNN regression loss adopt Huber loss;
[0025] The threshold hyperparameter in Huber loss is dynamically adjusted according to the change in loss function values between adjacent training rounds.
[0026] Optionally, the adjustment process of the threshold hyperparameter in Huber loss includes:
[0027] ;
[0028] Where, is the threshold hyperparameter for the t+1th round, is the threshold hyperparameter for round t, is the minimum value of the threshold hyperparameter, is the maximum value of the threshold hyperparameter, is the growth factor, is the reduction factor, is the training loss of the tth round, is the training loss of the t-1th round, is the loss change threshold.
[0029] A second aspect of the present invention provides a microscopic fission track detection device, comprising:
[0030] An image acquisition module is used to acquire the image of the fission track to be measured under a microscope;
[0031] A track detection module is used to locate the fission tracks on the fission track image to be tested using the trained Mask R-CNN model, and output the masks and boundary coordinates of multiple fission tracks;
[0032] A track fitting module, configured to perform ellipse fitting on the contours of the associated fission tracks using the boundary coordinates, and determine the major axis and minor axis of the ellipse of each fission track;
[0033] An overlap identification module, configured to calculate a track overlap coefficient based on the correspondence between the ellipse major axis and the ellipse minor axis of each fission track;
[0034] The track identification module is used to adjust the color of the associated mask according to the comparison result of each track overlap coefficient and the overlap threshold, map each mask to the fission track image to be measured, and output an identified fission track image.
[0035] A third aspect of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for detecting fission tracks under a microscope as described in any one of the above items.
[0036] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the method for detecting fission tracks under a microscope as described in any one of the above items.
[0037] A fifth aspect of the present invention provides a computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the method for detecting fission tracks under a microscope as described in any one of the above items.
[0038] It can be seen from the above technical solutions that the present invention has the following advantages:
[0039] The above scheme of the present invention provides a method for detecting fission tracks under a microscope, including: obtaining a fission track image to be tested under a microscope; locating the fission tracks on the fission track image to be tested by using a trained Mask R-CNN model, and outputting the masks and boundary coordinates to which multiple fission tracks belong; using each boundary coordinate to perform ellipse fitting on the contour of the associated fission track to determine the major axis and minor axis of the ellipse of each fission track; calculating the track overlap coefficient based on the major axis and minor axis of the ellipse of each fission track; adjusting the color of the associated mask according to the comparison result of each track overlap coefficient with the overlap threshold, and mapping each mask to the fission track image to be tested, and outputting an identified fission track image. Based on the above scheme, by combining the Mask R-CNN model recognition with the overlapping fission track determination of ellipse fitting, the fission track can be automatically identified and the single fission track and the overlapping fission track can be visually displayed, thereby achieving efficient and accurate identification of the fission track as a whole. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0041] Figure 1 A flowchart of the steps of a method for detecting fission tracks under a microscope provided by an embodiment of the present invention;
[0042] Figure 2 A flowchart of a method for detecting fission tracks under a microscope provided in an embodiment of the present application;
[0043] Figure 3 This is a processing flow chart of the Mask R-CNN model provided in an embodiment of the present application;
[0044] Figure 4This is a fission track annotation strategy diagram for an example embodiment of the present application;
[0045] Figure 5 This is a fission track identification effect diagram of an example embodiment of the present application;
[0046] Figure 6 This is a structural block diagram of a microscopic fission track detection device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0047] The embodiments of the present invention provide a method and device for detecting fission tracks under a microscope, which is used to solve the technical problem that traditional fission track analysis mainly relies on manual identification and measurement, has limitations such as low efficiency and strong subjectivity, and urgently needs to achieve technical problems such as efficient and accurate identification of fission tracks.
[0048] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0049] See also Figure 1 , Figure 1 A flowchart of the steps of a method for detecting fission tracks under a microscope provided by an embodiment of the present invention.
[0050] This embodiment provides a method for detecting fission tracks under a microscope, comprising:
[0051] Step 101: Acquire an image of a fission track to be measured under a microscope.
[0052] The fission track image to be tested refers to an image to be subjected to fission track detection.
[0053] It should be noted that the image acquisition system equipped with or connected to the microscope converts the visual information of the fission track of the sample to be tested observed under the microscope into a digital image, thereby obtaining the image of the fission track to be tested, so as to facilitate subsequent fission track analysis.
[0054] Step 102: Use the trained Mask R-CNN model to locate the fission track in the image to be tested, and output the masks and boundary coordinates of multiple fission tracks.
[0055] It should be noted that, in this embodiment, the trained Mask R-CNN model is used to perform instance segmentation on the fission tracks in the fission track image to be detected, thereby extracting the mask and boundary coordinates corresponding to each detected fission track.
[0056] It is understandable that the Mask R-CNN model is an extension of the Faster R-CNN model. On the basis of the original model, the RoI Align layer with bilinear interpolation replaces the RoI Pooling layer, and a fully convolutional network is used to generate masks for accurate instance segmentation to achieve accurate track contour recognition. The FPN used by Mask R-CNN can have a larger receptive field and better recognition ability for small-sized tracks, such as Figure 3 As shown: ResNet 101 is a residual network with 101 layers; FPN is a feature pyramid network; rpn_feature_maps is a feature map extracted from the feature pyramid network and used by the region proposal network (RPN) to generate candidate regions for target detection, mrcnn_feature_maps is a feature map extracted from the feature pyramid network for Mask R-CNN classification, bounding box regression and mask prediction. It can be understood that rpn_feature_maps covers more feature scales than mrcnn_feature_maps; RPN includes classification branch and regression branch, the classification branch determines whether each anchor box contains an object, and the regression branch adjusts the position and size of the anchor box; softmax is a commonly used activation function used to convert the score into a probability and make its sum 1, which is applied to classification tasks; bbox reg is a bounding box regression, which is used to predict the precise boundary coordinates for the detected area; the proposal layer is based on the probability value generated by the classification branch of the RPN and the coordinate offset generated by the regression branch. After screening, adjustment and non-maximum suppression, the candidate region proposal (region proposals); the detectiontarget layer matches the input candidate region with the real annotation box and generates the final training target; rpn_rois is the candidate box generated by the region proposal network, which contains the possible target area and is used for classification and mask prediction; FPN MASK refers to mask prediction using features extracted by the feature pyramid network; FC layers are fully connected layers used to process features in Mask R-CNN to achieve target classification and bounding box regression; RoI Align is the region of interest alignment layer used to extract aligned feature maps and unify the sizes of training targets of different sizes for subsequent classification, regression and adding masks to ensure the accuracy of detection and segmentation.
[0057] In a specific implementation of this embodiment, the training process of the Mask R-CNN model includes:
[0058] Using images marked with clear confining tracks, fuzzy confining tracks, clear radius tracks, and / or fuzzy radius tracks as training fission track images;
[0059] The training fission track images are used to train the Mask R-CNN model to be trained. The model parameters are iteratively optimized with the goal of minimizing the overall loss value. The number of identified fission tracks is counted for model verification, and the trained Mask R-CNN model is output.
[0060] It should be noted that the fission tracks identified by the Mask R-CNN model in this embodiment include confining tracks and radius tracks, which are characterized as single fission tracks, i.e., fission tracks that exist in a single independent form. When training the Mask R-CNN model, four types of track labels are defined for the fission tracks in the training fission track image, such as Figure 4 As shown in , including clear bounding paths, fuzzy bounding paths, clear radius paths and fuzzy radius paths, these labeled data are used to supervise the training of the Mask R-CNN model, enabling it to achieve multi-category instance segmentation on unseen data; Figure 2 As shown in the figure, the labeled training fission track image is input into the Mask R-CNN model to be trained for track recognition. The overall loss value is calculated based on the recognition output using the overall loss function. The model parameters are iteratively optimized with the goal of minimizing the overall loss value. The model is verified by calculating verification indicators such as precision and recall by counting the number of identified fission tracks, and the trained Mask R-CNN model is output.
[0061] In a more specific implementation of this embodiment, the overall loss function for calculating the overall loss value includes RPN classification loss, RPN bounding box regression loss, MRCNN classification loss, MRCNN regression loss and mask loss;
[0062] The RPN bounding box regression loss and MRCNN regression loss adopt Huber loss.
[0063] It should be noted that if Figure 3 As shown, the overall loss function of the Mask R-CNN model consists of five parts: RPN classification loss , RPN bounding box regression loss , MRCNN classification loss , MRCNN regression loss and mask loss , the specific calculation process can refer to the existing technology; and The smooth L1 loss is usually used, which is defined as:
[0064] (1)
[0065] Where, is the difference between the predicted value and the true value; when the error value is less than 1, it is expressed as L2 loss, and when the error value is greater than 1, it is expressed as L1 loss; in view of the large difference in fission track size, the smooth L1 loss in the regression loss is replaced by the Huber loss to improve the model's detection ability for short tracks (radius tracks). Therefore, this embodiment will and It is changed to Huber loss, which is defined as:
[0066] (2)
[0067] Where, is a threshold hyperparameter used to control error processing. Since fixed threshold hyperparameters cannot adapt well to this regression problem, in order to improve adaptability, the threshold hyperparameters in the Huber loss can be dynamically adjusted according to the change in the loss function values between adjacent training rounds. The adjustment process of the threshold hyperparameters in the Huber loss is as follows:
[0068] (3)
[0069] Where, is the threshold hyperparameter for the t+1th round, is the threshold hyperparameter for round t, is the minimum value of the threshold hyperparameter, is the maximum value of the threshold hyperparameter, is the growth factor, is the reduction factor, is the training loss of the tth round, is the training loss of the t-1th round, is the loss change threshold;
[0070] The above adjustment strategy shows that the threshold hyperparameter in the Huber loss is dynamically adjusted according to the loss changes observed during training. Specifically, when the loss fluctuates greatly, a growth factor slightly larger than 1 (for example, 1.1) is applied to gradually relax the error tolerance to enhance the robustness of the model while avoiding the introduction of drastic changes. Conversely, when the loss tends to be stable, a reduction factor slightly smaller than 1 (for example, 0.9) is used to gradually tighten the error tolerance to improve prediction accuracy while avoiding slow convergence.
[0071] Furthermore, to maintain the effectiveness of the Huber loss, the maximum value of the threshold hyperparameter is limited to 5.0. Beyond this range, the loss function will gradually approximate the mean absolute error (MAE), thereby reducing sensitivity to small residual errors. The minimum value of the threshold hyperparameter is set to 0.5 to prevent the loss function from approaching the mean squared error (MSE), which is more sensitive to outliers. This setting provides a balanced initialization scheme that takes into account both robustness and sensitivity to medium residual errors.
[0072] In addition, this embodiment can also use a loss change threshold of 0.1 to identify significant fluctuations, which means that the dynamic adjustment mechanism is triggered when the relative loss change between two adjacent epochs exceeds 10%.
[0073] Step 103: Using each boundary coordinate to perform ellipse fitting on the contours of the associated fission tracks, and determining the major axis and minor axis of the ellipse of each fission track.
[0074] It should be noted that this embodiment uses the identified fission track boundary coordinates to perform elliptical fitting on the fission track contours, determining the contour ellipse, and further extracting the corresponding ellipse major and minor axes based on each contour ellipse. It is worth mentioning that compared to circular or rectangular fitting, this embodiment considers the use of elliptical fitting to better fit the slender and variable track morphology, with higher geometric expression capabilities and technical adaptability, thereby achieving better recognition results.
[0075] In a specific implementation of this embodiment, the least squares method is used to calculate the fitting ellipse to perform ellipse fitting on the identified fission track boundary coordinates. By establishing a general ellipse equation and converting it into a matrix form, a boundary coordinate matrix D and a constraint matrix C are constructed. The ellipse fitting problem is converted into a constrained least squares optimization problem. After constructing the Lagrangian function, the eigenvalue problem is solved to obtain the ellipse parameters:
[0076] First, let's clarify that the general equation of an ellipse is:
[0077] (3)
[0078] Where, 、 、 、 、 and are all ellipse parameters, are the contour coordinates, which satisfy:
[0079] (4)
[0080] Converting formula (3) into matrix form, we can get:
[0081] (5)
[0082] Among them, the contour coordinate matrix , the ellipse parameter matrix , is the transpose of the matrix;
[0083] To simplify the calculation, formula (4) can be transformed into:
[0084] (6)
[0085] The boundary coordinate matrix D is constructed using the boundary coordinates of the fission tracks:
[0086] (7)
[0087] Where, Boundary coordinate index , is the number of boundary coordinates; this matrix is used to fit the quadratic curve describing the track shape and then extract the ellipse parameters;
[0088] Construct the constraint matrix C:
[0089] (8)
[0090] The construction of matrix C is to introduce constraints in the ellipse fitting process to ensure that the fitted quadratic curve is indeed an ellipse and not other types of quadratic curves (such as hyperbolas or parabolas);
[0091] Specifically, the ellipse satisfies the discriminant The matrix C is designed based on the quadratic expression of this discriminant, and its non-zero elements correspond to the parameters in the parameter vector. 、 、 By fitting the parameter vector Substitute the constraint form During the optimization process, the track profile is forced to satisfy the geometric characteristics of the ellipse, thus ensuring that the fitted curve is physically and mathematically reasonable. The introduction of this constraint matrix makes the fitting result more stable and consistent with the actual shape of the fission track. Therefore, the ellipse fitting problem can be expressed as:
[0092] subject to (9)
[0093] In order to find the optimal solution, construct the Lagrangian function , is the Lagrange multiplier:
[0094] (10)
[0095] Take the partial derivative of formula (8) and set it to 0:
[0096] (11)
[0097] Then the ellipse fitting problem can be transformed into the eigenvalue obtaining problem, namely:
[0098] (11)
[0099] Then, six ellipse parameters are calculated to obtain the outline ellipse. It is understood that the specific calculation process can be referred to the prior art.
[0100] Step 104: Calculate the track overlap coefficient based on the correspondence between the major axis and the minor axis of the ellipse of each fission track.
[0101] The track overlap coefficient is a quantitative indicator that characterizes the presence of overlapping fission tracks. Overlapping tracks refer to two or more single fission tracks that intersect and overlap in space.
[0102] It should be noted that, in this embodiment, the major axis and the minor axis of the ellipse of each fission track are obtained by fitting, so as to quantify the spatial intersection between the fission tracks and identify whether the fission tracks overlap.
[0103] In a specific implementation of this embodiment, step 104 includes the following sub-steps:
[0104] The ratio of the minor axis of each ellipse to the major axis of the associated ellipse is calculated to determine the corresponding radial-axial ratio;
[0105] Call the cv2.contourArea function to calculate the corresponding fission track contour area based on the boundary coordinates of each fission track;
[0106] The ellipse area is calculated based on the minor axis of each ellipse and the associated major axis of the ellipse to determine the corresponding fission track fitting area;
[0107] The ratio of each fission track contour area to the corresponding fission track fitting area is calculated to generate the corresponding convexity;
[0108] The elliptical minor axis, radial-axial ratio and convexity of each fission track are calculated comprehensively to generate the corresponding track overlap coefficient.
[0109] The fission track contour area refers to the area determined based on the contour of the identified fission track.
[0110] The fission track fitting area refers to the area determined based on the contour ellipse of the fission track fitting.
[0111] In a more specific implementation of this embodiment, the calculation process of the track overlap coefficient includes:
[0112] ;
[0113] in, , ;
[0114] Where, is the track overlap coefficient (dimensionless), is the reference threshold of the fission track (in this embodiment, the value can be 35), is the major axis of the ellipse, is the minor axis of the ellipse, is the radial-axial ratio, is the convexity, is the exponential coefficient for adjusting the convexity influence weight (in this embodiment, the value can be 1), is the fission track contour area, is the base of the natural logarithm; it can be understood that when the minor axis of the ellipse is large, the radial-axial ratio is large, and the convexity deviates from 1, it means that the identified track is likely to be the overlap of multiple tracks.
[0115] Step 105 : Adjust the color of the associated mask according to the comparison result of each track overlap coefficient and the overlap threshold, map each mask to the fission track image to be measured, and output the identified fission track image.
[0116] It should be noted that overlapping tracks are determined by setting an overlap threshold, and the mask of fission tracks with a track overlap coefficient greater than the overlap threshold (i.e., overlapping fission tracks) is marked with a first color, and the mask of fission tracks with a track overlap coefficient less than or equal to the overlap threshold (i.e., single fission tracks) is marked with a second color. After the color adjustment is completed, each mask is mapped back to the original fission track image to be tested to obtain a marked fission track image, thereby visually indicating the distribution of fission tracks. For example, when the first color is green and the second color is red, the visualization result is as follows: Figure 5 shown.
[0117] In an embodiment of the present invention, the Mask R-CNN model recognition is combined with the overlapping fission track determination based on ellipse fitting, so that fission tracks can be automatically identified and single fission tracks and overlapping fission tracks can be visually displayed, providing a solution to the difficulty in identifying short tracks and overlapping tracks. Because short tracks are fuzzy and have low contrast under a microscope, traditional manual identification methods are prone to omission or misjudgment, resulting in deviations in thermal history interpretation. The distinction between overlapping tracks is also conducive to finding confined tracks. Therefore, efficient and accurate identification of fission tracks is achieved overall, and more reliable data support can be provided for basin evolution, tectonic activity and oil and gas resource evaluation, which has important scientific significance and application value.
[0118] See also Figure 6 , Figure 6 This is a structural block diagram of a microscopic fission track detection device provided by an embodiment of the present invention.
[0119] The present invention provides a microscopic fission track detection device, comprising:
[0120] An image acquisition module 601 is used to acquire an image of the fission track to be measured under a microscope;
[0121] Track detection module 602, used to locate fission tracks on the fission track image to be detected using the trained Mask R-CNN model, and output the masks and boundary coordinates of multiple fission tracks;
[0122] A track fitting module 603 is used to perform ellipse fitting on the contours of the associated fission tracks using the boundary coordinates to determine the major axis and minor axis of the ellipse of each fission track;
[0123] An overlap identification module 604 is configured to calculate a track overlap coefficient based on the correspondence between the major axis and the minor axis of the ellipse of each fission track;
[0124] The track identification module 605 is used to adjust the color of the associated mask according to the comparison result of each track overlap coefficient and the overlap threshold, map each mask to the fission track image to be measured, and output the identified fission track image.
[0125] Optionally, calculating the track overlap coefficient based on the correspondence between the major axis and the minor axis of the ellipse of each fission track includes:
[0126] The ratio of the minor axis of each ellipse to the major axis of the associated ellipse is calculated to determine the corresponding radial-axial ratio;
[0127] Call the cv2.contourArea function to calculate the corresponding fission track contour area based on the boundary coordinates of each fission track;
[0128] The ellipse area is calculated based on the minor axis of each ellipse and the associated major axis of the ellipse to determine the corresponding fission track fitting area;
[0129] The ratio of each fission track contour area to the corresponding fission track fitting area is calculated to generate the corresponding convexity;
[0130] The elliptical minor axis, radial-axial ratio and convexity of each fission track are calculated comprehensively to generate the corresponding track overlap coefficient.
[0131] Optionally, the calculation process of the track overlap coefficient includes:
[0132] ;
[0133] in, , ;
[0134] Where, is the track overlap coefficient, is the reference threshold of fission track, is the major axis of the ellipse, is the minor axis of the ellipse, is the radial-axial ratio, is the convexity, To adjust the exponential coefficient of the convexity influence weight, is the fission track contour area, is the base of natural logarithms.
[0135] Optionally, the training process of the Mask R-CNN model includes:
[0136] Using images marked with clear confining tracks, fuzzy confining tracks, clear radius tracks, and / or fuzzy radius tracks as training fission track images;
[0137] The training fission track images are used to train the Mask R-CNN model to be trained. The model parameters are iteratively optimized with the goal of minimizing the overall loss value. The number of identified fission tracks is counted for model verification, and the trained Mask R-CNN model is output.
[0138] Optionally, the overall loss function for calculating the overall loss value includes RPN classification loss, RPN bounding box regression loss, MRCNN classification loss, MRCNN regression loss and mask loss;
[0139] The RPN bounding box regression loss and MRCNN regression loss use Huber loss;
[0140] The threshold hyperparameter in Huber loss is dynamically adjusted according to the change in loss function values between adjacent training rounds.
[0141] Optionally, the adjustment process of the threshold hyperparameter in Huber loss includes:
[0142] ;
[0143] Where, is the threshold hyperparameter for the t+1th round, is the threshold hyperparameter for round t, is the minimum value of the threshold hyperparameter, is the maximum value of the threshold hyperparameter, is the growth factor, is the reduction factor, is the training loss of the tth round, is the training loss of the t-1th round, is the loss change threshold.
[0144] An embodiment of the present invention further provides a computer device comprising a memory and a processor, wherein a computer program is stored in the memory; when the computer program is executed by the processor, the processor executes the steps of the method for detecting fission tracks under a microscope as described in any of the above embodiments.
[0145] An embodiment of the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the steps of the method for detecting fission tracks under a microscope as described in any of the above embodiments are implemented.
[0146] An embodiment of the present invention further provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for detecting fission tracks under a microscope as described in any of the above embodiments.
[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0148] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0149] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0150] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.
[0151] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0152] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for detecting fission tracks under a microscope, characterized in that: include: Acquire the image of the fission track to be measured under a microscope; Performing fission track positioning on the fission track image to be tested using the trained Mask R-CNN model, and outputting masks and boundary coordinates to which multiple fission tracks belong; performing ellipse fitting on the contours of the associated fission tracks using the boundary coordinates to determine the major axis and minor axis of the ellipse of each fission track; Calculating a track overlap coefficient based on the correspondence between the major axis of the ellipse and the minor axis of the ellipse of each fission track; The color of the associated mask is adjusted according to the comparison result of each track overlap coefficient and the overlap threshold, and each mask is mapped to the fission track image to be measured, and the marked fission track image is output.
2. The method for detecting fission tracks under a microscope according to claim 1, wherein: The calculating of the track overlap coefficient based on the correspondence between the major axis of the ellipse and the minor axis of the ellipse of each fission track includes: Performing a ratio operation on the minor axis of each ellipse and the major axis of the associated ellipse to determine the corresponding radial-axial ratio; Calling the cv2.contourArea function to calculate the corresponding fission track contour area based on the boundary coordinates of each fission track; Calculating the area of an ellipse based on the minor axis of each ellipse and the associated major axis of the ellipse to determine the corresponding fission track fitting area; Performing a ratio operation on the contour area of each fission track and the corresponding fission track fitting area to generate a corresponding convexity; The elliptical minor axis, radial-axial ratio and convexity of each fission track are respectively calculated comprehensively to generate the corresponding track overlap coefficient.
3. The method for detecting fission tracks under a microscope according to claim 1, wherein: The calculation process of the track overlap coefficient includes: ； in, , ; Where, is the track overlap coefficient, is the reference threshold of fission track, is the major axis of the ellipse, is the minor axis of the ellipse, is the radial-axial ratio, is the convexity, To adjust the exponential coefficient of the convexity influence weight, is the fission track contour area, is the base of natural logarithms.
4. The method for detecting fission tracks under a microscope according to claim 1, wherein: The training process of the Mask R-CNN model includes: Using images marked with clear confining tracks, fuzzy confining tracks, clear radius tracks, and / or fuzzy radius tracks as training fission track images; The training fission track images are used to train the Mask R-CNN model to be trained, and the model parameters are iteratively optimized with the goal of minimizing the overall loss value. The number of identified fission tracks is counted for model verification, and the trained Mask R-CNN model is output.
5. The method for detecting fission tracks under a microscope according to claim 1, wherein: The overall loss function for calculating the overall loss value includes RPN classification loss, RPN bounding box regression loss, MRCNN classification loss, MRCNN regression loss and mask loss; The RPN bounding box regression loss and the MRCNN regression loss adopt Huber loss; The threshold hyperparameter in Huber loss is dynamically adjusted according to the change in loss function values between adjacent training rounds.
6. The method for detecting fission tracks under a microscope according to claim 5, characterized in that: The adjustment process of the threshold hyperparameters in Huber loss includes: ； Where, is the threshold hyperparameter for the t+1th round, is the threshold hyperparameter for round t, is the minimum value of the threshold hyperparameter, is the maximum value of the threshold hyperparameter, is the growth factor, is the reduction factor, is the training loss of the tth round, is the training loss of the t-1th round, is the loss change threshold.
7. A fission track detection device under a microscope, characterized in that: include: An image acquisition module is used to acquire the image of the fission track to be measured under a microscope; A track detection module is used to locate the fission tracks on the fission track image to be tested using the trained Mask R-CNN model, and output the masks and boundary coordinates of multiple fission tracks; A track fitting module, configured to perform ellipse fitting on the contours of the associated fission tracks using the boundary coordinates, and determine the major axis and minor axis of the ellipse of each fission track; An overlap identification module, configured to calculate a track overlap coefficient based on the correspondence between the ellipse major axis and the ellipse minor axis of each fission track; The track identification module is used to adjust the color of the associated mask according to the comparison result of each track overlap coefficient and the overlap threshold, map each mask to the fission track image to be measured, and output an identified fission track image.
8. A computer device, characterized in that: It comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for detecting fission tracks under a microscope as described in any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method for detecting fission tracks under a microscope as described in any one of claims 1 to 6 are implemented.
10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method for detecting fission tracks under a microscope as described in any one of claims 1 to 6 are implemented.