Data processing method and system for shooting efficiency evaluation

Abstract

The invention provides a data processing method and system for shooting efficiency evaluation, and relates to the technical field of weapon equipment.The method comprises the steps that the model of a weapon to be evaluated is obtained, and a shooting efficiency estimation model matched with the model is determined; obtaining vulnerability parameters of the shooting target, intersection data of ammunition of the to-be-evaluated weapon and the shooting target, and shooting data of an operator using the to-be-evaluated weapon in the shooting; and the shooting data, the vulnerability parameters and the intersection data are input into a pre-trained shooting efficiency estimation model, and the shooting efficiency of the operator using the weapon to be evaluated in the shooting process is obtained. Wherein a dense convolutional network is configured in the shooting efficiency estimation model, and multi-feature extraction can be carried out on shooting data based on vulnerability parameters and rendezvous data. According to the method, through the dense connection characteristic of the dense convolutional network, the extraction capability of high-dimensional nonlinear features is improved, and the stability of model training and the accuracy of an estimation result are improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of weaponry and equipment, and particularly relates to a data processing method and system for shooting efficiency evaluation. Background Art

[0002] Shooting efficiency is an index that measures the relationship between the damage effect on a target and the ammunition consumption under specific shooting conditions. The improvement of shooting efficiency means that a large damage effect can be achieved with small ammunition consumption.

[0003] Currently, shooting efficiency is mostly evaluated by numerical integration methods. Based on ballistics and damage models, this method decomposes the calculation of shooting efficiency into a multi-step iterative process. By inputting parameters such as weapon power parameters, target vulnerability, and environmental conditions, the damage index and ammunition consumption index are gradually solved. This method has a complex calculation process. Especially under high-dimensional parameter conditions, the number of iterations is large, and the calculation amount increases significantly. Moreover, this method relies on approximate solutions, and error accumulation is introduced during the iterative process, resulting in poor calculation accuracy.

[0004] Evaluating shooting efficiency through a BP neural network improves the evaluation efficiency to a certain extent and reduces the problem of error accumulation. However, the feature extraction ability of the BP neural network is limited, it is difficult to process high-dimensional and multi-modal input data, and as the network depth increases, problems such as gradient disappearance or explosion are likely to occur, affecting the training effect. In small-sample data scenarios, the generalization ability is poor, it is difficult to adapt to complex shooting conditions, and the calculation accuracy is unstable.

[0005] In summary, the related technologies have a series of problems such as high computational complexity, weak model feature extraction ability, insufficient generalization ability, scarce training data, and inability to guarantee calculation accuracy, and it is difficult to meet the requirements of quickly and accurately evaluating shooting efficiency under complex shooting conditions. Summary of the Invention

[0006] The main purpose of the present disclosure is to provide a data processing method and system for shooting efficiency evaluation to solve the technical problems of weak model feature extraction ability and inability to guarantee calculation accuracy in the related technologies.

[0007] To achieve the above object, the first aspect of the present disclosure provides a data processing method for shooting efficiency evaluation, including:

[0008] Obtain the model data of the weapon to be evaluated, and determine a shooting efficiency estimation model matching the model data, wherein a dense convolutional network matching the model data is configured in the shooting efficiency estimation model;

[0009] Obtain the vulnerability parameters of the shooting target, the intersection data between the ammunition of the weapon to be evaluated and the shooting target, and the shooting data of the operator using the weapon to be evaluated in this shooting; and,

[0010] Input the shooting data, the vulnerability parameters, and the intersection data into the pre-trained shooting efficiency estimation model respectively, and obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting. Among them, the shooting efficiency estimation model is configured to perform multiple feature extractions on the shooting data based on the vulnerability parameters and the intersection data through the dense convolutional network, and obtain the shooting efficiency result data according to the extracted multiple features.

[0011] Further, the method further includes:

[0012] Establish independent shooting efficiency estimation models for different models of the weapon to be evaluated respectively;

[0013] For each shooting efficiency estimation model, train the shooting efficiency estimation model with the training data samples of the weapon to be evaluated with the matching model until the output of the trained shooting efficiency estimation model reaches the set accuracy, and obtain the pre-trained shooting efficiency estimation model.

[0014] Further, the shooting efficiency estimation model includes a sequentially arranged dense convolutional network and a feature fusion module; and the step of inputting the shooting data, the vulnerability parameters, and the intersection data into the pre-trained shooting efficiency estimation model respectively to obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting includes:

[0015] In response to the received shooting efficiency evaluation instruction, start the dense convolutional network to perform shallow feature extraction and deep feature extraction on the shooting data based on the vulnerability parameters and the intersection data respectively, and obtain a shallow feature time series and a deep feature time series containing the degradation trend information of the shooting data;

[0016] Start the feature fusion module, and perform feature fusion on the shallow feature time series and the deep feature time series through the fully connected layer in the feature fusion module to obtain the shooting efficiency result data.

[0017] Further, each convolutional layer in the dense convolutional network adopts the same configuration. Among them, the first convolutional layer in the dense convolutional network serves as the input convolutional layer for shallow feature extraction, and the other convolutional layers except the first convolutional layer are stacked as dense blocks for deep feature extraction. Each convolutional layer in the dense block is configured to perform zero-padding operation on the convolutional result after each convolution; and,

[0018] The shooting efficiency estimation model further includes a switching module; the switching module is configured to perform feature fusion on multiple deep feature time series output by the dense block, and send the new time series after feature fusion to the feature fusion module to replace the deep feature time series for feature fusion.

[0019] Further, for each shooting efficiency estimation model, obtaining training data samples of the weapon to be evaluated with a matching model to train the shooting efficiency estimation model includes:

[0020] Performing data augmentation processing on the training data samples to generate virtual data samples with the same degradation trend as the training data samples; and,

[0021] Mixing the virtual data samples with the training data samples, and training the shooting efficiency estimation model with the mixed data samples.

[0022] Further, performing data augmentation processing on the training data samples to generate virtual data samples with the same degradation trend as the training data samples includes:

[0023] Obtaining the similarity between any two time series in the training data samples;

[0024] Grouping the entire training data samples based on the similarity, so that the similarity between any two time series in each group of samples is higher than a set threshold;

[0025] Generating a virtual group based on each group of samples, and integrating to obtain virtual data samples, so that the virtual data samples have the same degradation trend as the training data samples.

[0026] Further, the training data samples contain time series with different data lengths, and the similarity Dis j and O k between any two time series in the training data samples DTW is configured as:

[0027] Dis DTW = sqrt(dp)

[0028] Among them,

[0029]

[0030] d(O j ,O k ) = (O j -O k ) 2

[0031] In the formula, dp is a matrix, the minimum value of m is 2, the maximum value is the data length of the O j , the minimum value of n is 2, the maximum value is the data length of the Ok, d(O j ,O k ) is the square of the Euclidean distance between the O j and the Ok, dp[m, n] is the element in the m-th row and n-th column of the dp, dp[m - 1, n] is the element in the (m - 1)-th row and n-th column of the dp, dp[m, n - 1] is the element in the m-th row and (n - 1)-th column of the dp, and dp[m - 1, n - 1] is the element in the (m - 1)-th row and (n - 1)-th column of the dp.

[0032] Furthermore, the training data samples contain time series with the same data length, and, j and O k The similarity Dis DTW of is configured as:

[0033]

[0034] In the formula, O j (t), O k (t) respectively represent the values of the time series O j and Ok at the time point t, and TC represents the data length of the time series O j , O k .

[0035] Furthermore, the intersection data includes the relative motion trajectory and attitude of the ammunition and the shooting target; and,

[0036] The input of the shooting efficiency estimation model further includes the damage power parameter and climate condition of the weapon to be evaluated;

[0037] The shooting efficiency estimation model is trained by the butterfly optimization algorithm to determine the optimal hyperparameter combination in the shooting efficiency estimation model.

[0038] The second aspect of the present disclosure provides a data processing system for shooting efficiency evaluation, including:

[0039] A model screening unit, configured to obtain the model data of the weapon to be evaluated, and determine a shooting efficiency estimation model that matches the model data, wherein a dense convolutional network that matches the model data is configured in the shooting efficiency estimation model;

[0040] A data acquisition unit, configured to obtain the vulnerability parameters of the shooting target, the intersection data between the ammunition of the weapon to be evaluated and the shooting target, and the shooting data of the operator using the weapon to be evaluated in this shooting; and,

[0041] A data processing unit, configured to respectively input the shooting data, the vulnerability parameters, and the intersection data into the pre-trained shooting efficiency estimation model, and obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting, wherein the shooting efficiency estimation model is configured to perform multi-feature extraction on the shooting data based on the vulnerability parameters and the intersection data through the dense convolutional network, and obtain the shooting efficiency result data according to the extracted multi-features.

[0042] In the data processing method for shooting efficiency evaluation provided by the embodiments of the present disclosure, the method of using a dense convolutional network for shooting efficiency evaluation, by introducing a dense convolutional network, utilizes its dense connection characteristics to enhance the information flow between layers in the network, improves the extraction ability of high-dimensional non-linear features, effectively solves the problem of gradient disappearance, and improves the stability of model training and the accuracy of shooting efficiency result data, achieving the purpose of efficient calculation of shooting efficiency, thereby realizing the technical effect of quickly and accurately evaluating shooting efficiency under complex shooting conditions, and further solving the technical problems of high computational complexity and weak model feature extraction ability caused by the related art using numerical integration algorithms or BP neural networks. Brief Description of the Drawings

[0043] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the related art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the related art. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0044] Figure 1 It is a schematic flowchart of the data processing method for shooting efficiency evaluation provided by the embodiments of the present disclosure;

[0045] Figure 2 It is a schematic flowchart of the training process of the shooting efficiency estimation model provided by the embodiments of the present disclosure;

[0046] Figure 3Schematic diagram of the training process of the shooting efficiency estimation model provided by the embodiments of the present disclosure;

[0047] Figure 4 Schematic diagram of the process of training the shooting efficiency estimation model by the butterfly optimization algorithm provided by the embodiments of the present disclosure;

[0048] Figure 5 Block diagram of the data processing system for shooting efficiency evaluation provided by the embodiments of the present disclosure;

[0049] Figure 6 Block diagram of the electronic device provided by the embodiments of the present disclosure.

[0050] Reference numerals

[0051] P - flavor concentration; rand - set value. Detailed implementation manners

[0052] In order to enable those skilled in the art to better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

[0053] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above - mentioned drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so as to describe the embodiments of the present disclosure here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non - exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these process, method, product or device.

[0054] In the present disclosure, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", etc. is the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. These terms are mainly used to better describe the present disclosure and its embodiments, and are not used to limit that the indicated device, element or component must have a specific orientation, or be constructed and operated in a specific orientation.

[0055] Moreover, in addition to being used to indicate orientation or positional relationships, some of the above terms may also be used to indicate other meanings. For example, the term "upper" may also be used to indicate a certain attachment relationship or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in this disclosure can be understood according to specific circumstances.

[0056] In addition, the terms "configured", "set", "provided with", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or there is internal communication between two devices, components or parts. For those of ordinary skill in the art, the specific meanings of the above terms in this disclosure can be understood according to specific circumstances.

[0057] The term "Dense Convolutional Network" refers to a deep learning model that combines a Convolutional Neural Network (CNN) and a dense connection strategy. Dense connection means that each layer in the network is connected to all the previous layers, so that each layer can receive all the features from the previous layer. This structure effectively enhances the information flow, avoids the problem of gradient vanishing, and improves the expression ability of the network through multiple feature reuses.

[0058] The term "firing efficiency" refers to an index that measures the relationship between the damage effect on the target and the ammunition consumption under specific firing conditions. Firing efficiency is usually composed of two key indexes: the damage index and the ammunition consumption index. The damage index reflects the degree of damage caused to the target after firing, while the ammunition consumption index measures the amount of ammunition required to complete the firing task. The improvement of firing efficiency means that the maximum damage effect can be achieved with the minimum ammunition consumption. On the battlefield, quickly and accurately evaluating firing efficiency is of great significance for formulating combat training plans, optimizing weapon configurations, and improving strike accuracy.

[0059] The term "hyperparameter optimization" refers to an important link in the training of deep learning models, which involves optimizing various hyperparameters (such as learning rate, regularization parameter, etc.) in the model training process to improve the performance of the model. The selection of hyperparameters has a crucial impact on the convergence speed, training stability, and final estimation accuracy of the model.

[0060] The term "gradient vanishing problem" refers to the problem that during the training of a dense convolutional network, the gradient gradually becomes smaller during the backpropagation process, eventually causing the network to be unable to effectively update the weights and affecting the learning ability of the model. In traditional convolutional neural networks, due to the stacking of multiple layers and complex activation functions, the gradient vanishing problem is particularly prominent.

[0061] The term "Dynamic Time Warping" refers to an algorithm for measuring the similarity between two time series of different lengths. It has extensive applications in fields such as speech recognition and signal processing. The core idea is to stretch or compress one of the time series on the time axis to find the best matching way between the two.

[0062] The term "weighted average method" refers to calculating the numerical result of the change of a set of indicators or data driven by multiple factors based on statistical principles. It can not only represent the average value of a set of data, but also take into account the individual values, and can effectively solve the situation where there is a certain influence between different items.

[0063] The term "feature extraction" refers to automatically extracting key information helpful for the task from the original data.

[0064] The term "regression task" refers to a task of predicting a continuous numerical value based on the input data. In this disclosure, a dense convolutional network is used for the regression task, that is, predicting the shooting efficiency. The regression task usually needs to learn the relationship between the input and the output through training data, and requires the model to make accurate predictions for unseen data. The dense convolutional network extracts multi-level feature information and inputs this information into the regression layer to predict the specific value of the shooting efficiency. During the training process, the dense convolutional network adopts the hyperbolic tangent (Tanh) activation function and the RMSProp optimization algorithm, further improving the convergence speed and stability of the model. Finally, the regression result output by the model is used to judge the efficiency under specific shooting conditions, providing a decision-making basis for combat training command.

[0065] The term "pooling operation" refers to an important part in a convolutional neural network (CNN), mainly used to reduce the dimension of the feature map, thereby reducing the computational complexity and memory consumption of the model.

[0066] The term "Butterfly Optimization Algorithm" refers to a heuristic algorithm that simulates the food search and environmental adaptation of a butterfly group. The algorithm simulates how butterflies perform local and global searches according to external conditions when looking for the optimal food position, so as to find the optimal solution. The algorithm has few parameters, simple principles, and is easy to implement.

[0067] The term "optimization algorithm with adaptive learning rate" (Root Mean Square Propagation, abbreviated as RMSProp optimization algorithm) is mainly used for parameter update in deep learning. The RMSProp optimization algorithm adjusts historical information by introducing a decay coefficient, making the adjustment of the learning rate smoother, thus avoiding the problem of continuous decline of the learning rate. The core idea is to use different learning rates for each parameter, and these learning rates are adaptively adjusted according to the magnitude of the recent gradient of the parameter. The exponential weighted moving average of the squared gradient can be used to adjust the learning rate, making the adjustment of the learning rate smoother.

[0068] It should be noted that, without conflict, the embodiments in this disclosure and the features in the embodiments can be combined with each other. The following will detail this disclosure with reference to the accompanying drawings and in combination with the embodiments.

[0069] There are technical problems in the related art such as weak model feature extraction ability and inability to guarantee calculation accuracy. In particular, when using the numerical integration method, the calculation of shooting efficiency involves multi-dimensional factors such as ballistic trajectory, environmental disturbance, and target characteristics, and the cost of obtaining complete and balanced data is too high. The data processing method for shooting efficiency evaluation based on the BP neural network often has difficulty in ensuring the robustness of the evaluation model in the case of insufficient data or large noise, which limits its application in actual combat training.

[0070] To solve the above technical problems, the embodiments of this disclosure provide a data processing method for shooting efficiency evaluation. Figure 1 It is a schematic flowchart of the data processing method for shooting efficiency evaluation provided by the embodiments of this disclosure. As Figure 1 shown, the method includes the following steps S11 to step S13.

[0071] Step S11: Obtain the model data of the weapon to be evaluated, and determine a shooting efficiency estimation model matching the model data. Among them, the shooting efficiency estimation model is configured with a Dense Convolutional Network (DenseNet) matching the model data.

[0072] Step S12: Obtain the vulnerability parameters of the shooting target, the intersection data of the ammunition of the weapon to be evaluated and the shooting target, and the shooting data of the operator using the weapon to be evaluated in this shooting.

[0073] Step S13: Input the shooting data, vulnerability parameters, and intersection data into the pre-trained shooting efficiency estimation model respectively, and obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting. Among them, the shooting efficiency estimation model is configured to perform multiple feature extractions on the shooting data based on the vulnerability parameters and intersection data through the dense convolutional network, and obtain the shooting efficiency result data according to the extracted multiple features.

[0074] It should be noted that the shooting data may include, in addition to the shooting result data (target ring scores, with the outermost target ring being 1 point and the bull's-eye being 10 points) of the operator using the weapon to be evaluated for the shooting target, shooting postures (standing posture, kneeling posture, and lying posture), etc. Among them, the input end of each convolutional layer of the dense convolutional network is set to be connected to the output ends of all the convolutional layers located in front of the convolutional layer. In the shooting efficiency estimation model, the function of obtaining the shooting efficiency result data based on the extracted multiple features can be completed through the output convolutional layer of the dense convolutional network, or the function of obtaining the shooting efficiency result data based on the extracted multiple features can be completed through a separately set module.

[0075] To better capture the shooting data features in the shooting task, a dense convolutional network is configured in the shooting efficiency estimation model of the present disclosure, which is one of the important inventive points of the present disclosure. Different from the traditional convolutional neural network, the dense convolutional network introduces a dense connection design in the network structure, enabling each convolutional layer to be connected to all the previous layers, thereby effectively transmitting information between layers and avoiding the problem of gradient disappearance. The dense connection design enables the model to continuously reuse the features extracted by the previous layers, thereby improving the information flow efficiency and enhancing the model's ability to extract high-dimensional non-linear features. The above embodiments provide a data processing method for shooting efficiency evaluation based on a dense convolutional network, aiming to improve the evaluation accuracy and calculation speed of the shooting efficiency result data through deep learning technology. The obtained shooting efficiency is an important indicator for measuring the completion degree of the shooting task and serves as an indicator for measuring the relationship between the target damage effect and the ammunition consumption under specific shooting conditions.

[0076] The above data processing method for shooting efficiency evaluation provided by the embodiments of the present disclosure is applicable to the shooting efficiency evaluation scenarios of any existing weapon. This method has universality, and the embodiments of the present disclosure do not limit this.

[0077] The embodiments of the present disclosure also provide a data processing system, a storage medium, and an electronic device for shooting efficiency evaluation corresponding to the above method.

[0078] Compared with the related art, the above data processing method for shooting efficiency evaluation provided by the embodiments of the present disclosure uses a dense convolutional network for shooting efficiency evaluation. By introducing a dense convolutional network and utilizing its dense connection characteristics to enhance the information flow between layers in the network, the ability to extract high-dimensional non-linear features is improved, the problem of gradient disappearance is effectively solved, and the stability of model training and the accuracy of shooting efficiency result data are improved, achieving the purpose of efficient calculation of shooting efficiency. Thus, the technical effect of quickly and accurately evaluating shooting efficiency under complex shooting conditions is realized, and further the technical problems of high computational complexity and weak model feature extraction ability caused by the related art using numerical integration algorithms or BP neural networks are solved.

[0079] Based on the method shown in Figure 1 Preferably, the shooting data in steps S12 and S13 is set as the processed data obtained by preprocessing the original data acquired by the sensor. According to actual data processing requirements, the preprocessing can be set to include data cleaning, data conversion, data description, etc., and the embodiments of the present disclosure do not limit this.

[0080] Based on the method shown in Figure 1 Preferably, the method further includes the following steps S01 to S02 to implement the training of the shooting efficiency estimation model, as Figure 2 shown.

[0081] Step S01: For different models of weapons to be evaluated, establish independent shooting efficiency estimation models respectively.

[0082] Step S02: For each shooting efficiency estimation model, train the shooting efficiency estimation model with the training data samples of the weapons to be evaluated with matching models until the output of the trained shooting efficiency estimation model reaches the set accuracy, and obtain a pre-trained shooting efficiency estimation model.

[0083] The embodiments of the present disclosure establish independent shooting efficiency estimation models for each model of weapon to be evaluated to improve the accuracy of the shooting efficiency evaluation results. The process of model training is also one of the important inventive points of the present disclosure and will be introduced in detail later. It will not be elaborated here.

[0084] Based on the method shown in Figure 1Based on the method shown, preferably, the shooting efficiency estimation model includes a densely connected convolutional network and a feature fusion module arranged in sequence. Among them, the densely connected convolutional network is configured to perform shallow feature extraction and deep feature extraction on shooting data based on vulnerability parameters and intersection data respectively, and obtain a shallow feature time series and a deep feature time series containing the degradation trend information of the shooting data. The feature fusion module is configured to perform feature fusion on the shallow feature time series and the deep feature time series through the fully connected layer in the feature fusion module to obtain shooting efficiency result data. The convolutional layer for shallow feature extraction can be set as the first convolutional layer of the densely connected convolutional network or other convolutional layers, and the convolutional layer for deep feature extraction can be set as the output convolutional layer of the densely connected convolutional network and the intermediate convolutional layers added. The embodiments of the present disclosure do not limit this.

[0085] Correspondingly, step S13 further includes sub-step S131 and sub-step S132.

[0086] Sub-step S131: In response to the received shooting efficiency evaluation instruction, start the densely connected convolutional network to perform shallow feature extraction and deep feature extraction on shooting data based on vulnerability parameters and intersection data respectively, and obtain a shallow feature time series and a deep feature time series containing the degradation trend information of the shooting data.

[0087] Sub-step S132: Start the feature fusion module, and perform feature fusion on the shallow feature time series and the deep feature time series through the fully connected layer in the feature fusion module to obtain shooting efficiency result data.

[0088] Exemplarily, in most cases, it can be set that: the first convolutional layer in the densely connected convolutional network is used as the input convolutional layer for shallow feature extraction, and the other convolutional layers except the first convolutional layer are stacked as a DenseBlock for deep feature extraction. The input shooting data first undergoes shallow feature extraction through the first convolutional layer, and this convolutional layer uses multiple one-dimensional filters to perform convolution operations on the time series data to extract the time series features in the original data. Then, further deep feature extraction is performed through a DenseBlock containing multiple stacked convolutional layers. Each convolutional layer uses the same configuration as the input convolutional layer, and zero padding is used for the convolution operation to maintain the dimension of the feature map.

[0089] The dense block is the core module of the dense convolutional network and one of the core technologies of the entire data processing method for shooting efficiency evaluation. It can achieve dense connections by reusing feature maps. Through the design of the dense block, close connections are established between each layer of the network, effectively promoting the flow and sharing of information, so as to better capture complex temporal and spatial features. Compared with traditional networks, the dense convolutional network with this preferred setting can retain the fine-grained features of the input data while extracting richer deep semantic information, improving the accuracy and robustness of shooting efficiency evaluation. It can be found that the output H of the nth convolutional layer in the dense convolutional network n can be expressed as:

[0090] H n = C([H 0 , H 1 , H 2 ,..., H n-1 ) (1)

[0091] In the formula, C([H 0 , H 1 , H 2 ,..., H n-1 ) is the operation of concatenating feature maps before the nth convolutional layer, and H 0 is the output of the first convolutional layer, and so on.

[0092] The difference between the input and output of the dense block lies in the number of channels. Assume that the number of input channels of the dense block is k 0 , then the number of output channels of the dense block of the nth layer is k 0 + k × n, where k is the growth rate.

[0093] The feature maps of the input convolutional layer and the dense block are input into the fully connected layer in the feature fusion module for the final shooting efficiency regression calculation, and accurate shooting efficiency result data can be quickly obtained. The fully connected layer fuses shallow features and deep features and outputs the final shooting efficiency result data. A neuron is connected at the end of the fully connected layer to perform the regression operation of shooting efficiency and obtain an accurate shooting efficiency value. During the training process, all layers use the hyperbolic tangent (Tanh) activation function for calculation to ensure that the model can converge quickly during the learning process. The training can be carried out through the RMSProp optimization algorithm, which further improves the performance of the model.

[0094] As a preferred implementation, each convolutional layer in the dense convolutional network adopts the same configuration, and each convolutional layer in the dense block is configured to perform zero-padding operation on the convolution result after each convolution. This design ensures that the spatial features of the shooting data will not be lost and can extract the subtle changes in the shooting data.

[0095] As another preferred embodiment, the shooting efficiency estimation model further includes a switching module. The switching module is configured to perform feature fusion on the multiple deep feature time series output by the dense block, and send the new time series after feature fusion to the feature fusion module for further feature fusion instead of the original deep feature time series. To reduce the dimension of the feature map extracted from the dense block, the switching module can convert the large-size multi-channel feature map into a small-size few-channel feature map, thereby constructing a more compact shooting efficiency estimation model. Among them, the switching module, as a switching layer, can also be integrated into the dense convolutional network, and the switching layer is inserted at the dense block of the dense convolutional network.

[0096] Specifically, the switching module can fuse multiple feature maps output by the dense block. The switching module can adopt a convolutional layer with a size of 3×1 to merge multiple feature maps from the dense block into a unique feature map. Through the switching module, the spatial dimension of the feature map is compressed, reducing the number of network parameters, avoiding overfitting, and accelerating the model training process. It is worth noting that the switching module removes the pooling operation, which helps to retain more useful feature information, thereby improving the prediction performance of the network.

[0097] Based on the method shown in Figure 2 preferably, the training process in step S02 further includes the following sub-steps S021 and S022.

[0098] Sub-step S021: Perform data augmentation on the training data samples to generate virtual data samples with the same degradation trend as the training data samples.

[0099] Sub-step S022: Mix the virtual data samples with the training data samples, and train the shooting efficiency estimation model with the mixed data samples.

[0100] Among them, the first key stage of the training process in sub-step S021, that is, the data augmentation stage, aims to expand the training data samples by generating virtual data samples with the same degradation characteristics. The original shooting data often affects the training effect and generalization ability of the model due to limited sample quantity. Therefore, preferably, data augmentation is performed by introducing dynamic time warping and weighted average methods, which can effectively expand the scale of the training data samples and improve the evaluation accuracy.

[0101] Specifically, sub-step S021 further includes: obtaining the similarity between any two time series in the training data samples; grouping the entire training data samples based on the similarity, so that the similarity between any two time series in each grouped sample is higher than the set threshold; generating a virtual group based on each grouped sample, and integrating to obtain virtual data samples, so that the virtual data samples have the same degradation trend as the training data samples.

[0102] Since the training data samples may contain time series with the same data length and time series with different data lengths, various methods can be used to obtain the similarity between any two time series.

[0103] Example 1: For the case where the training data samples contain time series with different data lengths, the similarity Dis j and O k between any two time series O in the training data samples can be configured as: DTW Dis

[0104] Dis DTW = sqrt(dp) (2)

[0105] where

[0106]

[0107] d(O j , O k ) = (O j - O k ) 2

[0108] In the formula, dp is a matrix, the minimum value of m is 2, the maximum value is the data length of O j , the minimum value of n is 2, the maximum value is the data length of O k , d(O j , O k ) is the square of the Euclidean distance (Euclidean distance) between O j and O k , dp[m,n], dp[m-1,n], dp[m,n-1], dp[m-1,n-1] are the corresponding elements of the matrix dp respectively, where dp[m,n] is the element in the m-th row and n-th column of the matrix dp, dp[m-1,n] is the element in the (m-1)-th row and n-th column of the matrix dp, dp[m,n-1] is the element in the m-th row and (n-1)-th column of the matrix dp, and dp[m-1,n-1] is the element in the (m-1)-th row and (n-1)-th column of the matrix dp.

[0109] The similarity between different time series can be calculated through the dynamic time warping of formula (2). After obtaining the optimal pair of two time series data points, the dynamic similarity of the time series can be obtained during the pairing process. Given two time series S = s 1 , s 2 ,..., s m and T = t 1 , t 2 ,..., t n, construct a matrix dp of size m×n to calculate the similarity between these time series. dp[m,n] is an element of the matrix dp. The similarity between any two time series O j and O k can be set as the square root of each element of the matrix dp. The smaller the similarity value, the higher the similarity.

[0110] Example 2: For the case where the training data samples contain time series with the same data length, the similarity Dis j between any two time series O k and O DTW in the training data samples is configured as:

[0111]

[0112] where O j (t) and O k (t) represent the values of the time series O j and Ok at time point t respectively, and TC represents the data length of the time series O j , O k .

[0113] The similarity between different time series can be calculated by the dynamic time warping of formula (3). It allows stretching or compressing the data of the time series, solving the problem of inconsistent shooting data lengths. By calculating Dis DTW to evaluate the similarity between two time series, data with a higher similarity can be used for data augmentation. The time series Q j = s 1 , s 2 ,..., s m , and the time series Q k = t 1 , t 2 ,..., t n .

[0114] Through the above formulas (2) and (3), the embodiments of the present disclosure introduce a time series generation method based on weighted dynamic time warping (DBA). By calculating the similarity between data, virtual data with similar degradation characteristics is generated, expanding the training data set and enhancing the generalization ability of the model. Through the above data augmentation technology of the weighted average method based on a dense convolutional network and dynamic time warping, a new data processing method for shooting efficiency evaluation is constructed. At the same time, the original training data samples are expanded through virtual data samples, alleviating the problem of insufficient small sample data and significantly enhancing the generalization ability of the model. In the overall model design, the dense convolutional network is combined with the fully connected layer, and through the fusion of deep features and shallow features, the efficient calculation of shooting efficiency is realized, meeting the requirements of real-time damage effectiveness evaluation in modern combat training.

[0115] This data augmentation technology enables the DCRN to maintain a high evaluation accuracy when facing complex shooting tasks.

[0116] Regarding the generation method of virtual data, time series of original shooting data with relatively high similarity are selected and combined into new virtual data. For this purpose, a weighted average method is used to synthesize multiple similar time series into one virtual data. Suppose N similar time series O 1 , O 2 ,..., O N are selected, and a weight w i is assigned to each time series. Then the virtual data V j can be expressed as:

[0117]

[0118] Among them,

[0119]

[0120] In the formula, w i is the weight of each original time series, and the sum of the weights is 1, where i = 1,..., N.

[0121] Through the above process, within each virtual group, the time series corresponding to multiple shooting data with similar characteristics can be synthesized into a virtual data, and the generated virtual data has the same degradation trend and characteristics as the original shooting data. The weighted average method aims to reduce the size of the time series dataset by calculating the similarity between multiple time series for classification tasks. It first selects a time series as the initial average sequence and then iteratively updates it to minimize the sum of the similarities between the average sequence and all other sequences. The above data augmentation method is designed for time series classification tasks and aims to generate data for specific categories. Data with multiple consecutive shooting efficiency labels can be generated for accurate shooting efficiency estimation.

[0122] To enhance data diversity, a similarity-based weighted assignment strategy can be adopted, that is, smaller samples with higher similarity will be assigned larger weights to ensure that the generated virtual data does not deviate from the degradation trend of the original shooting data. In actual operation, the weights can be assigned in the following way: from N similar time series, select two time series with a weight of 0.5, and the sum of the weights of the other time series is 0.5, which can be evenly distributed or weighted according to similarity.

[0123] Sub-step S022 is the second key stage of the training process, that is, the model training stage. The generated virtual data samples for training are combined with the original training data samples to form an augmented dataset. The augmented dataset contains more data samples with similar degradation characteristics, providing rich and diverse training data for the subsequent shooting efficiency estimation model. Through this data augmentation method, the scale of the training dataset is effectively expanded, ensuring that the shooting efficiency estimation model can still maintain high accuracy and robustness in the case of insufficient data. At the same time, the augmented dataset can help the model better adapt to various shooting conditions and improve the accuracy and stability of the shooting efficiency result data.

[0124] Figure 3 This is a schematic diagram of the training process of the shooting efficiency estimation model provided by the embodiments of the present disclosure. The model training process can be divided into two stages: data augmentation and model training. In this framework, the raw data obtained from the sensor undergoes data preprocessing. Then, virtual data samples with similar degradation characteristics to the training data samples are generated. Next, the generated virtual data samples and the actual training data samples (such as fault data) are mixed to train the shooting efficiency estimation model.

[0125] In Figure 1 Based on the method shown, preferably, the intersection data can be set to include the relative motion trajectory and attitude of the ammunition and the shooting target.

[0126] In Figure 1Based on the method shown, preferably, the input of the shooting efficiency estimation model further includes the damage power parameter of the weapon to be evaluated and the climate conditions.

[0127] On the Figure 1 Based on the method shown, preferably, in step S02, the shooting efficiency estimation model is trained by the butterfly optimization algorithm to determine the optimal combination of hyperparameters in the shooting efficiency estimation model. Figure 4 FIG. is a schematic flowchart of training a shooting efficiency estimation model by the butterfly optimization algorithm provided by an embodiment of the present disclosure. Through a further data processing method for shooting efficiency evaluation based on a dense convolutional network and the butterfly optimization algorithm, it aims to improve the accuracy and calculation speed of shooting efficiency evaluation by optimizing the combination of hyperparameters of the model, and achieve fast and accurate evaluation. The butterfly optimization algorithm is used to optimize the key hyperparameters in the dense convolutional network, thereby effectively improving the accuracy of the estimation results of the evaluation model in a complex combat training environment.

[0128] The butterfly optimization algorithm is an intelligent optimization algorithm based on the foraging behavior of butterflies in nature, and is used to optimize the key hyperparameters of the dense convolutional network. Specifically, by simulating the interaction mechanism of butterflies using "perception" information during the process of searching for food, the dynamic balance between global and local searches is achieved. These hyperparameters, such as the number of convolutional layers, the size of convolutional kernels, the learning rate, the regularization parameter, and the batch size, etc., are used to improve the performance of the model. Hyperparameter optimization is a key step in model training. The butterfly optimization algorithm can find the optimal combination of hyperparameters through global search, thereby improving the performance of the model and the accuracy of the estimation results. Compared with traditional optimization methods, the butterfly optimization algorithm has a strong global search ability when dealing with high-dimensional and complex parameter spaces, and can effectively avoid the local optimum problem.

[0129] First, the butterfly optimization algorithm randomly generates a set of initial solutions. Each solution corresponds to a set of hyperparameter configurations of the dense convolutional network, and these solutions are called "butterflies". Each solution is evaluated through a fitness function (such as the loss value or estimation accuracy of the model on the validation set) to measure its performance for the shooting efficiency estimation task. Solutions with high fitness represent better combinations of hyperparameters.

[0130] Next, the butterfly optimization algorithm utilizes the behavioral characteristics of butterflies to perform searches globally and fine-tune locally. During the search process, butterflies will dynamically update the hyperparameter configurations according to their own states and the states of other butterflies, thereby continuously optimizing the solutions. By introducing a trade-off mechanism between global search and local search, the butterfly optimization algorithm can efficiently approach the optimal hyperparameters in a complex high-dimensional search space.

[0131] Finally, after multiple iterations, the Butterfly Optimization Algorithm finds a set of hyperparameter configurations that can maximize the performance of the Dense Convolutional Network, thereby significantly improving the accuracy of the model's estimation results and the training stability in the shooting efficiency evaluation. This optimization method effectively solves the problems of low efficiency and easy entrapment in local optima in traditional manual hyperparameter tuning, providing an efficient and reliable solution for real-time damage effectiveness evaluation in complex combat training environments.

[0132] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0133] From the above description, it can be seen that the present disclosure achieves the following technical effects:

[0134] 1. To achieve accurate and rapid shooting efficiency evaluation, a method based on the Dense Convolutional Network is proposed. By combining the characteristics of the Dense Convolutional Network, an efficient shooting efficiency estimation model is designed, which can quickly evaluate the damage effect under given shooting conditions. Through data augmentation and deep learning techniques, deep features and shallow features in the shooting data are extracted, thereby realizing high-precision and high-efficiency shooting efficiency calculation. An efficient and accurate method for rapid shooting efficiency evaluation is provided, which can provide real-time shooting efficiency evaluation support for combat training commanders, assisting in decision-making and optimizing combat training strategies.

[0135] 2. In the first stage of model training, namely the data augmentation stage, the original training data samples are extended through data augmentation techniques, thereby overcoming the problem of insufficient original training data samples. By introducing the weighted dynamic time warping and weighted average methods, virtual data samples consistent with the degradation characteristics of the original training data samples are generated and mixed with the original training data samples. The generated virtual data samples not only retain the degradation trend of the original training data samples but also enhance the generalization ability of the model, enabling the network to maintain a high estimation accuracy under different shooting conditions and solving the problem of insufficient sample data.

[0136] 3. In the second stage of model training, namely the shooting efficiency evaluation stage, the augmented data samples are then input into the shooting efficiency estimation model based on the Dense Convolutional Network to extract high-dimensional feature information in the shooting data. By combining data augmentation and feature extraction, the accuracy and robustness of shooting efficiency evaluation can be effectively improved.

[0137] 4. High - efficient feature extraction is achieved. The dense convolutional network extracts time - series features in the shooting task through multiple convolutional layers, including shallow features and deep features. These features help the model accurately predict the shooting efficiency under different shooting conditions. Shallow features usually refer to local patterns and simple rules in the input data, while deep features are complex patterns and global trends extracted by the network through multiple convolutional operations. These features help the model identify and understand the complex relationships in the data, thereby improving the accuracy of shooting efficiency evaluation. The dense connection mechanism of the dense convolutional network enables the network to reuse shallow features and deep features, thus effectively improving the accuracy of shooting efficiency evaluation. Through the dense connection structure of the dense convolutional network, the information flow between layers is strengthened, ensuring that the gradient can be effectively transmitted and accelerating the training of the network. This design avoids the problem of gradient vanishing, ensuring that the dense convolutional network can fully learn the features in the data and obtain high - precision estimation results in shooting efficiency evaluation. The dense block can effectively extract deep features in the time series and capture subtle degradation information in the shooting task. The key advantages of the dense convolutional network lie in its strong feature extraction ability and good gradient propagation characteristics, which can quickly and efficiently evaluate the shooting efficiency.

[0138] 5. The problem of information flow and gradient vanishing is solved. The dense connection design ensures the effective transmission of information between layers, avoids the problem of gradient vanishing, and improves the training efficiency and the stability of the model.

[0139] 6. Fast and accurate shooting efficiency evaluation is achieved. Through the combination of data augmentation technology and the dense convolutional network, the shooting efficiency can be quickly and accurately evaluated, meeting the requirements of real - time damage effectiveness evaluation in modern combat training.

[0140] 7. Overfitting is reduced and training is optimized. The switching module effectively reduces the network parameters, avoids overfitting, and accelerates the training speed of the model.

[0141] 8. Through the combination of the dense convolutional network and the butterfly optimization algorithm, the shooting efficiency can be evaluated in real - time and accurately, providing decision - making support for precise strikes in modern warfare. The hyperparameters of the dense convolutional network are optimized by using the butterfly optimization algorithm. By simulating the search behavior of butterflies when looking for food, the optimal combination of hyperparameters is found, thus avoiding the limitation of being easily trapped in local optimal solutions in traditional methods. The result of hyperparameter optimization is that the model can converge faster and avoid overfitting during the training process, thereby improving the accuracy of shooting efficiency evaluation.

[0142] 9. The key advantages of the dense convolutional network lie in its strong feature extraction ability and good gradient propagation characteristics, which can quickly and efficiently evaluate the shooting efficiency.

[0143] The embodiments of the present disclosure also provide a data processing system for shooting efficiency evaluation for implementing the above method embodiments, as Figure 5 shown. The data processing system 10 for shooting efficiency evaluation includes a model screening unit 101, a data acquisition unit 102, and a data processing unit 103.

[0144] The model screening unit 101 is configured to obtain the model data of the weapon to be evaluated and determine a shooting efficiency estimation model that matches the model data. Among them, the shooting efficiency estimation model is configured with a dense convolutional network that matches the model data.

[0145] The data acquisition unit 102 is configured to obtain the vulnerability parameters of the shooting target, the intersection data of the ammunition of the weapon to be evaluated and the shooting target, and the shooting data of the operator using the weapon to be evaluated in this shooting.

[0146] The data processing unit 103 is configured to input the shooting data, vulnerability parameters, and intersection data into the pre-trained shooting efficiency estimation model respectively to obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting. Among them, the shooting efficiency estimation model is configured to perform multi-feature extraction on the shooting data based on the vulnerability parameters and intersection data through a dense convolutional network, and obtain the shooting efficiency result data according to the extracted multi-features.

[0147] The specific manners of the operations performed by each unit in the above device embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0148] The embodiments of the present disclosure also provide an electronic device, as Figure 6 shown. The electronic device includes one or more processors 21 and a memory 22, Figure 6 taking one processor 21 as an example.

[0149] The controller may further include: an input device 23 and an output device 24.

[0150] The processor 21, the memory 22, the input device 23, and the output device 24 may be connected through a bus or other means, Figure 6 taking connection through a bus as an example.

[0151] The processor 21 may be a Central Processing Unit (CPU), or it may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., or a combination of the above types of chips. The general-purpose processor may be a microprocessor or any conventional processor.

[0152] The memory 22, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the control method in the embodiments of the present disclosure. The processor 21 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 22, that is, implements the data processing method for shooting efficiency evaluation in the above method embodiments.

[0153] The memory 22 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the processing device of the server operation, etc. In addition, the memory 22 may include high-speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory 22 may optionally include a memory remotely set relative to the processor 21, and these remote memories can be connected to the network connection device through a network. Examples of the above networks include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0154] The input device 23 can receive input digital or character information and generate key signal inputs related to the user settings and function controls of the processing device of the server. The output device 24 may include display devices such as a display screen.

[0155] One or more modules are stored in the memory 22 and, when executed by one or more processors 21, execute the method as Figure 1 shown.

[0156] Those skilled in the art can understand that to implement all or part of the processes in the above method embodiments, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes in the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (FM), a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memories.

[0157] Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A data processing method for shooting efficiency evaluation, characterized in that: include: Acquire model data of the weapon to be evaluated, and determine a shooting efficiency estimation model that matches the model data, wherein the shooting efficiency estimation model is configured with a dense convolutional network that matches the model data; Acquiring vulnerability parameters of a shooting target, intersection data between the ammunition of the weapon to be evaluated and the shooting target, and shooting data of an operator using the weapon to be evaluated in this shooting; and, The shooting data, the vulnerability parameters, and the intersection data are respectively input into the pre-trained shooting efficiency estimation model to obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting, wherein the shooting efficiency estimation model is configured to perform multiple feature extraction on the shooting data based on the vulnerability parameters and the intersection data through the dense convolutional network, and obtain the shooting efficiency result data based on the extracted multiple features.

2. The method according to claim 1, characterized in that The method further comprises: For different types of weapons to be evaluated, independent shooting efficiency estimation models are established respectively; For each of the shooting efficiency estimation models, the shooting efficiency estimation model is trained by using training data samples of the matching model of the weapon to be evaluated until the output of the trained shooting efficiency estimation model reaches a set accuracy, thereby obtaining the pre-trained shooting efficiency estimation model.

3. The method according to claim 2, characterized in that The shooting efficiency estimation model includes the dense convolution network and the feature fusion module which are arranged in sequence; and the shooting data, the vulnerability parameter, and the intersection data are respectively input into the pre-trained shooting efficiency estimation model to obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting, including: In response to the received shooting efficiency evaluation instruction, the dense convolution network is started to perform shallow feature extraction and deep feature extraction on the shooting data based on the vulnerability parameter and the intersection data, respectively, to obtain a shallow feature time series and a deep feature time series containing degradation trend information of the shooting data; The feature fusion module is started, and the shallow feature time series and the deep feature time series are feature-fused through the fully connected layer in the feature fusion module to obtain the shooting efficiency result data.

4. The method according to claim 3, characterized in that Each convolution layer in the dense convolution network adopts the same configuration, wherein the first convolution layer in the dense convolution network is used as an input convolution layer for shallow feature extraction, and the other convolution layers except the first convolution layer are stacked as a dense block for deep feature extraction, and each of the convolution layers in the dense block is configured to perform a zero padding operation on the convolution result after each convolution; and The shooting efficiency estimation model also includes a switching module; the switching module is configured to perform feature fusion on the multiple deep feature time series output by the dense block, and send the new time series after feature fusion to the feature fusion module to replace the deep feature time series for feature fusion.

5. The method according to any one of claims 1 to 4, characterized in that: For each of the shooting efficiency estimation models, obtaining training data samples of the weapon to be evaluated of a matching model to train the shooting efficiency estimation model includes: Performing data enhancement processing on the training data samples to generate virtual data samples having a consistent degradation trend with the training data samples; and The virtual data samples are mixed with the training data samples, and the shooting efficiency estimation model is trained by the mixed data samples.

6. The method according to claim 5, characterized in that The performing data enhancement processing on the training data sample to generate a virtual data sample having a consistent degradation trend with the training data sample includes: Obtaining the similarity between any two time series in the training data samples; Grouping the entire training data samples based on the similarity, so that the similarity between any two time series in each grouped sample is higher than a set threshold; and, A virtual group is generated based on each group sample, and a virtual data sample is obtained after integration, so that the virtual data sample and the training data sample have a consistent degradation trend.

7. The method according to claim 6, characterized in that The training data samples contain time series with different data lengths, and any two time series in the training data samples are j and O k The similarity Dis DTW is configured as: Dis DTW =sqrt(dp) in, d(O j ,O k )=(O j -O k ) 2 Wherein, dp is a matrix, the minimum value of m is 2, and the maximum value is O j The minimum value of n is 2, and the maximum value is the data length of Ok, d(O j ,O k ) is the O j and the square of the Euclidean distance between dp and Ok, dp[m,n] is the element in the mth row and nth column of dp, dp[m-1,n] is the element in the m-1th row and nth column of dp, dp[m,n-1] is the element in the mth row and n-1th column of dp, and dp[m-1,n-1] is the element in the m-1th row and n-1th column of dp.

8. The method according to claim 6, characterized in that The training data samples contain time series with the same data length, and any two time series in the training data samples are j and O k The similarity Dis DTW is configured as: In the formula, O j (t), O k (t) respectively represent the time series O j and O k The value at time point t, TC represents the time series O j , O k The data length.

9. The method according to any one of claims 6 to 8, characterized in that: The intersection data includes the relative motion trajectory and posture of the ammunition and the shooting target; and, The input of the shooting efficiency estimation model also includes the destructive power parameters of the weapon to be evaluated and the climatic conditions; The shooting efficiency estimation model is trained by a butterfly optimization algorithm to determine the optimal hyperparameter combination in the shooting efficiency estimation model.

10. A data processing system for shooting efficiency evaluation, characterized in that: include: A model screening unit is configured to obtain model data of a weapon to be evaluated and determine a shooting efficiency estimation model that matches the model data, wherein the shooting efficiency estimation model is configured with a dense convolutional network that matches the model data; a data acquisition unit configured to acquire vulnerability parameters of a shooting target, intersection data between the ammunition of the weapon to be evaluated and the shooting target, and shooting data of an operator using the weapon to be evaluated in this shooting; and The data processing unit is configured to input the shooting data, the vulnerability parameter, and the intersection data into the pre-trained shooting efficiency estimation model, respectively, to obtain the shooting efficiency result data of the operator using the weapon to be evaluated in this shooting, wherein the shooting efficiency estimation model is configured to perform multiple feature extraction on the shooting data based on the vulnerability parameter and the intersection data through the dense convolutional network, and obtain the shooting efficiency result data based on the extracted multiple features.

Images