Rail damage judgment method and device based on large model

Abstract

The present application relates to a kind of based on large model's rail damage judgment method and device, belong to artificial intelligence technical field, the method first according to the rail damage result judged according to ultrasonic signal feature, then further judge according to the rail damage judgment large model trained by historical data, finally, the result for clutter in the result of large model judgment is detected and corrected to clutter, obtains the final rail damage judgment result.The method and device described in the present application can improve the efficiency and accuracy of rail damage judgment.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, and in particular relates to a method and device for judging rail damage based on a large model. Background Technology

[0002] With the expansion of the high-speed rail network and the increase in operating speed, the accurate identification of rail damage has become crucial. Traditional manual inspection and simple image processing techniques suffer from low efficiency, limited accuracy, and inconsistent damage assessment standards when dealing with complex damage types such as core defects, bolt hole damage, and railbed anomalies. Summary of the Invention

[0003] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method, device, and storage medium for judging rail damage based on a large model. Using the method of this invention, the judgment standard can be unified, improving the accuracy and efficiency of rail damage judgment.

[0004] In a first aspect, the present invention proposes a method for judging rail damage based on a large model, comprising:

[0005] The damage to the rail is determined based on the characteristics of the ultrasonic signal, and a first judgment result is obtained;

[0006] The rail damage judgment model is trained by training a pre-set large model based on historical rail damage data. The first judgment result is then input into the rail damage judgment model to obtain the second judgment result.

[0007] The second judgment result is corrected to obtain the final injury judgment result.

[0008] Furthermore, based on the characteristics of the ultrasonic signals, the damage to the rail is determined, and the first judgment results include:

[0009] Define the rectangular defect detection area and move it according to a fixed step size S;

[0010] Each step of movement extracts the image of the damage detection area and obtains the initial damage detection result based on the ultrasonic signal characteristics. The initial damage detection result includes no damage, core damage, bolt hole damage, and rail bottom anomaly. The rectangular damage detection area has a height of H and a width of W.

[0011] Acquire images of the damaged area;

[0012] The acquisition of the damaged area image includes:

[0013] When the initial damage assessment result is no damage, continue the detection of the next detection area;

[0014] When the initial damage assessment result is nuclear damage, the damaged area image is determined according to the first method;

[0015] When the initial damage assessment result is screw hole damage, the damage area image is determined according to the second method;

[0016] When the initial damage assessment result is an anomaly at the track bottom, the image of the damaged area is determined according to the third method.

[0017] The first, second, and third methods are not entirely the same.

[0018] A second aspect of the present invention provides an apparatus for implementing a rail damage assessment method based on a large model, comprising:

[0019] The initial assessment module is configured to determine the damage to the rail based on the characteristics of the ultrasonic signal and obtain the first assessment result;

[0020] The large model damage assessment module is configured to train a preset large model based on historical rail damage assessment data to obtain a rail damage assessment large model, and input the first assessment result into the rail damage assessment large model to obtain a second assessment result;

[0021] The correction module is configured to correct the damage result of the second judgment result to obtain the final damage result.

[0022] In a third aspect, the present invention provides an apparatus for implementing a rail damage assessment method based on a large model, comprising: a memory, a processor, and a user interface;

[0023] The memory is used to store computer programs;

[0024] The user interface is used to interact with the user;

[0025] The processor is used to read the computer program in the memory, and when the processor executes the computer program, it implements the above-mentioned rail damage judgment method based on a large model.

[0026] In a fourth aspect, the present invention provides a processor-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the above-described rail damage assessment method based on a large model.

[0027] The beneficial effects of this invention are as follows:

[0028] By combining deep learning and computer vision, this invention first performs ultrasonic-based rail damage assessment, then rail damage assessment based on a large model, and finally corrects the assessment results. This approach accurately identifies and classifies various types of damage, unifies the standards for damage assessment, and improves the accuracy and efficiency of damage assessment. Using this method, the development trend of rail damage can also be predicted, providing data support for maintenance decisions and thus improving railway safety and maintenance efficiency. Attached Figure Description

[0029] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. It is obvious that the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings.

[0030] Figure 1 This is a flowchart of a rail damage assessment method based on a large model, according to an embodiment of the present invention.

[0031] Figure 2 This is a schematic diagram of the initial assessment process for rail damage according to an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the clutter correction process according to an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of a rail damage assessment device based on a large model according to an embodiment of the present invention;

[0034] Figure 5 This is a schematic diagram of another rail damage assessment device based on a large model according to an embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0036] Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts disclosed in this invention.

[0037] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The terms "installed," "connected," and "linked" should be interpreted broadly; for example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0038] Damage assessment is a shorthand for assessing the extent of damage, and is used to determine the damage to rails.

[0039] Damage results, that is, the results of rail damage, include no damage, core damage, bolt hole damage, and rail base anomalies.

[0040] Damage types, that is, the types of rail damage, include core damage, bolt hole damage, and rail base anomalies.

[0041] Nuclear damage refers to metal fatigue damage to the contact surface of the rail head caused by long-term wheel rolling. It manifests as spalling, cracks, or dents on the rail surface, and usually occurs in the center of the rail head, distributed like a "core," hence the name nuclear damage.

[0042] Bolt hole damage refers to cracks or damage appearing around the bolt holes used to fix fishplate connections on the track. This damage is mainly caused by metal fatigue due to repeated stress around the bolt holes, or by improper bolt installation or poor tightening.

[0043] Rail base anomalies refer to damage or defects appearing at the bottom of the rail. These mainly include rail base cracks, corrosion, and wear. The rail base bears the weight and bending stress of the rail; any anomalies there will seriously affect the overall strength and safety of the track.

[0044] The damage detection area is the area where the rail is inspected for damage.

[0045] Damage area image, that is, when damage is determined to exist, the image of the damage detection area is obtained.

[0046] MiniCPM is a large language model.

[0047] In this invention, unless otherwise specified, the unit of length is millimeters.

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of methods and systems consistent with some aspects of the invention as detailed in the appended claims.

[0049] This invention provides a method for judging rail damage based on a large model, specifically, as follows: Figure 1 As shown, the method includes steps S101 to S103:

[0050] S101. Based on the characteristics of ultrasonic signals, determine the damage to the rail and obtain the first judgment result;

[0051] S102. Train the preset large model based on the historical rail damage assessment data to obtain a rail damage assessment large model, and input the first assessment result into the rail damage assessment large model to obtain the second assessment result.

[0052] S103. Correct the damage assessment result of the second judgment result to obtain the final damage assessment result.

[0053] In this invention, step S101 is called initial damage assessment. The damage detection area is a rectangular window that moves from left to right along the width of the rail image. The window has a height of H and a width of W, and moves at a fixed step size S. Each step extracts an image of the damage detection area and obtains the rail damage result determined based on the ultrasonic signal characteristics; this is called the initial damage assessment result. If the initial damage assessment result shows damage, an image of the damaged area is acquired for further assessment using a larger model. Specifically, as shown... Figure 2 As shown, it includes:

[0054] S201. Determine the rectangular defect detection area and move it according to a fixed step size S;

[0055] S202. For each step of movement, extract the image of the damage detection area and obtain the initial damage assessment result based on the ultrasonic signal characteristics. The initial damage assessment result includes no damage, core damage, bolt hole damage, and rail bottom anomaly.

[0056] S203. Obtain the image of the damaged area based on the initial damage assessment results.

[0057] In this step, according to the different types of initial damage assessment results, different methods are used to obtain images of the damaged area. The first, second, and third methods for obtaining the damaged area images are not entirely the same.

[0058] S203-1: If the initial damage assessment result is no damage, there is no need to acquire an image of the damaged area; continue with the detection of the next detection area. That is, move one step size S to detect the next area.

[0059] S203-2: If the initial damage assessment result is nuclear damage, then the damaged area image is determined according to the first method. Specifically, it could be:

[0060] Obtain the height h corresponding to the nuclear damage. i The damaged area O that needs to be inspected is obtained. i :

[0061]

[0062] in dh1 represents the horizontal cut-off length corresponding to the nuclear damage. i =max(0,h) i -Δh1) is the lower bound for the cutoff corresponding to the nuclear damage, uh1 i =min(H,h) i +Δh1) is the upper limit of the cut-off corresponding to the nuclear damage, Δh1 is the longitudinal cut-off length corresponding to the nuclear damage, w1 and Δh1 are multiples of 10, and the unit is millimeters;

[0063] when At that time, add a block with width W and height 2Δh1-uh1 above the intercepted area. i +dh1 i The black areas form the image of the damaged area;

[0064] when At that time, add a block with width W and height 2Δh1-uh1 below the intercepted area. i +dh1 i The black areas form the image of the damaged area.

[0065] Where i is the number of the rectangular defect detection region, starting with 0 and increasing by 1 with each step. For example, at the beginning of detection, the number of the rectangular defect detection region is 0. After detection is completed and the region moves one step S, the number of the rectangular defect detection region becomes 1, and so on.

[0066] S203-3: If the initial damage assessment result is screw hole damage, then determine the damage area image according to the second method. Specifically, it could be:

[0067] Obtain the height h corresponding to the screw hole damage iThe damaged area O that needs to be inspected is obtained. i ;

[0068]

[0069] in dh2 is the transverse section length corresponding to the screw hole damage. i =max(0,h) i -Δh2) is the lower bound for the cutoff corresponding to the screw hole damage, uh2 i =min(H,h) i +Δh2) is the upper limit of the cut corresponding to the screw hole damage, Δh2 is the longitudinal cut length corresponding to the screw hole damage, and w2 and Δh2 are multiples of 10;

[0070] when At that time, add a block with width W and height 2Δh²-uh² above the intercepted area. i +dh2 i The black areas form the image of the damaged area;

[0071] when At that time, add a block with width W and height 2Δh²-uh² below the intercepted area. i +dh2 i The black areas form the image of the damaged area.

[0072] S203-3: If the initial damage assessment result is an anomaly at the track bottom, then the damaged area image is determined according to the third method. Specifically, it could be:

[0073] Obtain the height h corresponding to the track bottom anomaly i The damaged area O that needs to be inspected is obtained. i ;

[0074]

[0075] in dh3 represents the lateral cut-off length corresponding to the track bottom anomaly. i =max(0,h) i -Δh3) is the lower bound for the intercept corresponding to the track bottom anomaly, uh3 i =min(H,h) i +Δh3) is the upper limit of the cut-off corresponding to the rail bottom anomaly, Δh3 is the longitudinal cut-off length corresponding to the rail bottom anomaly, and w3 and Δh3 are multiples of 10;

[0076] when At that time, add a block with width W and height 2Δh3-uh3 above the intercepted area. i +dh3 i The black area represents the image of the damaged area;

[0077] when At that time, add a block with width W and height 2Δh3-uh3 below the intercepted area. i +dh3 i The black area represents the image of the damaged area.

[0078] Preferably, in S102, a large-scale rail damage judgment model is obtained by training a preset large-scale model based on historical rail damage assessment data. The first judgment result is then input into the large-scale rail damage judgment model to obtain a second judgment result, including:

[0079] S102-1: Model Training: A large-scale damage assessment model is established based on the MiniCPM model. The training dataset includes damage images and a damage description dataset, which includes the damage type and whether it is clutter. The MiniCPM model is trained using the training dataset, and the model's performance is evaluated. Based on the evaluation results, the model is optimized to obtain the large-scale damage assessment model.

[0080] S102-2: Model-based damage assessment: A large-scale rail damage assessment model is trained based on historical rail damage assessment data. The first assessment result is then input into this model to obtain a second assessment result. Specifically, the damaged area image obtained in step 101 and the initial damage assessment result are input into the model to obtain the second assessment result. The second assessment result includes whether the damage corresponding to the damaged area image is clutter.

[0081] In other words, the second judgment result includes a specific damage type or clutter, that is, the second judgment result is one of the following: no damage, nuclear damage, bolt hole damage, rail bottom anomaly or clutter;

[0082] When the second judgment result is clutter, clutter detection and correction are performed to obtain the final damage judgment result.

[0083] Preferably, in S103, correcting the damage assessment result of the second judgment result to obtain the final damage assessment result may include:

[0084] When the second judgment result is clutter, clutter detection and correction are performed, specifically as follows: Figure 3 As shown, it includes S301, S302 and S303:

[0085] S301, Damage feature extraction.

[0086] The damaged area image is converted into a grayscale image. First, the area is divided into L equal parts according to its width, and then into L equal parts according to its height, resulting in L*L rectangular areas of the same size, where L is an integer greater than or equal to 5. For example, the area could be divided into 10 equal parts according to its width, and then into 10 equal parts according to its height, resulting in 100 rectangular areas of the same size.

[0087] The average pixel value for each rectangular region is calculated to be g. j,k Where j is the index of the rectangular region from left to right horizontally, and k is the index of the rectangular region from top to bottom vertically, the feature set {g} of the damaged region is obtained. j,k}, where j and k are integers from 0 to L-1.

[0088] Construct the characteristic matrix Where ts1 is the set first judgment threshold, and i is the sequence number of the damaged area image, which is also the number of the rectangular damage detection area.

[0089] For example, when L = 10, the feature matrix of the i-th damaged area image is:

[0090]

[0091] S302, Damage Clustering.

[0092] Clustering the feature matrix yields m clusters, and the cluster center of each cluster is extracted and denoted as . dp j,k These are the elements corresponding to the center matrix, where j and k are integers from 0 to L-1. For example, when L=10, the cluster center of each cluster is denoted as:

[0093]

[0094] The feature matrix D i With each cluster center ΔD q Similarity calculation is performed to obtain the classification feature score gd q :

[0095]

[0096] For example, when L = 10, gd q The expression is:

[0097]

[0098] Select score gd q The highest-ranking cluster is used to classify the detection region;

[0099] S303, Damage Correction;

[0100] Obtain the historical verification count gt of the damaged area image. q Number of historical misjudgments q ;

[0101] when At that time, acquire the historical damage records n of the detection area.i With historical clutter record n1 i ,when When the detected area image is determined to be clutter, the detected area image is added to the cluster's tag and stored; when If manual verification is performed and the verification result is noise, then the historical verification count gt q Increase by 1, historical misjudgment count ge q The number of historical verifications remains unchanged; if the verification result indicates damage, then the number of verifications is greater than or equal to gt. q Increase by 1, historical misjudgment count ge q Add 1; where the cluster's label includes the damage type; where a cluster is a collection of detection areas for a specific damage type, including core damage, bolt hole damage, or railbed anomaly. For example, all detection areas with core damage as the first cluster, all detection areas with bolt hole damage as the second cluster, and all detection areas with railbed anomaly as the third cluster.

[0102] when At that time, the image is sent to designated personnel for manual review. If the review result is noise, the historical verification count is gt. q Increase by 1, historical misjudgment count ge q The number of historical verifications remains unchanged; if the verification result indicates damage, then the number of verifications is greater than or equal to gt. q Increase by 1, historical misjudgment count ge q Increase by 1;

[0103] Where ts2 is the set second judgment threshold, and ts3 is the set third judgment threshold;

[0104] After clutter correction following S301, S302, and S303, the final rail damage assessment result is obtained, including core damage, bolt hole damage, rail bottom anomaly, or clutter.

[0105] The method of this invention first determines the rail damage based on ultrasonic signal characteristics, then further assesses it using a large-scale rail damage assessment model trained on historical data, and finally corrects any clutter findings in the model's assessment to obtain the final rail damage assessment result. This method combines deep learning and computer vision to accurately identify and classify various types of damage, unifying the standards for damage assessment and improving its accuracy and efficiency. Furthermore, this method can predict the development trend of rail damage, providing data support for maintenance decisions, thereby improving railway safety and maintenance efficiency.

[0106] Another specific embodiment of the present invention discloses a rail damage assessment device based on a large model, such as... Figure 4 As shown, it includes.

[0107] The initial judgment module 401 is configured to judge the damage of the rail based on the characteristics of the ultrasonic signal and obtain the first judgment result;

[0108] The large model damage assessment module 402 is configured to train a preset large model based on historical rail damage assessment data to obtain a rail damage assessment large model, and input the first assessment result into the rail damage assessment large model to obtain a second assessment result;

[0109] The correction module 403 is configured to correct the damage result of the second judgment result to obtain the final damage result.

[0110] Preferably, the preliminary judgment module 401 is also configured to:

[0111] Define the rectangular defect detection area and move it according to a fixed step size S;

[0112] Each step of movement extracts the image of the damage detection area and obtains the initial damage detection result based on the ultrasonic signal characteristics. The initial damage detection result includes no damage, core damage, bolt hole damage, and rail bottom anomaly. The rectangular damage detection area has a height of H and a width of W.

[0113] Based on the initial damage assessment results, obtain an image of the damaged area;

[0114] The step of obtaining the damaged area image based on the initial damage assessment result includes:

[0115] When the initial damage assessment result is no damage, continue the detection of the next detection area;

[0116] When the initial damage assessment result is nuclear damage, the damaged area image is determined according to the first method;

[0117] When the initial damage assessment result is screw hole damage, the damage area image is determined according to the second method;

[0118] When the initial damage assessment result is an anomaly at the track bottom, the image of the damaged area is determined according to the third method.

[0119] The first, second, and third methods are not entirely the same.

[0120] Preferably, determining the damaged area image according to the first method includes:

[0121] Obtain the height h corresponding to the nuclear damage. i The damaged area O that needs to be inspected is obtained. i :

[0122]

[0123] in dh1 represents the horizontal cut-off length corresponding to the nuclear damage. i =max(0,h) i -Δh1) is the lower bound for the cutoff corresponding to the nuclear damage, uh1 i =min(H,h) i +Δh1) is the upper bound of the cutoff corresponding to the nuclear injury, Δh1 is the longitudinal cutoff length corresponding to the nuclear injury, and w1 and Δh1 are multiples of 10;

[0124] when At that time, add a block with width W and height 2Δh1-uh1 above the intercepted area. i +dh1 i The black areas form the image of the damaged area;

[0125] when At that time, add a block with width W and height 2Δh1-uh1 below the intercepted area. i +dh1 i The black areas form the image of the damaged area;

[0126] Where i is the number of the rectangular defect detection area.

[0127] Preferably, determining the damaged area image according to the second method includes:

[0128] Obtain the height h corresponding to the screw hole damage i The damaged area O that needs to be inspected is obtained. i ;

[0129]

[0130] in dh2 is the transverse section length corresponding to the screw hole damage. i =max(0,h) i -Δh2) is the lower bound for the cutoff corresponding to the screw hole damage, uh2 i =min(H,h) i +Δh2) is the upper limit of the cut corresponding to the screw hole damage, Δh2 is the longitudinal cut length corresponding to the screw hole damage, and w2 and Δh2 are multiples of 10;

[0131] when At that time, add a block with width W and height 2Δh²-uh² above the intercepted area. i +dh2 i The black areas form the image of the damaged area;

[0132] when At that time, add a block with width W and height 2Δh²-uh² below the intercepted area. i +dh2 i The black areas form the image of the damaged area;

[0133] Where i is the number of the rectangular defect detection area.

[0134] Preferably, the method for determining the damaged area image according to the third method includes:

[0135] Obtain the height h corresponding to the track bottom anomaly i The damaged area O that needs to be inspected is obtained. i ;

[0136]

[0137] in dh3 represents the lateral cut-off length corresponding to the track bottom anomaly. i =max(0,h) i -Δh3) is the lower bound for the intercept corresponding to the track bottom anomaly, uh3 i =min(H,h) i +Δh3) is the upper limit of the cut-off corresponding to the rail bottom anomaly, Δh3 is the longitudinal cut-off length corresponding to the rail bottom anomaly, and w3 and Δh3 are multiples of 10;

[0138] when At that time, add a block with width W and height 2Δh3-uh3 above the intercepted area. i +dh3 i The black area represents the image of the damaged area;

[0139] when At that time, add a block with width W and height 2Δh3-uh3 below the intercepted area. i +dh3 i The black area represents the image of the damaged area;

[0140] Where i is the number of the rectangular defect detection area.

[0141] Preferably, the correction module 403 is further configured to: when the second judgment result is clutter, perform clutter detection correction to obtain the final damage judgment result.

[0142] Preferably, the clutter detection and correction includes:

[0143] Damage feature extraction, damage clustering, and damage correction;

[0144] The damage feature extraction includes:

[0145] The damaged area image is converted into a grayscale image. First, the area is divided into L equal parts according to its width, and then the area is divided into L equal parts according to its height, resulting in L*L rectangular areas of the same size, where L is an integer greater than or equal to 10.

[0146] The average pixel value for each rectangular region is calculated to be g. j,k Where j is the index of the rectangular region from left to right horizontally, and k is the index of the rectangular region from top to bottom vertically, the feature set {g} of the damaged region is obtained. j,k}, where j and k are integers from 0 to L-1;

[0147] Construct the characteristic matrix Where ts1 is the set first judgment threshold;

[0148] The damage clustering includes:

[0149] Clustering the feature matrix yields m clusters, and the cluster center of each cluster is extracted and denoted as . dp j,k These are the elements corresponding to the central matrix;

[0150] The feature matrix D i With each cluster center ΔD q Similarity calculation is performed to obtain the classification feature score gd q :

[0151]

[0152]

[0153] Select score gd q The highest-ranking cluster is used to classify the detection region;

[0154] The damage correction includes:

[0155] Obtain the historical verification count gt of the damaged area image. q Number of historical misjudgments q ;

[0156] when At that time, acquire the historical damage records n of the detection area. i With historical clutter record n1 i ,when When the detected area image is determined to be clutter, the detected area image is added to the cluster's tag and stored; when If manual verification is performed and the verification result is noise, then the historical verification count gt q Increase by 1, historical misjudgment count ge q The number of historical verifications remains unchanged; if the verification result indicates damage, then the number of verifications is greater than or equal to gt. q Increase by 1, historical misjudgment count ge q Increase by 1; where the cluster's label includes the damage type; a cluster is a collection of detection regions for a specific damage type;

[0157] when At that time, the image is sent to designated personnel for manual review. If the review result is noise, the historical verification count is gt. q Increase by 1, historical misjudgment count ge q The number of historical verifications remains unchanged; if the verification result indicates damage, then the number of verifications is greater than or equal to gt. q Increase by 1, historical misjudgment count ge q Increase by 1;

[0158] Where ts2 is the set second judgment threshold and ts3 is the set third judgment threshold.

[0159] It should be noted that the preliminary judgment module 401 can implement all the method steps included in S101 in the above method embodiment, and the similarities will not be repeated.

[0160] It should be noted that the large model damage assessment module 402 can implement all the method steps included in S102 in the above method embodiment, and the similarities will not be repeated here.

[0161] It should be noted that the correction module 403 can implement all the method steps included in S103 in the above method embodiment, and the similarities will not be repeated.

[0162] It should be noted that the embodiments of this device and the above-described method belong to the same inventive concept, solve the same technical problem, and achieve the same technical effect. The similarities will not be repeated here.

[0163] Thirdly, the present invention proposes an electronic device, such as... Figure 5 As shown, it includes: memory and one or more processors.

[0164] The memory stores one or more application programs, which are adapted to be executed by the one or more processors to implement the rail damage assessment method based on a large model as described in the first aspect:

[0165] The damage to the rail is determined based on the characteristics of the ultrasonic signal, and a first judgment result is obtained;

[0166] The rail damage judgment model is trained by training a pre-set large model based on historical rail damage data. The first judgment result is then input into the rail damage judgment model to obtain the second judgment result.

[0167] The second judgment result is corrected to obtain the final injury judgment result.

[0168] like Figure 5 As shown, the electronic device includes a processor 501 and a memory 502. The processor 501 and the memory 502 are connected, for example, via a bus interface.

[0169] The structure of this electronic device does not constitute a limitation on the embodiments of the present invention.

[0170] Processor 501 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 501 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0171] A bus interface may include a pathway for transmitting information between the aforementioned components. The bus interface can be a PCI bus or an EISA bus, etc. Bus interfaces can be categorized into address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0172] The memory 502 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0173] It should be noted that, in this embodiment of the device, the one or more application programs are adapted to be executed by the one or more processors to implement all the rail damage judgment methods based on large models described in the first aspect, and the similarities will not be repeated.

[0174] It should be noted that the embodiments of this device and the above-described method belong to the same inventive concept, solve the same technical problem, and achieve the same technical effect. The similarities will not be repeated here.

[0175] Fourthly, the present invention proposes a computer-readable storage medium having a computer program stored thereon, the computer program being loaded and executed by a processor for the rail damage assessment method based on a large model as described in the first aspect.

[0176] The applicant of this invention has provided a detailed description of the embodiments of the invention in conjunction with the accompanying drawings. However, those skilled in the art should understand that the above embodiments are merely preferred embodiments of the invention. The detailed description is only intended to help readers better understand the spirit of the invention and is not intended to limit the scope of protection of the invention. On the contrary, any improvements or modifications made based on the inventive spirit of the invention should fall within the scope of protection of the invention.

[0177] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A large model-based rail damage judgment method, characterized by, The method comprises the following steps: determining the damage of the rail according to the ultrasonic signal characteristics to obtain a first judgment result; training a preset large model according to historical rail damage data to obtain a rail damage judgment large model, inputting the first judgment result into the rail damage judgment large model, and obtaining a second judgment result; correcting the second judgment result to obtain a final damage judgment result; The method for determining the damage of the rail according to the ultrasonic signal characteristics to obtain a first judgment result comprises the following steps: determining a rectangular damage detection area and moving according to a fixed step S; extracting the picture of the damage detection area every step and obtaining an initial damage judgment result according to the ultrasonic signal characteristics, wherein the initial damage judgment result comprises no damage, core damage, screw hole damage, and rail bottom abnormality; wherein the height of the rectangular damage detection area is H and the width is W; obtaining a damage area image according to the initial damage judgment result; The method for obtaining a damage area image according to the initial damage judgment result comprises the following steps: when the initial damage judgment result is no damage, continuing to detect the next detection area; when the initial damage judgment result is core damage, determining a damage area image according to a first method; when the initial damage judgment result is screw hole damage, determining a damage area image according to a second method; when the initial damage judgment result is rail bottom abnormality, determining a damage area image according to a third method; wherein the first method, the second method, and the third method are not completely the same; The method for determining a damage area image according to the first method comprises the following steps: Obtaining the height h corresponding to the core wound i Obtaining the damaged area O that needs to be detected i : in dh1 represents the horizontal cut-off length corresponding to the nuclear damage. i =max(0,h) i -Δh1) is the lower bound for the cutoff corresponding to the nuclear damage, uh1 i =min(H,h) i +Δh1) is the upper bound of the cutoff corresponding to the nuclear injury, Δh1 is the longitudinal cutoff length corresponding to the nuclear injury, and w1 and Δh1 are multiples of 10; When uh1 i -dh1 i <2Δh1 and uh1+dh1 i +dh1 i a black area with width W and height 2Δh1-uh1 When uh1 i -dh1 i <2Δh1 and uh1+dh1 i +dh1 i a black area with width W and height 2Δh1-uh1 wherein i is the number of the rectangular damage detection area; The method for determining a damage area image according to the second method comprises the following steps: Obtaining the height h corresponding to the screw hole damage i , obtaining the damage area O that needs to be detected i ; in dh2 is the transverse section length corresponding to the screw hole damage. i =max(0,h) i -Δh2) is the lower bound for the cutoff corresponding to the screw hole damage, uh2 i =min(H,h) i +Δh2) is the upper limit of the cut corresponding to the screw hole damage, Δh2 is the longitudinal cut length corresponding to the screw hole damage, and w2 and Δh2 are multiples of 10; When uh2 i -dh2 i <2Δh2 and A black area with width W and height 2Δh2-uh2 i +dh2 i is added above the intercept area to form the damage area image. When uh2 i -dh2 i <2Δh2 and uh2 i +dh2 i a black area with width W and height 2Δh2-uh2 wherein i is the number of the rectangular damage detection area; The method for determining a damage area image according to the third method comprises the following steps: Obtaining the height h corresponding to the rail bottom anomaly i Obtaining the damage area O to be detected i ; in dh3 represents the lateral cut-off length corresponding to the track bottom anomaly. i =max(0,h) i -Δh3) is the lower bound for the intercept corresponding to the track bottom anomaly, uh3 i =min(H,h) i +Δh3) is the upper limit of the cut-off corresponding to the rail bottom anomaly, Δh3 is the longitudinal cut-off length corresponding to the rail bottom anomaly, and w3 and Δh3 are multiples of 10; When uh3 i -dh3 i <2Δh3 and At that time, add a block with width W and height 2Δh3-uh3 above the intercepted area. i +dh3 i The black areas form the image of the damaged area; When uh3 i -dh3 i <2Δh3 and A black area with width W and height 2Δh3-uh3 i +dh3 i is added below the intercept area to form the image of the damaged area. wherein i is the number of the rectangular damage detection area.

2. The method according to claim 1, wherein The method for correcting the second judgment result to obtain a final damage judgment result comprises the following steps: The second judgment result comprises no damage, core damage, screw hole damage, rail bottom abnormality, or clutter; when the second judgment result is clutter, performing clutter detection correction to obtain a final damage judgment result.

3. The method according to claim 2, wherein The method for performing clutter detection correction comprises the following steps: damage feature extraction, damage clustering, and damage correction; The method for damage feature extraction comprises the following steps: converting the damage area image into a gray image, dividing the area into L parts according to the width, and dividing the area into L parts according to the height to obtain L*L rectangular areas of the same size, wherein L is an integer greater than or equal to 10; The pixel average value of each rectangular region is calculated as g j,k where j is the sequence number of the rectangular region counted from left to right in the horizontal direction, and k is the sequence number of the rectangular region counted from top to bottom in the vertical direction, to obtain a feature set {g j,k} of the damaged area, where j and k are both integers from 0 to L-1. establishing a feature matrix where ts1 is a set first determination threshold value; The method for damage clustering comprises the following steps: clustering the feature matrix to obtain m clusters, and extracting the clustering center of each cluster as dp j,k is the element corresponding to the center matrix; The feature matrix D is calculated i The similarity is calculated with each cluster center ΔD q The classification feature score gd is obtained by similarity calculation q : Selection score gd q The highest cluster is classified as the detection region. The method for damage correction comprises the following steps: the number of times of historical verification gt of the image of the damaged area q the number of times of historical misjudgment ge q ; When the history damage record n of the detection area is obtained i and the history clutter record n1 i , when the detection area image is determined as clutter, and the detection area image is added to the mark of the cluster and stored; when , if manual review is performed, the history check times gt q is increased by 1, and the history misjudgment number ge q is not changed, if the review result is damage, the history check times gt q is increased by 1, and the history misjudgment number ge q is increased by 1; wherein the mark of the cluster includes the damage type; the cluster is a collection of detection areas of a specific damage type. When the image is sent to designated personnel for manual review, and if the review result is clutter, the historical verification times gt q is increased by 1, and the historical misjudgment number ge q is not changed, and if the review result is damage, the historical verification times gt q is increased by 1, and the historical misjudgment number ge q is increased by 1; wherein ts2 is a second judgment threshold value, and ts3 is a third judgment threshold value.

4. A large model-based rail damage judgment device characterized by comprising: The device is used to implement the large model-based rail damage judgment method of any one of claims 1 to 3, and the device comprises: a preliminary judgment module configured to determine the damage of the rail according to the ultrasonic signal characteristics to obtain a first judgment result; a large model damage judgment module configured to train a preset large model according to historical rail damage data to obtain a rail damage judgment large model, input the first judgment result into the rail damage judgment large model, and obtain a second judgment result; A correction module is configured to correct the second judgment result to obtain a final damage judgment result. The first judgment result is obtained by judging the damage of the steel rail according to the ultrasonic signal characteristics, and the method comprises the following steps: A rectangular damage judgment detection area is determined, and the area is moved by a fixed step S; An initial damage judgment result is obtained by judging the damage according to the ultrasonic signal characteristics, and the initial damage judgment result comprises no damage, core damage, screw hole damage and rail bottom abnormality, wherein the height of the rectangular damage judgment detection area is H and the width is W; An image of the damage area is obtained according to the initial damage judgment result; The image of the damage area is obtained according to the initial damage judgment result, and the method comprises the following steps: When the initial damage judgment result is no damage, the detection of the next detection area is continued; When the initial damage judgment result is core damage, the image of the damage area is determined according to a first mode; When the initial damage judgment result is screw hole damage, the image of the damage area is determined according to a second mode; When the initial damage judgment result is rail bottom abnormality, the image of the damage area is determined according to a third mode; The first mode, the second mode and the third mode are not completely the same; The image of the damage area is determined according to the first mode, and the method comprises the following steps: Obtaining the height h corresponding to the core wound i , obtaining the damaged area O that needs to be detected i : wherein is a lateral intercept length corresponding to the core damage, dh1 i = max(0, h i - Δh1) is a lower intercept limit corresponding to the core damage, uh1 i = min(H, h i + Δh1) is an upper intercept limit corresponding to the core damage, Δh1 is a longitudinal intercept length corresponding to the core damage, w1, Δh1 are multiples of 10; When uh1 i -dh1 i <2Δh1 and At that time, add a block with width W and height 2Δh1-uh1 above the intercepted area. i +dh1 i The black areas form the image of the damaged area; When uh1 i -dh1 i <2Δh1 and uh1+dh1 i +dh1 i a black area with width W and height 2Δh1-uh1 Wherein, i is the number of the rectangular damage judgment detection area; The image of the damage area is determined according to the second mode, and the method comprises the following steps: Obtaining the height h corresponding to the screw hole damage i , obtaining the damage area O that needs to be detected i ; wherein is a transverse length of the hole damage, dh2 i = max(0, h i - Δh2) is a lower boundary of the hole damage, uh2 i = min(H, h i + Δh2) is an upper boundary of the hole damage, and Δh2 is a longitudinal length of the hole damage, w2 and Δh2 are multiples of 10. When uh2 i -dh2 i <2Δh2 and A black area with width W and height 2Δh2-uh2 i +dh2 i is added above the intercept area to form the damage area image. When uh2 i -dh2 i <2Δh2 and uh2 i +dh2 i a black area with width W and height 2Δh2-uh2 Wherein, i is the number of the rectangular damage judgment detection area; The image of the damage area is determined according to the third mode, and the method comprises the following steps: Obtaining the height h corresponding to the rail bottom anomaly i Obtaining the damage area O to be detected i ; in dh3 represents the lateral cut-off length corresponding to the track bottom anomaly. i =max(0,h) i -Δh3) is the lower bound for the intercept corresponding to the track bottom anomaly, uh3 i =min(H,h) i +Δh3) is the upper limit of the cut-off corresponding to the rail bottom anomaly, Δh3 is the longitudinal cut-off length corresponding to the rail bottom anomaly, and w3 and Δh3 are multiples of 10; When uh3 i -dh3 i <2Δh3 and At that time, add a block with width W and height 2Δh3-uh3 above the intercepted area. i +dh3 i The black areas form the image of the damaged area; When uh3 i -dh3 i <2Δh3 and uh3+dh3 i +dh3 i a black area with width W and height 2Δh3-uh3+dh3 Wherein, i is the number of the rectangular damage judgment detection area.

5. A large model-based rail damage judgment device characterized by comprising: The device comprises a memory, a processor and a user interface; The memory is used to store a computer program; The user interface is used to realize interaction with the user; The processor is used to read the computer program in the memory, and the processor executes the computer program to realize the method for judging the damage of the steel rail based on the large model according to any one of claims 1 to 3.

6. A processor-readable storage medium, characterized in that, The processor readable storage medium stores a computer program, and the processor executes the computer program to realize the method for judging the damage of the steel rail based on the large model according to any one of claims 1 to 3.

Images