Mine ecological environment assessment method and device based on remote sensing detection data analysis

Abstract

The present invention provides a method and device for evaluating the ecological environment of a mine based on remote sensing detection data analysis, which relates to the field of data processing technology. The model for extracting features from the spectral vector data of a mine area can fully consider the dynamic changes or local complexity of the data, and improve the model's ability to identify key features from the spectral data. The angle parameters corresponding to the key features in the preset data distribution dimensions are also calculated and mapped to the quantum state, and classification decisions are made based on the quantum state. The parallelism and efficiency of quantum computing are used to directly participate in the classification process, and the characteristics of different data sets can be adapted to improve the accuracy and flexibility of classification. In summary, the embodiments of the present invention can adapt to the different characteristics of the data, effectively improve the data extraction capability, and improve the accuracy of model identification, providing a novel and potentially efficient approach for the evaluation of the ecological environment of the mine.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a mine ecological environment assessment method and device based on remote sensing detection data analysis. Background Art

[0002] With the increase of mining activities, continuous monitoring of the mine environment has become particularly important to ensure that the goals of environmental protection and sustainable development are achieved. Remote sensing technology has become an important means of studying and monitoring the earth's surface and environmental changes because it can provide large-scale and efficient surface monitoring.

[0003] Remote sensing uses sensors carried by satellites or aircraft to collect spectral data of the surface. These data can reflect the spectral characteristics of different surface materials such as vegetation, water bodies, rocks, etc. By analyzing these spectral vector data, different surface types in the mining area can be identified and classified, and the impact of mining activities on the ecological environment can be evaluated. However, traditional data processing methods often face the problem of insufficient feature extraction and inability to adapt to the different characteristics of the data, which limits the generalization ability and accuracy of the model. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a mine ecological environment assessment method and device based on remote sensing detection data analysis, which can effectively improve data extraction capabilities and improve the accuracy of model recognition.

[0005] In a first aspect, an embodiment of the present invention provides a method for evaluating a mine ecological environment based on remote sensing detection data analysis, wherein the method includes: obtaining pre-collected spectral vector data of a mine area; inputting the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model; wherein the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized through the fire ant algorithm; calculating the angle parameters corresponding to the key features in the preset data distribution dimensions, and mapping the key features to quantum states based on the angle parameters; based on the quantum states, making classification decisions on the key features to determine the mine ecological category indicated by the spectral vector data of the mine area; and evaluating the ecological environment of the mine area based on the mine ecological category.

[0006] In combination with the first aspect, an embodiment of the present invention also provides a first implementation method of the first aspect, wherein, based on the quantum state, a classification decision is made on the key features to determine the mining ecological category indicated by the spectral vector data, including: performing quantum logic gate operations on the quantum state, measuring the quantum state, and determining the category label indicated by the quantum state; based on the category label, determining the mining ecological category indicated by the spectral vector data of the mining area.

[0007] In combination with the first aspect, an embodiment of the present invention also provides a second implementation of the first aspect, wherein the step of evaluating the ecological environment of a mining area based on the mining ecological category includes: calculating the vegetation index corresponding to the spectral vector data; and performing a mining ecological environment evaluation on the mining area according to the vegetation index.

[0008] In combination with the first aspect, an embodiment of the present invention also provides a third implementation method of the first aspect, wherein a method for constructing a feature extraction model includes: training a preset feature extraction neural network through a preset training sample set, and evaluating the recognition differences of the feature extraction neural network; adjusting the learning rate of the feature extraction neural network according to the recognition differences using an adaptive feedback mechanism; searching for the optimal solution for the parameters of the feature extraction neural network using a fire ant algorithm, combining the gradient descent method to calculate the error gradient of the feature extraction neural network, and updating the network parameters of the feature extraction neural network; until the feature extraction neural network meets the preset training requirements, a feature extraction model is constructed based on the feature extraction neural network.

[0009] In combination with the first aspect, an embodiment of the present invention also provides a fourth implementation of the first aspect, wherein the method also includes: dynamically adjusting the distance between the vectors of the training sample set, and monitoring the current recognition difference of the feature extraction neural network to determine the learning progress of the feature extraction neural network.

[0010] In combination with the first aspect, an embodiment of the present invention also provides a fifth implementation of the first aspect, wherein the steps of calculating the angle parameters corresponding to the key feature in multiple data distribution dimensions, and mapping the key feature to the quantum state based on the angle parameters, include: calculating the first angle parameter corresponding to the key feature based on the standard deviation of the key feature; calculating the second angle parameter corresponding to the key feature based on the mean of the key feature; calculating the third angle parameter corresponding to the key feature based on the skewness of the key feature; and mapping the key feature to the quantum state through a parameterized single-qubit rotation gate based on the first angle parameter, the second angle parameter, and the third angle parameter.

[0011] In combination with the first aspect, an embodiment of the present invention also provides a sixth implementation of the first aspect, wherein a classification decision is made on the key features through a support vector machine, and the method also includes: training a preset support vector machine using a preset training sample set, and, during the training process of the support vector machine, measuring the state of the quantum register corresponding to the quantum state of the training sample set; optimizing the parameters of the support vector machine according to the state; wherein the parameters are optimized based on a loss function, and dynamically adjusted according to the importance evaluation results of the key features.

[0012] In combination with the first aspect, an embodiment of the present invention also provides a seventh implementation method of the first aspect, wherein the method also includes: obtaining pre-collected initial samples of the mining area, extracting sequence characteristics and frequency characteristics of the initial samples of the mining area, and generating a composite feature vector; inputting the initial samples of the mining area and the composite feature vector into a preset generative adversarial network to generate preliminary expanded samples; evaluating the quality of the preliminary expanded samples; when the quality meets the preset quality requirements, merging the preliminary expanded samples with the initial samples of the mining area to construct a training sample set.

[0013] In combination with the first aspect, an embodiment of the present invention also provides an eighth implementation of the first aspect, wherein the method also includes: calculating the regularization coefficient of the generator of the generative adversarial network based on the preliminary expanded samples; based on the regularization coefficient and the output of the discriminator of the generative adversarial network, adjusting the parameters of the generator and the parameters of the discriminator of the generative adversarial network so that the quality of the preliminary expanded samples meets the preset quality requirements.

[0014] In the second aspect, an embodiment of the present invention also provides a mine ecological environment assessment device based on remote sensing detection data analysis, wherein the device includes: a data acquisition module, used to acquire pre-collected spectral vector data of a mine area; a feature extraction module, used to input the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model; wherein the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized through the fire ant algorithm; a data processing module, used to calculate the angle parameters corresponding to the key features in the preset data distribution dimension, and based on the angle parameters, map the key features to quantum states; a classification module, used to make classification decisions on the key features based on the quantum state, and determine the mine ecological category indicated by the spectral vector data of the mine area; an assessment module, used to assess the ecological environment of the mine area based on the mine ecological category.

[0015] The embodiments of the present invention bring the following beneficial effects: the method and device for evaluating the ecological environment of a mine based on remote sensing detection data analysis provided by the present invention extracts features from the spectral vector data of the mine area through a feature extraction model, and the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized through the fire ant algorithm, so that the network can automatically adjust its parameters when facing the local complexity of the spectral vector data, optimize the learning efficiency, and improve the model's ability to identify key features from the spectral data. In addition, the parameters of the feature extraction model are updated based on the learning progress of the training sample set, which can fully consider the dynamic changes or local complexity of the data and improve the recognition accuracy. Further, the embodiment of the present invention also uses the superposition and entanglement characteristics of quantum bits to process information, converts data into quantum states by calculating angle parameters, covers the probability amplitudes of different quantum bits, and directly participates in the classification process using the parallelism and efficiency of quantum computing, which helps to adapt to the characteristics of different data sets, thereby improving the accuracy and flexibility of classification. In summary, the embodiments of the present invention can effectively improve the data extraction capability, improve the accuracy of model recognition, and provide a novel and potentially efficient approach for the evaluation of the ecological environment of mines.

[0016] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A flowchart of a mine ecological environment assessment method based on remote sensing detection data analysis provided by an embodiment of the present invention;

[0020] Figure 2 A flowchart of another method for evaluating the ecological environment of a mine based on remote sensing data analysis provided by an embodiment of the present invention;

[0021] Figure 3 A flowchart of a method for constructing a training sample set provided by an embodiment of the present invention;

[0022] Figure 4 A flowchart of a mine ecological environment assessment method based on remote sensing detection data analysis provided by an embodiment of the present invention;

[0023] Figure 5 A schematic diagram of the structure of a mine ecological environment assessment device based on remote sensing detection data analysis provided by an embodiment of the present invention;

[0024] Figure 6 A schematic diagram of the structure of another mine ecological environment assessment device based on remote sensing detection data analysis provided by an embodiment of the present invention;

[0025] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the following embodiments of the present disclosure are described by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0027] In response to the above technical problems, an embodiment of the present invention provides a method and device for evaluating the ecological environment of a mine based on remote sensing detection data analysis, which can effectively improve data extraction capabilities and improve the accuracy of model recognition.

[0028] To facilitate understanding of this embodiment, firstly, a mine ecological environment assessment method based on remote sensing detection data analysis disclosed in an embodiment of the present invention is described. Figure 1 A flowchart of a mine ecological environment assessment method based on remote sensing detection data analysis provided by an embodiment of the present invention is shown. Figure 1 As shown, the method includes:

[0029] Step S102, obtaining the spectral vector data of the mining area collected in advance.

[0030] As a non-destructive means of earth observation, spectral data can reveal the composition and state of surface materials and provide basic information for subsequent analysis. The data of the present invention comes from the spectral vector data obtained by satellite remote sensing detection. The spectral vector data is collected by the spectrometer of the remote sensing satellite and can reflect the spectral reflectance characteristics of different materials in the mining area. The spectral vector data is stored as a structured one-dimensional array, and each data point contains information about its spectral reflectance. The data is stored in a standard CSV format.

[0031] Step S104: input the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model.

[0032] Traditional feature extraction fails to fully consider the local complexity and dynamic changes of data, resulting in low efficiency and insufficient accuracy of the model when processing complex or dynamically changing data. In this regard, the feature extraction model used in the embodiment of the present invention uses the fire ant algorithm to optimize parameters, and based on the feature extraction model, the learning progress of the preset training sample set is updated.

[0033] The fire ant algorithm is a swarm intelligence optimization method that simulates the foraging behavior of fire ants to find the optimal solution to the problem. It is applied to the parameter optimization of the feature extraction model to improve the model's ability to identify key features from spectral data. In addition, the model dynamically updates parameters based on the learning progress of the preset training sample set to further enhance its performance, fully consider the dynamic changes of the data, and, through the learning progress, consider the local complexity of the data, thereby improving the recognition accuracy.

[0034] Step S106, calculating the angle parameter corresponding to the key feature in the preset data distribution dimension, and mapping the key feature to the quantum state based on the angle parameter.

[0035] Step S108, based on the quantum state, classification decisions are made on the key features to determine the mining ecological category indicated by the spectral vector data of the mining area.

[0036] Conventional evaluation methods usually directly classify and evaluate the data after feature extraction. However, traditional methods are often unable to effectively adapt to actual application scenarios with variable data distribution and characteristics, which in turn affects the final classification effect and the breadth of application. The embodiment of the present invention utilizes the superposition and entanglement characteristics of quantum bits to process information. The angle parameter, as an abstract representation of the key feature, is converted into a quantum state, which helps to utilize the parallelism and efficiency of quantum computing. In addition, the use of quantum states to directly participate in the classification process can adapt to the characteristics of different data sets, thereby improving the accuracy and flexibility of classification.

[0037] Step S110, based on the mine ecological category, evaluate the ecological environment of the mine area.

[0038] The classification results include vegetation covered areas, water areas, bare land areas, rock areas, etc. In one embodiment, for the spectral vector data R b =[0.3,0.45,0.6,0.2,......,0.2,0.8,0.55] is classified as “vegetation covered area”.

[0039] Furthermore, the embodiment of the present invention also conducts a mine ecological environment assessment according to the classification category, and evaluates the ecological environment status by calculating the area of ​​regional vegetation coverage areas, water areas, and rock areas. In one embodiment, the method of evaluating ecological quality uses the normalized vegetation index remote sensing ecological indicator to further analyze the health status of vegetation. If the normalized vegetation index is higher than the preset threshold, it means that the vegetation is healthy and has a high coverage.

[0040] In addition, by comparing with historical vegetation coverage data, we analyze coverage change trends, such as the increase or decrease in vegetation coverage, the expansion of exposed areas, etc., in order to evaluate ecological and environmental changes.

[0041] In summary, the embodiment of the present invention provides a mine ecological environment assessment method based on remote sensing detection data analysis, which combines advanced machine learning technology (feature extraction model, fire ant algorithm) and quantum computing principles (quantum state mapping), providing a novel and potentially efficient approach for mine ecological environment assessment.

[0042] Furthermore, based on the above method embodiment, the embodiment of the present invention also provides another mine ecological environment assessment method based on remote sensing detection data analysis. Figure 2 FIG. 4 is a flowchart of another method for evaluating the ecological environment of a mine based on remote sensing data analysis provided by an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0043] Step S202, obtaining the spectral vector data of the mining area collected in advance.

[0044] Step S204: input the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model.

[0045] In the prior art, some solutions use neural networks for feature extraction. In some neural network structures, they may encounter problems such as gradient vanishing, gradient exploding, or falling into local optimal solutions, which will affect the stability of training and the performance of the model.

[0046] The present invention adopts a 6-layer fully connected neural network for feature extraction, and adopts a fire ant optimization neural network based on adaptive feedback as a feature extraction model, performs parameter optimization through the fire ant algorithm, and updates the parameters of the learning progress of the preset training sample set based on the feature extraction model. In the specific implementation, the learning rate and weight are adaptively adjusted according to the different numbers of iterations in the training process. By monitoring the learning progress of each layer of the neural network, the learning rate and weight update strategy are dynamically adjusted, so that the network can automatically adjust its parameters and optimize the learning efficiency when facing the local complexity of the spectral vector data.

[0047] Specifically, the training process of the fire ant optimization neural network algorithm based on adaptive feedback is as follows:

[0048] 1) Train the preset feature extraction neural network with the preset training sample set and evaluate the recognition difference of the feature extraction neural network.

[0049] First, initialize the parameters of the neural network. In one embodiment, the initialization method is expressed as:

[0050]

[0051]

[0052] In the formula, is the initial weight of the neural network, represents the standard deviation of the weight initialization of the neural network, is the bias of the neural network, represents uniform distribution. Preferably, Set to 0.01.

[0053] Further, a training sample set is input. In one embodiment, the training sample set corresponds to the spectral vector data of the above-mentioned mining area. The input spectral vector data is processed by a feature extraction neural network, and is passed through the neural network layer step by step. After applying the activation function to each layer, the corresponding feature representation is obtained. The forward propagation method is expressed as:

[0054]

[0055]

[0056] In the formula, is the spectral vector data input to the neural network, is the linear transformation result of the neural network layer l, is the activation value of the feature of the lth layer of the neural network calculated using the activation function, is the weight parameter of the lth layer of the neural network, is the bias parameter of the lth layer of the neural network, is the input feature of the l-1th layer of the neural network, It is the dynamically adjusted activation function of the neural network.

[0057] In one embodiment, the present invention also uses a dynamically adjusted activation function, which is dynamically adjusted based on the training progress of the network to optimize the model's responsiveness to different features. The calculation method of the activation function is expressed as:

[0058]

[0059] In the formula, is a dynamic adjustment parameter of the activation function, used to control the nonlinearity of the activation function. Preferably, Set to 0.05.

[0060] After each forward pass, the network's output is evaluated against the expected output.

[0061] 2) According to the recognition differences, an adaptive feedback mechanism is used to adjust the learning rate of the feature extraction neural network.

[0062] In the specific implementation, the adaptive feedback mechanism adjusts the learning rate according to the error feedback, and the adjustment method is expressed as:

[0063]

[0064] In the formula, is the initial learning rate of the neural network, is the learning rate of the neurons in the lth layer of the neural network, is the decay coefficient of the learning rate of the neural network, is the number of iterations of the neural network layer l. Preferably, Set to 0.01, Set to 0.95.

[0065] 3) The fire ant algorithm is used to search for the optimal solution of the parameters of the feature extraction neural network. Combined with the gradient descent method, the error gradient of the feature extraction neural network is calculated to update the network parameters of the feature extraction neural network.

[0066] In a specific training cycle, the embodiment of the present invention chooses to start the exploration process of the fire ant algorithm. The fire ants move in the weight space according to the set rules to search for possible better solutions to achieve parameter updates. The update method is expressed as:

[0067]

[0068] In the formula, is the update amount of the weight of the lth layer of the neural network, is the exploration rate of neural network parameter update, is the exploration contribution weight of the kth fire ant, is the potential optimization weight position explored by the kth fire ant, is the number of fire ants; is the feature distance penalty term of the lth layer of the neural network, Yes The gradient of is used to guide the weight update to reflect the optimization distance between features.

[0069] Furthermore, in order to improve the classification accuracy and generalization of the model, the embodiment of the present invention also dynamically adjusts the distance between the vectors of the training sample set. The adaptive feature distance penalty strategy is adopted to enhance the model's ability to maintain the distance between data categories in the feature space. The purpose of the adaptive feature distance penalty is to automatically adjust the distance between feature vectors during the model training process to ensure that the distance between similar features is minimized and the distance between heterogeneous features is maximized. The calculation method of the feature distance penalty term through the lth layer of the network is expressed as:

[0070]

[0071] In the formula, is the feature distance penalty term of the lth layer of the neural network, is the number of samples input to the neural network in the current batch, is the activation value of the i-th data point in the current batch in the l-th layer of the neural network, is the activation value of the jth data point in the current batch in the lth layer of the neural network, The weights are defined based on whether the labels are the same or different.

[0072] Furthermore, the calculation method of the weight defined based on the same or different labels is expressed as:

[0073]

[0074] In the formula, y i is the true label of the i-th spectral vector sample, y j is the true label of the jth spectral vector sample.

[0075] In one embodiment, the exploration contribution weight is adjusted based on the quality of the solution corresponding to each fire ant, and the calculation method is expressed as:

[0076]

[0077] Where, L k represents the quality evaluation of the solution found by the kth fire ant, L j represents the quality assessment of the solution found by the j-th fire ant; is a temperature parameter used to control the randomness of the exploration, similar to the annealing process in simulated annealing. Preferably, the quality assessment method uses the error size to measure. Preferably, Set to 10.

[0078] Furthermore, according to the results of the fire ant exploration and the gradient descent method, the error gradient is calculated and the network weight is updated to ensure more accurate feature extraction in the next training cycle. The update method of the weight parameters of the neural network is expressed as:

[0079]

[0080]

[0081] In the formula, is the loss function of the neural network, is the symbol of partial derivative, is the partial derivative of the loss function with respect to the activation value of the lth layer of the neural network, is the derivative of the dynamically adjusted activation function of the neural network, is the activation value of the feature of the l-1th layer of the neural network calculated using the activation function, Update the symbol for the parameter.

[0082] In one embodiment, the loss function of the neural network is combined with the Frobenius norm to perform constraint calculation to better perform error feedback and model optimization. The calculation method is expressed as:

[0083]

[0084] In the formula, is the true label of the i-th spectral vector sample, The number of samples input to the neural network for the current batch; It is the model prediction value, which is calculated by the Softmax function on the output feature vector of the neural network; is the regularization coefficient of the neural network loss function; The Frobenius norm of the weights of the lth layer of the neural network is used to control the complexity of the model and prevent overfitting. Preferably, Set to 0.01.

[0085] 4) Until the feature extraction neural network meets the preset training requirements, a feature extraction model is constructed based on the feature extraction neural network.

[0086] Repeat the above steps until the preset stop iteration condition is met, which means that the model training is completed. In one embodiment, the preset stop iteration condition is reaching a preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0087] Step S206, calculating the angle parameter corresponding to the key feature in the preset data distribution dimension, and mapping the key feature to the quantum state based on the angle parameter.

[0088] The present invention adopts a quantum-inspired support vector machine algorithm as a classifier model. On the basis of the traditional support vector machine, it dynamically optimizes the decision boundary by adopting the probability amplitude adjustment in quantum computing to adapt to the characteristics of different data sets, thereby improving the flexibility and accuracy of classification.

[0089] First, the embodiment of the present invention maps the data after feature extraction to the quantum state, each feature corresponds to a quantum bit, and the feature value determines the probability amplitude of the corresponding quantum bit. In the specific implementation, the first angle parameter corresponding to the key feature is calculated based on the standard deviation of the key feature; the second angle parameter corresponding to the key feature is calculated based on the mean of the key feature; the third angle parameter corresponding to the key feature is calculated based on the skewness of the key feature; based on the first angle parameter, the second angle parameter, and the third angle parameter, the key feature is mapped to the quantum state through a parameterized single-qubit rotation gate.

[0090] The way to convert real data into quantum data through quantum gates is expressed as:

[0091]

[0092] In the formula, is a parameterized single-qubit rotation gate used to map real eigenvectors to qubits, is the first angle parameter calculated based on the input data, is the second angle parameter calculated based on the input data, is the third angle parameter calculated based on the input data; Is an imaginary unit.

[0093] In one embodiment, the first angle parameter calculated based on the input data is calculated based on the standard deviation of the feature vector to ensure that the rotation can cover different probability amplitudes, thereby reflecting the distribution characteristics of the feature. The calculation method is expressed as:

[0094]

[0095] In the formula, Represents the standard deviation of the input feature vector, multiplied by π to normalize it to the range [0,π]; is the standard deviation function; is the input data of the support vector machine.

[0096] Furthermore, in order to adapt to the asymmetry and complexity of the input data, the second angle parameter and the third angle parameter calculated based on the input data are calculated based on the mean and skewness of the eigenvector, and the calculation method is expressed as follows:

[0097]

[0098]

[0099] In the formula, and Represent the mean and skewness of the input feature vector, respectively, multiplied by 2π to normalize to the range of [0,2π]; is the mean function, is the skewness function.

[0100] Step S208, performing quantum logic gate operation on the quantum state, measuring the quantum state, and determining the category label indicated by the quantum state.

[0101] Furthermore, quantum logic gates are used to operate quantum states and adjust them according to the category labels of the data to form quantum states that can effectively distinguish different categories. Specifically, the logic gate uses a controlled NOT gate operation to achieve entanglement and interaction between quantum bits. By measuring the quantum state multiple times, the collapsed state of the quantum bit is obtained. These states are directly related to the classification label, so as to make classification decisions, which can be expressed as:

[0102]

[0103] In the formula, CNOT is the control NOT gate operation, is the sequence of quantum gates used to generate entanglement between qubits, and I is the identity matrix.

[0104] Among them, the embodiment of the present invention uses a support vector machine to classify and decide on key features, and converts data into quantum states through quantum registers. In the training process of quantum registers, first, the quantum registers are initialized. The function of the quantum registers is to store data features in the form of quantum states, and to process and calculate these data through quantum operations. Each quantum bit represents an aspect of the data feature, ensuring that the quantum state can fully cover the feature space. Specifically, the present invention calculates by converting features into quantum states, which is expressed as:

[0105]

[0106] In the formula, represents the initialized quantum state, is the dimension of the feature; p iis the initialization probability corresponding to the i-th feature, obtained from the statistical distribution of the data; is the symbol for the tensor product of quantum states; Indicates that the quantum bit is in the ground state, corresponding to the 0 of the classical bit; Indicates that the quantum bit is in an excited state, corresponding to 1 of the classical bit.

[0107] In the training process of the support vector machine, the preset training sample set is used to train the preset support vector machine, and, in the training process of the support vector machine, the state of the quantum register corresponding to the quantum state of the training sample set is measured. According to the state, the parameters of the support vector machine are optimized. The parameters are optimized based on the loss function, and are dynamically adjusted according to the importance evaluation results of the key features.

[0108] Specifically, the training process of the support vector machine is as follows:

[0109] During the training process, the state of the quantum register is measured multiple times, and the parameters of the support vector machine are adjusted according to the measurement results. The parameters of the support vector machine include the weight and bias of the support vector machine to minimize the classification error. The classification decision process involves measurement operations. Classification information is obtained by performing quantum measurements. The calculation method of the measurement results of the quantum state is expressed as:

[0110]

[0111] In the formula, Z i is the Pauli-Z observable of the ith qubit; It is the measurement result of the quantum state on this qubit, and takes the value of +1 or -1, representing the two categories of classification; is the measurement result of the quantum state.

[0112] Furthermore, the final classification decision is calculated based on the measurement results, and the category of the data point is determined using the decision function, which is expressed as:

[0113]

[0114] In the formula, f svm () is the decision function of the support vector machine, wu i is the i-th weight parameter of the support vector machine, sign() is the sign function, yu s is the output sample category of the support vector machine.

[0115] During the training process of the support vector machine, the parameters of the support vector machine are optimized through the loss function. The loss function of the support vector machine is expressed as:

[0116]

[0117] Where, L SVM is the loss function of the support vector machine, n nu is the number of samples input into the support vector machine in the current batch, is the true label of the i-th sample input to the support vector machine, It is the measurement result of the quantum state corresponding to the i-th sample input to the support vector machine.

[0118] Furthermore, a dynamic feature importance evaluation mechanism is used to evaluate and adjust the contribution of each feature in the model training process, and the weight parameters of the support vector machine are dynamically adjusted according to its contribution. The calculation method of the feature importance score is expressed as:

[0119]

[0120] In the formula, is the importance score of the ith feature input to the support vector machine, is the i-th eigenvalue of the input data of the support vector machine, is a contribution adjustment factor. Preferably, Set to 2.

[0121] Furthermore, the weight parameters of the support vector machine are updated according to the importance score of the feature, and the updating method is expressed as:

[0122]

[0123] In the formula, is the parameter update operation, is the learning rate for updating the parameters of the support vector machine. Preferably, Set to 0.1.

[0124] Furthermore, the embodiment of the present invention also uses a validation data set separated from the training sample to test the performance of the support vector machine model, and iteratively optimizes until a preset performance indicator is reached. In one embodiment, the preset performance indicator is that the loss function of the support vector machine is less than 0.05, that is, the iteration is stopped, indicating that the model training is completed.

[0125] Step S210: determining the mining ecological category indicated by the mining area spectral vector data based on the category label.

[0126] Step S212, calculating the vegetation index corresponding to the spectral vector data.

[0127] Step S214, conducting a mine ecological environment assessment on the mine area according to the vegetation index.

[0128] In summary, the embodiments of the present invention utilize a trained model to perform a mine ecological environment assessment. In one embodiment, the newly collected spectral vector data is input into a trained feature extraction model for feature processing. Furthermore, the extracted features are input into a classifier model for classifier training to obtain classification results, such as vegetation coverage areas, water areas, bare land areas, rock areas, etc.

[0129] Furthermore, the ecological environment of the mine is assessed according to the classification category. For example, if the classification category of the spectral vector data is "vegetation coverage area", the ecological quality is assessed by using the normalized vegetation index remote sensing ecological indicator to further analyze the health status of the vegetation. If the normalized vegetation index is higher than the preset threshold, it means that the vegetation is healthy and has a high coverage. In addition, by comparing with the historical vegetation coverage data, the coverage change trend is analyzed, such as the increase or decrease of vegetation coverage, the expansion of exposed areas, etc., to assess the ecological environment change.

[0130] In summary, another method for assessing the ecological environment of a mine based on remote sensing data analysis provided by an embodiment of the present invention adopts a neural network based on a fire ant optimization algorithm with adaptive feedback in the feature extraction stage, and automatically adapts to the local complexity of the data by dynamically adjusting the learning rate and weight update strategy, thereby improving the accuracy of feature extraction and the efficiency of network training. In classifier training, a heuristic support vector machine algorithm based on quantum computing is adopted, and the decision boundary is dynamically optimized by adjusting the probability amplitude of the quantum state, adapting to the characteristics of different data sets, thereby improving the accuracy and flexibility of classification.

[0131] Furthermore, based on the above embodiment, the embodiment of the present invention further designs a training sample set, wherein the embodiment of the present invention also provides another method for evaluating the ecological environment of a mine based on remote sensing detection data analysis. Figure 3 The flowchart of the method is shown to illustrate the steps of constructing the training sample set. Figure 4 The flowchart of the mine ecological environment assessment method based on remote sensing detection data analysis provided by the embodiment of the present invention is shown. Figure 3 , the method comprises the following steps:

[0132] Step S10, obtaining the pre-collected initial samples of the mining area, extracting the sequence characteristics and frequency characteristics of the initial samples of the mining area, and generating a composite feature vector.

[0133] The data of the present invention comes from the spectral vector data obtained by satellite remote sensing detection. The spectral vector data is collected by the spectrometer of the remote sensing satellite and can reflect the spectral reflectance characteristics of different materials in the mining area. The spectral vector data is stored as a structured one-dimensional array, and each data point contains information about its spectral reflectance. The data is stored in a standard CSV format.

[0134] Furthermore, the collected data is annotated manually. Professional remote sensing data analysts classify and annotate the spectral vectors based on ground verification information and historical data. The annotated data is the initial sample of the mining area. In one embodiment, a specific spectral vector data R a It can be expressed as: R a =[0.3, 0.45, 0.6, 0.2, ..., 0.2, 0.8, 0.55]. This data reflects the spectral reflectance of the surface material at a specific wavelength. In this embodiment, the labeled category of the data is "vegetation coverage area".

[0135] Furthermore, the embodiment of the present invention also extracts the sequence characteristics and frequency characteristics of the data, generates a composite feature vector, and performs data expansion based on the composite feature vector. It is understandable that in the task of the present invention, the collection, acquisition, annotation and preprocessing of training data are time-consuming and labor-intensive, and insufficient training samples easily lead to poor generalization ability of the model and affect the accuracy of the model. The embodiment of the present invention optimizes the quality of the generated data by fusing the sequence characteristics and frequency domain characteristics of the spectral data and then performing sample expansion.

[0136] Specifically, a batch of spectral vector data is randomly selected from the real spectral data set, and its sequence features and frequency features are extracted through the feature fusion module to form a composite feature vector, which is expressed as:

[0137]

[0138] In the formula, is a composite eigenvector, represents the true spectral vector, is the coefficient of the first characteristic contribution, is the coefficient of the second feature contribution, is the function of fast Fourier transform to extract frequency features, is the function of discrete cosine transform to extract sequence features, is a weighted bias. In one embodiment, the extracted features are as follows: frequency features extracted by fast Fourier transform, sequence features extracted by discrete cosine transform, fundamental frequency amplitude, sum of low-frequency coefficients, harmonic frequency, sum of high-frequency coefficients, energy spectrum density, energy ratio, power spectrum density, mean of low-frequency coefficients, frequency center, mean of high-frequency coefficients, spectrum flatness, standard deviation of low-frequency coefficients, spectrum centroid, standard deviation of high-frequency coefficients, spectrum entropy, low-frequency energy proportion, spectrum roll-off point, high-frequency energy proportion, spectrum contrast, coefficient entropy.

[0139] Furthermore, the coefficient of the first feature contribution and the coefficient of the second feature contribution are adjusted according to the dynamic characteristics of the data, and the adjustment method is expressed as:

[0140]

[0141]

[0142] In the formula, and Is based on the real spectral vector data of the input The score is calculated based on the statistical properties of the

[0143] Furthermore, the processing method based on the statistical characteristics of the input real spectral vector data increases the adaptability of the model to the dynamic changes of the input data. The calculation method is expressed as:

[0144]

[0145]

[0146] In the formula, is the hyperbolic tangent function, is the cosine function, is the i-th feature of the true spectral vector data.

[0147] Step S11, inputting the initial sample of the mining area and the composite feature vector into a preset generative adversarial network to generate a preliminary expanded sample.

[0148] The embodiment of the present invention performs data expansion through a generative adversarial network, wherein the generative adversarial network consists of a generator and a discriminator. The generator is responsible for generating realistic spectral vector data, and the discriminator is responsible for distinguishing the generated data from the real data. The present invention uses a feature fusion module to fuse data in the generator to obtain a composite feature vector to improve the generation quality and stability of the model.

[0149] In the specific implementation, the generator receives random noise input and generates new pseudo-spectral vector data in combination with composite features. The generation method is expressed as:

[0150]

[0151] In the formula, is a random noise vector, are the parameters of the generator, The data generated by the generator, For the generator, is a generator function.

[0152] Step S12, evaluating the quality of the initial expanded sample.

[0153] The discriminator evaluates the real samples and the generated fake samples respectively, and outputs the probability that the sample is a real sample, expressed as:

[0154]

[0155]

[0156] In the formula, is the discrimination result of the real sample, is the discrimination result of the fake sample, are the parameters of the discriminator, is the discriminator, is the discriminator function.

[0157] Step S13, when the quality reaches the preset quality requirement, the preliminary expanded samples are merged with the initial samples of the mining area to construct a training sample set.

[0158] Among them, in order to ensure the quality of the expanded data, the embodiment of the present invention calculates the regularization coefficient of the generator of the generative adversarial network based on the preliminary expanded samples during the training process of the generative adversarial network; and based on the regularization coefficient and the output of the discriminator of the generative adversarial network, adjusts the parameters of the generator and the discriminator of the generative adversarial network so that the quality of the preliminary expanded samples meets the preset quality requirements.

[0159] First, initialize the network parameters of the generator and the discriminator. In one embodiment, the initialization method is expressed as:

[0160]

[0161]

[0162] In the formula, is the initial parameter of the generator, is the initial parameter of the discriminator, A normal distribution with a mean of 0 and a variance of the identity matrix, is a normal distribution, It indicates that it obeys a specific distribution and I indicates the identity matrix.

[0163] Further, the generator is used to perform data expansion to obtain preliminary expanded samples, and the data expansion is referred to the above steps and will not be repeated here. Further, the discriminator is used to discriminate the preliminary expanded samples, and the parameters of the generator and the discriminator are adjusted according to the output of the discriminator.

[0164] For the loss function of the generator, the present invention uses a regularization term to control the quality of the generated samples. The loss value is calculated by combining the Euclidean distance between the generated samples and the real samples. The calculation method of the loss function of the generator is expressed as:

[0165]

[0166] In the formula, is the loss function of the generator; is the regularization coefficient of the generator, which controls the similarity between the generated samples and the real samples, helping to ensure that the generated samples can not only deceive the discriminator, but also be similar to the real samples in statistical properties; Generate loss weights as the basis; is the L2 norm. Preferably, Set to 0.3, Set to 0.7.

[0167] Furthermore, for the loss function of the discriminator, the present invention uses an additional stabilization term to ensure that the discriminator is not too powerful in the early stage of training, thereby preventing the training of the generator from stagnating. The calculation method of the loss function of the discriminator is expressed as:

[0168]

[0169] In the formula, represents the variance of the discriminator output; is the coefficient that controls the influence of variance, which is used to ensure the diversity and discrimination ability of the discriminator output. Further, Set to 0.3.

[0170] Furthermore, the generator and the discriminator respectively update their parameters using the gradient descent method according to their respective loss functions. Repeat the above steps until the preset stop iteration condition is met, indicating that the model training is completed. In one embodiment, the preset stop iteration condition is to reach a preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0171] In summary, an embodiment of the present invention provides another method for assessing the ecological environment of a mine based on remote sensing data analysis. The method uses a generative adversarial network algorithm based on spectral and time-frequency feature fusion for data expansion, which can effectively generate realistic data samples from limited spectral vector data. These data samples are highly similar to real data in frequency domain and time domain characteristics, and can solve the problems of insufficient model training and weak generalization ability caused by insufficient data in traditional methods.

[0172] Furthermore, an embodiment of the present invention also provides a mine ecological environment assessment device based on remote sensing detection data analysis, Figure 5 FIG. 1 shows a schematic diagram of a structure of a mine ecological environment assessment device based on remote sensing detection data analysis provided by an embodiment of the present invention. Figure 5As shown, the device includes: a data acquisition module 100, used to acquire the spectral vector data of the mining area collected in advance; a feature extraction module 200, used to input the spectral vector data into a pre-built feature extraction model to identify the key features of the feature extraction model; wherein the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized by the fire ant algorithm; a data processing module 300, used to calculate the angle parameters corresponding to the key features in the preset data distribution dimension, and based on the angle parameters, map the key features to the quantum state; a classification module 400, used to make classification decisions on the key features based on the quantum state, and determine the mining ecological category indicated by the spectral vector data of the mining area; an evaluation module 500, used to evaluate the ecological environment of the mining area based on the mining ecological category.

[0173] A mine ecological environment assessment device based on remote sensing detection data analysis provided in an embodiment of the present invention has the same technical features as a mine ecological environment assessment method based on remote sensing detection data analysis provided in the above embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0174] Furthermore, based on the above embodiment, the embodiment of the present invention also provides another mine ecological environment assessment device based on remote sensing detection data analysis. Figure 6 FIG. 1 shows a schematic diagram of another mine ecological environment assessment device based on remote sensing detection data analysis provided by an embodiment of the present invention, referring to FIG. Figure 6 The classification module 400 is also used to perform quantum logic gate operations on the quantum state, measure the quantum state, and determine the category label indicated by the quantum state; based on the category label, determine the mine ecological category indicated by the spectral vector data of the mine area. The evaluation module 500 is also used to calculate the vegetation index corresponding to the spectral vector data; and perform a mine ecological environment evaluation on the mine area according to the vegetation index.

[0175] The feature extraction module 200 is also used to train a preset feature extraction neural network through a preset training sample set, evaluate the recognition difference of the feature extraction neural network; according to the recognition difference, use an adaptive feedback mechanism to adjust the learning rate of the feature extraction neural network; use the fire ant algorithm to search for the optimal solution of the parameters of the feature extraction neural network, combine with the gradient descent method, calculate the error gradient of the feature extraction neural network, and update the network parameters of the feature extraction neural network; until the feature extraction neural network meets the preset training requirements, a feature extraction model is constructed based on the feature extraction neural network. The feature extraction module 200 is also used to dynamically adjust the distance between the vectors of the training sample set.

[0176] The above-mentioned data processing module 300 is also used to calculate a first angle parameter corresponding to the key feature based on the standard deviation of the key feature; calculate a second angle parameter corresponding to the key feature based on the mean of the key feature; calculate a third angle parameter corresponding to the key feature based on the skewness of the key feature; and map the key feature to a quantum state through a parameterized single-qubit rotation gate based on the first angle parameter, the second angle parameter, and the third angle parameter.

[0177] Furthermore, an embodiment of the present invention uses a support vector machine to make classification decisions on key features. The above-mentioned classification module 400 is also used to train a preset support vector machine using a preset training sample set, and, during the training process of the support vector machine, measure the state of the quantum register corresponding to the quantum state of the training sample set; according to the state, optimize the parameters of the support vector machine; wherein the parameters are optimized based on the loss function, and are dynamically adjusted according to the importance evaluation results of the key features.

[0178] The embodiment of the present invention also includes a construction module 600, which is used to obtain the pre-collected initial samples of the mining area, extract the sequence characteristics and frequency characteristics of the initial samples of the mining area, and generate a composite feature vector; input the initial samples of the mining area and the composite feature vector into a preset generative adversarial network to generate a preliminary expansion sample; evaluate the quality of the preliminary expansion sample; when the quality reaches the preset quality requirement, merge the preliminary expansion sample with the initial sample of the mining area to construct a training sample set. The above-mentioned construction module 600 is also used to calculate the regularization coefficient of the generator of the generative adversarial network based on the preliminary expansion sample; based on the regularization coefficient and the output of the discriminator of the generative adversarial network, adjust the parameters of the generator and the parameters of the discriminator of the generative adversarial network so that the quality of the preliminary expansion sample meets the preset quality requirement.

[0179] An embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above Figures 1 to 3 The embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to perform the above Figures 1 to 3 Any of the steps of the method shown.

[0180] The embodiment of the present invention also provides a schematic diagram of the structure of an electronic device, such as Figure 7 FIG. 1 is a schematic diagram of the structure of the electronic device, wherein the electronic device includes a processor 71 and a memory 70, the memory 70 stores computer executable instructions that can be executed by the processor 71, and the processor 71 executes the computer executable instructions to implement the above Figures 1 to 3 Any of the methods shown. Figure 7In the illustrated embodiment, the electronic device further includes a bus 72 and a communication interface 73, wherein the processor 71, the communication interface 73 and the memory 70 are connected via the bus 72. The memory 70 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 73 (which may be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. may be used. The bus 72 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. It may also be an AMBA (Advanced Microcontroller Bus Architecture) bus, where AMBA defines three types of buses, including an APB (Advanced Peripheral Bus) bus, an AHB (Advanced High-performance Bus) bus, and an AXI (Advanced eXtensible Interface) bus. The bus 72 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0181] The processor 71 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 71. The above processor 71 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor 71 reads the information in the memory and combines its hardware to complete the above Figures 1 to 3 Any of the methods shown.

[0182] A computer program product of a mine ecological environment assessment method and device based on remote sensing detection data analysis provided in an embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the method described in the previous method embodiment. The specific implementation can be referred to the method embodiment, which will not be repeated here. If the function is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code. Finally, it should be noted that the above embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above embodiments, those skilled in the art should understand that any person skilled in the art can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A mine ecological environment assessment method based on remote sensing detection data analysis, characterized in that: The method comprises: Obtaining pre-collected spectral vector data of the mining area; Inputting the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model; wherein the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized by the fire ant algorithm; Calculating an angle parameter corresponding to the key feature in a preset data distribution dimension, and mapping the key feature to a quantum state based on the angle parameter; Based on the quantum state, a classification decision is made on the key features to determine the mining ecological category indicated by the spectral vector data of the mining area; Based on the mine ecological category, assess the ecological environment of the mine area; Based on the quantum state, the step of making a classification decision on the key feature and determining the mine ecological category indicated by the spectral vector data comprises: Performing a quantum logic gate operation on the quantum state, measuring the quantum state, and determining a category label indicated by the quantum state; Based on the category label, determining the mine ecological category indicated by the mine area spectral vector data; The step of calculating the angle parameters corresponding to the key features in multiple data distribution dimensions, and mapping the key features to quantum states based on the angle parameters, comprises: Calculating a first angle parameter corresponding to the key feature based on a standard deviation of the key feature; Based on the mean of the key feature, calculating a second angle parameter corresponding to the key feature; Based on the skewness of the key feature, calculating a third angle parameter corresponding to the key feature; Based on the first angle parameter, the second angle parameter, and the third angle parameter, the key feature is mapped to a quantum state through a parameterized single-qubit rotation gate.

2. The method according to claim 1, characterized in that: Based on the mine ecological category, the step of evaluating the ecological environment of the mine area comprises: Calculating the vegetation index corresponding to the spectral vector data; Conduct a mine ecological environment assessment of the mining area based on the vegetation index.

3. The method according to claim 1, characterized in that The method for constructing the feature extraction model comprises: The preset feature extraction neural network is trained through a preset training sample set to evaluate the recognition difference of the feature extraction neural network; According to the recognition difference, an adaptive feedback mechanism is used to adjust the learning rate of the feature extraction neural network; The fire ant algorithm is used to search for the optimal solution of the parameters of the feature extraction neural network, and the error gradient of the feature extraction neural network is calculated in combination with the gradient descent method to update the network parameters of the feature extraction neural network; Until the feature extraction neural network meets the preset training requirements, a feature extraction model is constructed based on the feature extraction neural network.

4. The method according to claim 3, characterized in that The method further comprises: The distances between the vectors of the training sample set are dynamically adjusted.

5. The method according to claim 1, characterized in that Performing classification decisions on the key features by using a support vector machine, the method further includes: Using a preset training sample set to train a preset support vector machine, and, during the training process of the support vector machine, measuring the state of the quantum register corresponding to the quantum state of the training sample set; According to the state, the parameters of the support vector machine are optimized; wherein the parameters are optimized based on a loss function and dynamically adjusted according to the importance evaluation result of the key feature.

6. The method according to claim 1, characterized in that The method further comprises: Acquire the pre-collected initial samples of the mining area, extract the sequence characteristics and frequency characteristics of the initial samples of the mining area, and generate a composite feature vector; Inputting the initial sample of the mining area and the composite feature vector into a preset generative adversarial network to generate a preliminary expanded sample; assessing the quality of the initial expansion sample; When the quality reaches the preset quality requirement, the preliminary expanded samples are merged with the initial samples of the mining area to construct a training sample set.

7. The method according to claim 6, characterized in that The method further comprises: Based on the preliminary expanded samples, calculating a regularization coefficient of the generator of the generative adversarial network; Based on the regularization coefficient and the output of the discriminator of the generative adversarial network, the parameters of the generator and the discriminator of the generative adversarial network are adjusted so that the quality of the preliminary expanded samples meets the preset quality requirements.

8. A mine ecological environment assessment device based on remote sensing detection data analysis, characterized in that: The device comprises: A data acquisition module, used to acquire spectral vector data of the mining area collected in advance; A feature extraction module, used for inputting the spectral vector data into a pre-built feature extraction model to identify key features of the feature extraction model; wherein the neural network parameters of the feature extraction model are adaptively adjusted based on the learning progress, and the parameters are optimized by the fire ant algorithm; A data processing module, used to calculate an angle parameter corresponding to the key feature in a preset data distribution dimension, and map the key feature to a quantum state based on the angle parameter; A classification module, used to make a classification decision on the key features based on the quantum state, and determine the mine ecological category indicated by the spectral vector data of the mine area; An evaluation module, used for evaluating the ecological environment of the mining area based on the mining ecological category; The classification module is further used to perform quantum logic gate operations on the quantum state, measure the quantum state, and determine the category label indicated by the quantum state; based on the category label, determine the mine ecological category indicated by the mine area spectral vector data; The data processing module is further used to calculate a first angle parameter corresponding to the key feature based on the standard deviation of the key feature; calculate a second angle parameter corresponding to the key feature based on the mean of the key feature; calculate a third angle parameter corresponding to the key feature based on the skewness of the key feature; and map the key feature to a quantum state through a parameterized single-qubit rotation gate based on the first angle parameter, the second angle parameter, and the third angle parameter.

Images