Illegal excavation identification model training method and device, equipment and storage medium

Abstract

The invention provides an illegal excavation recognition model training method and device, equipment and a storage medium, and the method comprises the steps: obtaining excavation data of each project stage in a legal scene and an illegal scene as initial data, carrying out the preprocessing of the initial data, and obtaining the preprocessed data, the attribute types of the excavation data comprise basic characteristics and vibration signals; signal features are obtained through calculation based on the preprocessed vibration signals; and taking the preprocessed basic features and signal features as a training set, constructing a corresponding initial random forest for each engineering stage, correcting the initial random forest by using a test set to obtain a final random forest corresponding to each engineering stage, and integrating each final random forest into an illegal excavation identification model. According to the method, through supervised training, an illegal excavation identification model containing multiple groups of random forests is constructed, and the manufacturing cost of a hardware equipment layout method is effectively reduced by using a machine learning algorithm.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a training method, device, equipment and storage medium for an illegal excavation identification model. Background Art

[0002] With the development of urbanization, the density of subways in urban underground is getting higher and higher. By the end of 2020, the length of subways in operation in my country has reached 6,302.79 kilometers. During the construction and maintenance of subways, accidents of illegal penetration of subway lines often occur, which brings heavy pressure to daily inspection work.

[0003] Among the current relevant technologies, traditional manual patrols are often used, which leads to low efficiency in identifying illegal excavations, increases the economic cost and time of maintenance, and may even cause safety accidents of varying degrees; some intelligent detection methods applied to subway signals often need to be based on hardware deployment, and optical fibers and other devices need to be pre-installed in the tunnel, which has a high cost.

[0004] Based on the above analysis of the development status of this technology field, the intelligent methods in the existing technologies rely on hardware deployment, and lack machine learning detection solutions that do not require the deployment of hardware devices. Summary of the invention

[0005] The purpose of the present invention is to provide a training method, device, equipment and storage medium for an illegal excavation identification model, aiming to solve the above-mentioned problems in the prior art.

[0006] According to a first aspect of an embodiment of the present disclosure, a method for training an illegal excavation identification model is provided, comprising:

[0007] Acquire excavation data of each engineering stage under legal and illegal scenarios as initial data, preprocess the initial data to obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals;

[0008] Calculate the signal characteristics based on the preprocessed vibration signal;

[0009] The preprocessed basic features and signal features are used as training sets to construct the corresponding initial random forest for each engineering stage. The test set is used to correct the initial random forest to obtain the final random forest corresponding to each engineering stage. The final random forests are integrated into an illegal excavation recognition model.

[0010] According to a second aspect of an embodiment of the present disclosure, a training device for an illegal excavation identification model is provided, comprising:

[0011] A data preprocessing module is used to obtain excavation data of each engineering stage in legal and illegal scenarios as initial data, preprocess the initial data, and obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals;

[0012] A signal feature acquisition module, used for calculating the signal feature based on the preprocessed vibration signal;

[0013] The model building module is used to use the preprocessed basic features and signal features as training sets, build the corresponding initial random forest for each engineering stage, use the test set to correct the initial random forest, obtain the final random forest corresponding to each engineering stage, and integrate the final random forests into an illegal excavation recognition model.

[0014] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the training method for the illegal excavation identification model provided in the first aspect of the present disclosure are implemented.

[0015] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps of the training method for the illegal excavation identification model provided in the first aspect of the present disclosure are implemented.

[0016] The technical solution provided by the embodiment of the present invention includes the following beneficial effects: through supervised training, an illegal excavation identification model including multiple groups of random forests is constructed. The decision trees in the same group of random forests are for the same engineering stage. Therefore, no matter which engineering stage of the excavation construction activity to be detected can be obtained, the corresponding random forest can be used to accurately judge the construction activity. At the same time, the use of machine learning algorithms can effectively reduce the cost of the hardware equipment deployment method.

[0017] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 is a flow chart of a training method for an illegal excavation identification model according to an embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of a random forest group according to an embodiment of the present invention;

[0021] Figure 3 is a schematic diagram of a training device for an illegal excavation identification model according to an embodiment of the present invention;

[0022] Figure 4 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0024] Method Embodiment

[0025] According to an embodiment of the present invention, a method for training an illegal excavation recognition model is provided. Figure 1 is a flow chart of a training method for an illegal excavation identification model according to an embodiment of the present invention. Figure 1 As shown, according to an embodiment of the present invention Figure 1 The training method of the illegal excavation identification model of the embodiment of the present invention specifically includes:

[0026] In step S110, the excavation data of each engineering stage in the legal scenario and the illegal scenario are obtained as initial data, and the initial data is preprocessed to obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals, specifically including:

[0027] The excavation data of the drilling stage, excavation stage and blasting load stage in legal and illegal scenarios are obtained as initial data; the drilling stage data is the information collected during the drilling process, and the excavation stage data is a series of information collected during the excavation process of the excavator; the blasting load stage refers to a series of quantitative information related to blasting energy, efficiency, safety and effect during blasting;

[0028] Each engineering stage includes related attributes of basic features and related attributes of vibration signals. In the embodiment of the present invention, there is no requirement for the number of related attributes of basic features in the initial data set. The basic features of the drilling stage include drilling speed, the basic features of the excavation stage include excavation volume, and the basic features of the blasting load stage include the maximum or average size of the rock after crushing. That is to say, the basic features include a series of features, and the vibration signal is an independent feature.

[0029] The vibration signal is a vibration signal collected by a vibration sensor in a time or space sequence.

[0030] The data preprocessing process is:

[0031] Delete the data with missing basic features and anomalies in the initial data, that is, remove the entire data with missing values ​​to obtain the cleaned basic data. Since each piece of basic feature data is independent of each other, removing the entire data will not affect the training;

[0032] The cleaned basic data is standardized or normalized to obtain standard basic data. In the embodiment of the present invention, the minimum-maximum standardization method is used to scale the data to between 0 and 1, and labels of legal scenes or illegal scenes are added to the standard basic data to obtain the preprocessed basic features; data scaling is shown in Formula 4:

[0033]

[0034] Among them, x min Represents the minimum value in the basic data after cleaning, x max represents the maximum value of the basic data after cleaning, x represents the data to be standardized, that is, the basic data after cleaning, x norm Represents standard basic data.

[0035] A filter is used to remove noise from the vibration signal in the initial data to obtain a standard vibration signal, and a label of a legal scene or an illegal scene is added to the standard vibration signal to obtain a preprocessed vibration signal;

[0036] In step S120, the signal features are calculated based on the preprocessed vibration signal, specifically including:

[0037] According to formula 1, the preprocessed vibration signal is processed into a frequency domain signal using the Fourier transform method. According to formula 2, the maximum value in the frequency domain signal is obtained as the main frequency feature. According to formula 3, the amplitude feature is calculated based on the main frequency feature:

[0038]

[0039] f peak = arg max k|X[k]| Formula 2;

[0040] a peak =|X[f peak ]| Formula 3;

[0041] Where X[k] represents the discrete Fourier transform result of the vibration signal at discrete frequency k. After the Fourier transform, the main frequency characteristics of the vibration signal are determined by finding the maximum value of X[k]. x[n] represents the nth sampling point in the vibration signal time series, N represents the total number of sampling points, j represents the imaginary unit, and arg max k represents the maximum value, |.| represents the absolute value, and f peak represents the main frequency characteristics, X[f peak ] represents the discrete Fourier transform result at the main frequency feature, A peak Represents the amplitude feature, which is the intensity at the main frequency. Taking the main frequency feature and the amplitude feature as signal features is equivalent to the signal feature having two attributes: the main frequency feature and the amplitude feature.

[0042] e represents the base of natural logarithm, and π represents pi;

[0043] It should be noted that, since the vibration signal is discretely collected in a time series in practical applications, a discrete Fourier transform is used in the embodiment of the present invention. Preferably, a signal reconstruction technique can be used to process the discrete vibration signal into a time-continuous signal, and a frequency domain signal is obtained by continuous Fourier transform processing as a comparison signal, and then the value of the discrete collection corresponding time point in the comparison signal is obtained as a specific comparison point;

[0044] If the sampling rate is high enough and the spectral characteristics of the signal meet the sampling requirements, the frequency domain signal obtained by the discrete Fourier transform can be very close to the continuous Fourier transform. Therefore, preferably, when the frequency domain signal obtained by solving Formula 1 differs too much from the comparison point, the preprocessed vibration signal can be further optimized, for example: by estimating the noise spectrum and subtracting the noise spectrum from the signal spectrum to remove the noise.

[0045] The main frequency feature and amplitude feature are two important features in signal processing. The main frequency feature can be used to detect the frequency coding of the signal, and the amplitude feature can be used to judge the quality of the signal.

[0046] In step S130, the preprocessed basic features and signal features are used as training sets to construct the corresponding initial random forest for each engineering stage, and the test set is used to correct the initial random forest to obtain the final random forest corresponding to each engineering stage. The final random forests are integrated into an illegal excavation recognition model, which specifically includes:

[0047] In the process of constructing each decision tree in the initial random forest, a preset number of data are extracted from the training set as the construction set using the method of extraction with replacement;

[0048] Randomly select attributes from the construction set as construction attributes, where the construction attributes include at least one attribute of the basic feature and one attribute of the signal feature. When selecting the number of attributes, the threshold can be limited to prevent the number of decision tree layers from being too high. Although the specific selection of construction attributes is random, it cannot only include the data of the basic features or the data of the signal features. It is necessary to select at least one feature from the basic features and at least one feature from the signal features to ensure that the model fully utilizes the information of various features. Considering only one type of feature may lead to a decrease in the generalization ability and prediction accuracy of the model.

[0049] The Gini coefficient is used to select the order of constructing attributes to build a specific decision tree; the lower the Gini coefficient, the more discriminating the attribute is in the classification task;

[0050] The threshold division method of nodes in the decision tree adopts the conventional method. For example, for numerical attributes, the average value of the constructed attribute is taken as the node threshold. The decision tree can directly output the category label, such as legal or illegal;

[0051] Use the test set to verify each decision tree in the initial random forest, and use the pruning method to correct the decision trees whose verification accuracy is lower than the result threshold, until the accuracy of all decision trees is higher than or equal to the result threshold, and obtain the final random forest corresponding to each engineering stage;

[0052] Preferably, since the training data set of the present application is large in scale, in the embodiment of the present invention, in addition to adding pruning in the verification process, pre-pruning can be added in the process of initial random forest construction;

[0053] Due to the particularity of the hazards of illegal excavation activities, it is necessary to identify illegal excavation behaviors in real time. If all attribute types are used together to construct decision trees without distinguishing the engineering stages, a unified set of random forests will be formed. This will result in the inability to obtain accurate identification results when attribute types are missing during verification. For example, if the construction team only carried out illegal drilling, and both excavation and blasting data are missing, the random forest corresponding to the existing drilling can be used for judgment and identification. Figure 2 is a schematic diagram of a random forest group according to an embodiment of the present invention. Figure 2 As shown in the figure, the corresponding random forest is set in each engineering stage.

[0054] The method further comprises:

[0055] In step S140, the excavation data of the construction activity to be identified is obtained as a validation set, and each piece of data in the validation set is input into the corresponding final random forest to identify whether each piece of data is illegal excavation data. When the data exceeding the proportion threshold in the validation set is identified as illegal excavation data, the construction activity to be identified is determined to be an illegal excavation activity, specifically including:

[0056] The excavation data generated by the construction activity to be identified is used as the validation set. If only drilling is performed at the current progress, only the drilling data needs to be input into the corresponding final random forest for validation. Before validation, the same data preprocessing method as the training set needs to be used to process the initial validation set, and the same signal feature acquisition method needs to be adopted;

[0057] Each decision tree can output a legal or illegal identification result. In the final random forest, the probability of each piece of data being illegal can be obtained, that is, the proportion of decision trees that predict illegal results to the total decision trees in the final random forest. If the probability exceeds 50%, the data is identified as illegal excavation data.

[0058] In the embodiment of the present invention, the ratio threshold is also 50%, that is, if more than half of the data in the verification set are identified as illegal excavation, the construction activity to be identified is identified as illegal excavation. In reality, each engineering stage should include multiple decision trees, and there is a large amount of data in the verification set. However, for the sake of convenience, the embodiment of the present invention takes three engineering stages, each including 3 decision trees, and 4 data in the verification set as an example. Among these 4 data, 2 are drilling stage data and 2 are excavation stage data, numbered a, b, c, and d respectively;

[0059] The prediction results of a in the final random forest corresponding to drilling are illegal, illegal, and legal, respectively, so the probability of illegality is 66%, and data a is illegal excavation data;

[0060] If the prediction results of b in the final random forest corresponding to drilling are illegal, illegal, and illegal, the illegal probability is 100%, and data b is illegal excavation data;

[0061] The prediction results of c in the final random forest corresponding to mining are illegal, legal, and legal respectively, then the illegal probability is 33%, and data c is legal mining data;

[0062] The prediction results of d in the final random forest corresponding to the mining are illegal, illegal, and legal, respectively, then the illegal probability is 66%, and data d is illegal mining data;

[0063] In the verification set of the embodiment of the present invention, 75% of the data was judged to be illegal excavation data, which is greater than the proportion threshold of 50%. Therefore, the construction activity to be identified is determined to be an illegal excavation activity. The embodiment of the present invention does not need to be verified in the final random forest corresponding to the blasting load stage because there is no data from the blasting engineering stage in the verification set.

[0064] To sum up, in response to the existing problems, this paper invents a training method for an illegal excavation identification model, in which basic feature attributes and signal feature attributes that can represent excavation data are selected as the basis for constructing a decision tree. The above features are crucial for distinguishing legal from illegal excavations, and can effectively improve the training effect of the model; through supervised training, an illegal excavation identification model containing multiple groups of random forests is constructed. The decision trees in the same group of random forests are for the same engineering stage. Therefore, no matter which engineering stage of the excavation construction activity to be detected can be obtained, the corresponding random forest can be used to accurately judge the construction activity; the decision tree in the random forest is only constructed based on the corresponding features, which to a certain extent avoids too many layers of decision trees, and can fully obtain effective information in the attributes; the scheme generally uses a machine learning algorithm, which can effectively reduce the cost of the hardware equipment deployment method, and there is no need to pre-install hardware equipment such as optical fiber in the tunnel, reducing the intrusion into the tunnel structure.

[0065] Device Embodiment

[0066] According to an embodiment of the present invention, a training device for an illegal excavation recognition model is provided. Figure 3 is a schematic diagram of a training device for an illegal excavation identification model according to an embodiment of the present invention. Figure 3 As shown, according to an embodiment of the present invention Figure 3 The training device of the illegal excavation identification model of the embodiment of the present invention specifically includes:

[0067] The data preprocessing module 30 is used to obtain excavation data of each engineering stage in legal scenarios and illegal scenarios as initial data, preprocess the initial data to obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals, and are specifically used for:

[0068] The excavation data of the drilling stage, excavation stage and blasting load stage under legal and illegal scenarios are obtained as initial data.

[0069] Delete the data with missing basic features and abnormalities in the initial data to obtain the cleaned basic data, standardize or normalize the cleaned basic data to obtain standard basic data, add labels of legal scenarios or illegal scenarios to the standard basic data, and obtain the preprocessed basic features;

[0070] A filter is used to remove noise from the vibration signal in the initial data to obtain a standard vibration signal, and a label of a legal scene or an illegal scene is added to the standard vibration signal to obtain a preprocessed vibration signal.

[0071] The signal feature acquisition module 32 is used to calculate the signal feature based on the preprocessed vibration signal, specifically for:

[0072] According to formula 1, the preprocessed vibration signal is processed into a frequency domain signal using the Fourier transform method. According to formula 2, the maximum value in the frequency domain signal is obtained as the main frequency feature. According to formula 3, the amplitude feature is calculated based on the main frequency feature:

[0073]

[0074] f ppeak = arg max k |X[k]| Formula 2;

[0075] A pea k=|X[f peak ] Formula 3;

[0076] Where X[k] represents the discrete Fourier transform result of the vibration signal at discrete frequency k, x[n] represents the nth sampling point in the vibration signal time series, N represents the total number of sampling points, j represents the imaginary unit, and argmax k represents the maximum value, |.| represents the absolute value, and f peak represents the main frequency characteristics, X[f peak ] represents the discrete Fourier transform result at the main frequency feature, A peak Represents amplitude characteristics;

[0077] The main frequency characteristics and amplitude characteristics are used as signal features.

[0078] The model building module 34 is used to use the pre-processed basic features and signal features as training sets, build the corresponding initial random forest for each engineering stage, use the test set to correct the initial random forest, obtain the final random forest corresponding to each engineering stage, and integrate each final random forest into an illegal excavation recognition model, which is specifically used to:

[0079] In the process of constructing each decision tree in the initial random forest, a preset number of data are extracted from the training set as a construction set by using a replacement extraction method, and attributes are randomly selected from the construction set as construction attributes, wherein the construction attributes include at least one attribute of a basic feature and one attribute of a signal feature, and the construction order of the construction attributes is selected by using the Gini coefficient method;

[0080] Use the test set to verify each decision tree in the initial random forest, and use the pruning method to correct the decision trees whose verification accuracy is lower than the result threshold, until the verification accuracy results of all decision trees are higher than or equal to the result threshold, and obtain the final random forest corresponding to each engineering stage.

[0081] The device further comprises:

[0082] The model uses module 36 to obtain the excavation data of the construction activity to be identified as a verification set, input each data in the verification set into the corresponding final random forest, and identify whether each data is illegal excavation data. When the data in the verification set that exceeds the proportion threshold is identified as illegal excavation data, the construction activity to be identified is determined to be an illegal excavation activity.

[0083] To sum up, in response to the existing problems, this paper invents a training device for an illegal excavation identification model, which selects basic feature attributes and signal feature attributes that can represent excavation data as the basis for constructing a decision tree. The above features are crucial for distinguishing legal and illegal excavations, and can effectively improve the training effect of the model; through supervised training, an illegal excavation identification model containing multiple groups of random forests is constructed. The decision trees in the same group of random forests are for the same engineering stage. Therefore, no matter which engineering stage of the excavation construction activity to be detected can be obtained, the corresponding random forest can be used to accurately judge the construction activity; the decision tree in the random forest is only constructed based on the corresponding features, which to a certain extent avoids too many layers of decision trees, and can fully obtain effective information in the attributes; the scheme generally uses a machine learning algorithm, which can effectively reduce the cost of the hardware equipment deployment method, and there is no need to pre-install hardware equipment such as optical fiber in the tunnel, reducing the intrusion into the tunnel structure.

[0084] Electronic device embodiment

[0085] Figure 4 4 is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device 400 may include at least one processor 410 and a memory 420. The processor 410 may execute instructions stored in the memory 420. The processor 410 is connected to the memory 420 through a data bus. In addition to the memory 420, the processor 410 may also be connected to an input device 430, an output device 440, and a communication device 450 through a data bus.

[0086] The processor 410 may be any conventional processor, such as a commercially available CPU. The processor may also include a graphics processor (Graphic Process Unit, GPU), a field programmable gate array (Field Programmable Gate Array, FPGA), a system on chip (System on Chip, SOC), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC) or a combination thereof.

[0087] Memory 420 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0088] In the embodiment of the present disclosure, executable instructions are stored in the memory 420, and the processor 410 can read the executable instructions from the memory 420 and execute the instructions to implement all or part of the steps of the training method of the illegal excavation identification model in any of the above exemplary embodiments.

[0089] Computer Readable Storage Medium Embodiments

[0090] In addition to the above-mentioned methods and devices, the exemplary embodiments of the present disclosure may also be a computer program product or a computer-readable storage medium storing the computer program product. The computer product includes computer program instructions, which can be executed by a processor to implement all or part of the steps described in the training method of the illegal excavation identification model in any of the above-mentioned exemplary embodiments.

[0091] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages ​​such as Java, C++, etc., and also conventional procedural programming languages ​​such as "C" language or similar programming languages ​​and scripting languages ​​(e.g., Python). The program code may be executed entirely on the user computing device, partially on the user computing device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0092] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of readable storage media include: a static random access memory (SRAM) with one or more wires electrically connected, an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk, or any suitable combination of the above.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A training method for an illegal excavation identification model, characterized in that: include: Acquire excavation data of each engineering stage in legal scenarios and illegal scenarios as initial data, preprocess the initial data to obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals; Calculate the signal characteristics based on the preprocessed vibration signal; The preprocessed basic features and the signal features are used as training sets to construct a corresponding initial random forest for each engineering stage. The initial random forest is corrected using the test set to obtain the final random forest corresponding to each engineering stage. The final random forests are integrated into an illegal excavation recognition model.

2. The method according to claim 1, characterized in that The method further comprises: The excavation data of the construction activity to be identified is obtained as a verification set, and each data in the verification set is input into the corresponding final random forest to identify whether each data is illegal excavation data. When the data exceeding the proportion threshold in the verification set is identified as illegal excavation data, the construction activity to be identified is determined to be an illegal excavation activity.

3. The method according to claim 1, characterized in that The obtaining of excavation data of each engineering stage in legal scenarios and illegal scenarios as initial data specifically includes: obtaining excavation data of the drilling stage, the excavation stage and the blasting load stage in legal scenarios and illegal scenarios as initial data.

4. The method according to claim 1, characterized in that: The preprocessing of the initial data specifically includes: Deleting data in which basic features are missing or abnormal in the initial data to obtain cleaned basic data, standardizing or normalizing the cleaned basic data to obtain standard basic data, and adding labels of legal scenarios or illegal scenarios to the standard basic data to obtain preprocessed basic features; A filter is used to remove noise from the vibration signal in the initial data to obtain a standard vibration signal, and a label of a legal scene or an illegal scene is added to the standard vibration signal to obtain a preprocessed vibration signal.

5. The method according to claim 1, characterized in that The signal features calculated based on the preprocessed vibration signal specifically include: According to formula 1, the preprocessed vibration signal is processed into a frequency domain signal using the Fourier transform method. According to formula 2, the maximum value in the frequency domain signal is obtained as the main frequency feature. According to formula 3, the amplitude feature is calculated based on the main frequency feature: f peak = arg max k |X[k]| Formula 2; A peak =|X[f peak ]| Formula 3; Where X[k] represents the discrete Fourier transform result of the vibration signal at discrete frequency k, x[n] represents the nth sampling point in the vibration signal time series, N represents the total number of sampling points, j represents the imaginary unit, and argmax k represents the maximum value, |.| represents the absolute value, and f peak represents the main frequency characteristics, X[f peak ] represents the discrete Fourier transform result at the main frequency feature, A peak Represents amplitude characteristics; The main frequency feature and the amplitude feature are used as signal features.

6. The method according to claim 1, characterized in that The construction of the corresponding initial random forest for each engineering stage, and the use of the test set to correct the initial random forest to obtain the final random forest corresponding to each engineering stage specifically include: In each decision tree construction process in the initial random forest, a preset number of data are extracted from the training set as a construction set by using a replacement extraction method, and attributes are randomly selected from the construction set as construction attributes, wherein the construction attributes include at least an attribute of a basic feature and an attribute of a signal feature, and the construction order of the construction attributes is selected by using a Gini coefficient method; Use the test set to verify each decision tree in the initial random forest, and use the pruning method to correct the decision trees whose verification accuracy is lower than the result threshold, until the verification accuracy results of all decision trees are higher than or equal to the result threshold, and obtain the final random forest corresponding to each engineering stage.

7. A training device for an illegal excavation identification model, characterized in that: include: A data preprocessing module, used to obtain excavation data of each engineering stage in legal scenarios and illegal scenarios as initial data, preprocess the initial data to obtain preprocessed data, wherein the attribute types of the excavation data include basic features and vibration signals; A signal feature acquisition module, used for calculating the signal feature based on the preprocessed vibration signal; The model building module is used to use the preprocessed basic features and the signal features as training sets, build a corresponding initial random forest for each engineering stage, use the test set to correct the initial random forest, obtain the final random forest corresponding to each engineering stage, and integrate the final random forests into an illegal excavation identification model.

8. The device according to claim 7, characterized in that The device further comprises: The model uses a module to obtain excavation data of the construction activity to be identified as a verification set, input each data in the verification set into the corresponding final random forest, and identify whether each data is illegal excavation data. When the data exceeding the proportion threshold in the verification set is identified as illegal excavation data, the construction activity to be identified is determined to be an illegal excavation activity.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the training method for the illegal excavation identification model as described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by the processor, the steps of the training method for the illegal excavation identification model as described in any one of claims 1 to 6 are implemented.

Images