Intelligent shooting behavior prediction and optimization method based on deep learning

Abstract

The invention relates to the technical field of behavior prediction, in particular to an intelligent shooting behavior prediction and optimization method based on deep learning, which comprises the following steps: acquiring stretching data to analyze a displacement trend, identifying synclastic enhancement and sudden change synchronization, acquiring aiming speed, grip strength, recoil and respiration data to analyze collaborative conflicts, and extracting angle change to classify an aiming trend. Integrating the three types of labels to judge behavior deviation, and matching prompt content to output a prediction result. According to the method, the stretching trend change of the sensing area is recognized, the structure change state sensing is enhanced, the aiming speed, the holding force, the recoil response and the dynamic sequence of the respiratory rhythm are integrated, the cooperative relation of behavior driving factors is mined, the aiming angle deviation frequency is extracted, the behavior trend characteristics are recognized, and transverse comparison is conducted between multiple types of labels; the behavior deviation direction and the trend affiliation are determined and matched to the prompt content and then are output in a classified mode, a whole-process closed loop from multi-source data analysis to prediction guidance is achieved, and prediction accuracy, real-time performance and tactical adaptation degree are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of behavior prediction, and particularly to an intelligent shooting behavior prediction and optimization method based on deep learning. Background Art

[0002] The technical field of behavior prediction includes related methods and technologies for inferring the future behaviors of individuals or groups based on data analysis and pattern recognition. The core content of this technical field is to use artificial intelligence algorithms such as machine learning and deep learning to extract features from historical behavior data and achieve probabilistic prediction of future behaviors through modeling. Behavior prediction technology is mainly applied in scenarios such as intelligent security, traffic prediction, human-computer interaction, military simulation, and intelligent decision-making. Through continuous learning of multi-dimensional dynamic data, this field realizes accurate modeling of behavior patterns and combines perception data, time series analysis, and probabilistic inference to complete the prediction of the changing trends of target behaviors. This field features data-driven, model self-adaptation, and real-time inference, and its development relies on the continuous improvement of large-scale data acquisition capabilities, computing resource support, and algorithm optimization capabilities.

[0003] Among them, the intelligent shooting behavior prediction and optimization method based on deep learning refers to using a deep neural network structure to extract features from multi-source dynamic data during the shooting behavior process, predicting behavior parameters such as the spatial trajectory, posture changes, and firing intentions of shooting actions by constructing a time series correlation model, and simultaneously modeling the evolution trends of target recognition, target tracking, and shooting intentions in combination with historical shooting data. Specifically, this method relies on a convolutional neural network to extract visual information, processes time series data through a recurrent neural network, and jointly uses a reinforcement learning algorithm to continuously optimize the shooting strategy. In addition, this method also trains a high-dimensional action prediction model based on large-scale shooting behavior data to infer the state transition of individual behaviors during shooting and correct the prediction results in real time.

[0004] Most current technical approaches focus on probabilistic modeling based on static behavior data, lacking the ability to finely distinguish the evolution of microstructure states over time, and making it difficult to effectively capture the linkage change characteristics between multiple regions within the launcher. In the process of shooting behavior prediction, conventional schemes often make judgments based on single variables or low-dimensional combinations, ignoring the co-evolution mechanism of multiple behavior variables in the time series dimension, resulting in logical breaks when analyzing the causes of complex behaviors. The frequency of directional switching of aiming behavior has not been systematically modeled, which limits the in-depth identification of fluctuation patterns in behavior, resulting in trend misjudgment or lagging stability assessment. Most of the prediction results are static outputs, lacking the linkage structure of behavior prompt content, making it difficult for output information to connect operational decisions, and easily causing idling between identification content and action instructions. Faced with a tactical environment that requires high-speed feedback and real-time response, this type of technical model is difficult to ensure the continuity of prediction logic and the operational value of behavior identification, seriously affecting the application efficiency of the overall intelligent feedback system. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art and to propose an intelligent shooting behavior prediction and optimization method based on deep learning.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an intelligent shooting behavior prediction and optimization method based on deep learning, comprising the following steps: S1: Obtain the tensile data of the emission component area, analyze the change direction of three consecutive sets of data, compare the displacement trends of adjacent areas, determine whether there is unidirectional enhancement and mutation synchronization, and obtain the structural change performance; S2: Collect continuous data of aiming speed, grip strength, recoil response, and breathing rhythm, and after synchronous processing, combine aiming and grip strength trends, identify sudden recoil changes and compare breathing status, analyze the coordination and conflict between the four items, and obtain the performance of action impact; S3: Obtain the time series of aiming angles, calculate the angle difference and generate a direction change sequence, count the conversion frequency, classify it into one-way deviation, continuous oscillation and stabilization according to the distribution, and generate an aiming trend label; S4: Integrate the three types of labels, namely, the structural change performance, the action impact performance and the generated aiming trend label, and compare them horizontally to see if they are consistent. If they are of the same type, it is determined as a target behavior bias, and a shooting behavior bias label is generated; S5: According to the preset trend prompt table of the shooting behavior bias label, select the corresponding prompt content, correct the biased behavior, and classify and output the shooting behavior prediction and optimization results.

[0007] As a further solution of the present invention, the structural change manifestations include the stretching change direction type, the adjacent area displacement comparison mode, and the spatial distribution chain characteristics. The action influence manifestations include the trend combination type of aiming speed and holding force, the mutation characteristic points in the recoil response, and the breathing rhythm state label. The aiming trend labels include the single-direction offset trend, the continuous oscillation trend, and the slow stabilization trend. The shooting behavior deviation labels include the direction deviation classification, the perturbation performance classification, and the stabilization state classification. The shooting behavior prediction and optimization results include the trend category prompt items, the label selection range, and the classification output content; The trend prompt table refers to the corresponding relationship table of the preset trend categories and the corresponding prompt contents, which is divided into three text sets of deviation, perturbation, and stabilization according to the shooting behavior classification results; The construction method of the trend prompt table is as follows: based on the statistical analysis of shooting behavior data, the shooting behavior deviation labels are attributed and matched with the actual performance, the typical deviation types and correction suggestions are divided, and the mapping relationship from the labels to the prompt contents is established.

[0008] As a further solution of the present invention, the specific steps of S1 are as follows: S101: Obtain the stretching change value of the sensing area in the launching member. Based on the continuous three groups of collected data at each point, calculate the stretching direction angle change amplitude, summarize the point direction change data, and generate the stretching direction change average amplitude; S102: According to the stretching direction change average amplitude, compare the direction change values of adjacent sensing areas, screen the area combinations with the direction difference lower than the direction change threshold, and generate the same-direction trend ratio of adjacent areas; S103: Based on the same-direction trend ratio of adjacent areas, extract the points with the same direction and the stretching mutation value higher than the stretching mutation threshold, statistically analyze the spatial distribution and summarize and judge the overall change relationship to obtain the structural change manifestation; The direction change threshold refers to the maximum range of the stretching direction angle difference between adjacent areas, which is set as the angle deviation standard; The stretching mutation threshold refers to the numerical limit for determining that the stretching change amplitude of a point reaches the mutation standard during the continuous acquisition period, which is defined as the standard deviation multiple range of the stretching change values of all points in the area; The specific value of the standard deviation multiple range is statistically analyzed through historical stretching data, and the standard deviation multiple corresponding to the stretching mutation threshold is set to cover the normal fluctuation range and identify abnormal mutations.

[0009] As a further solution of the present invention, the specific formula for calculating the stretching direction angle change amplitude is as follows: ; Where: represents the The angular change amplitude of the stretching direction at a point in the time period, , respectively represent the stretching change amounts in the X and Y directions of the point at the time period, , are the stretching change amounts in the corresponding directions in the previous time period, is the balance factor, is the key weight factor of the point; The dimensional normalization process of the parameters can be carried out by dividing the physical quantity, displacement change amount and weight by the characteristic scale and the maximum value respectively for standardization, verifying the dimensionless input in the formula for differentiating dimensional units; The characteristic scale is obtained by extracting the mean and median of the displacement change amounts in the historical samples, and the maximum value is taken as the absolute maximum value of the parameter in the sample set.

[0010] As a further solution of the present invention, the specific steps of S2 are as follows: S201: Obtain continuous data sequences of aiming speed, holding force, recoil response, and breathing rhythm, perform time-axis alignment processing on the four items of data according to a unified time reference, merge the four numerical values at each time node into a unified data unit, call the data unit for index arrangement, and obtain a synchronous action sequence group; S202: According to the synchronous action sequence group, extract the change direction of the aiming speed and the increasing and decreasing trend of the holding force in the time node, perform paired classification based on the change relationship between the two, and mark the combinations with increasing holding corresponding to accelerating changes in the paired data to obtain an action coordination matching record; S203: Call the time period numbers of the marked combinations in the action coordination matching record, correspondingly extract the change break points of the recoil response and match the state of the breathing rhythm, make a determination according to the consistency between the break point fluctuation amplitude and the breathing state, and classify and file all combinations for conflict and coordination to obtain an action influence performance; The time-axis alignment and dimensional normalization of the four items of data are as follows: linearly standardize the numerical values according to the range in the overall acquisition period, so that the numerical values are uniformly mapped to the interval.

[0011] As a further solution of the present invention, the specific steps of S3 are as follows: S301: Obtain the aiming angle offset values in a continuous time period, calculate the angle change density values in adjacent time periods in chronological order, arrange the positive and negative change trends in sequence, and call the direction change identifiers generated in all time periods to construct a sequence index to obtain a direction change sequence; S302: According to the direction change sequence, count the number of executions at the positions where the direction positive and negative conversions occur in the sequence. Denote the segments with consistent directions continuously as zero conversions, include the conversion points in the change segments in the total frequency, classify and screen the frequencies in all time periods, and obtain the direction conversion frequency distribution; S303: Invoke the direction conversion frequency distribution, and based on whether the frequency of each segment exceeds the direction frequency definition threshold and combined with the change duration, determine three situations: single offset, continuous oscillation, and slow stabilization. Classify the annotations according to the determination results to obtain the aiming trend labels; The direction frequency definition threshold refers to the lower limit of the number of conversions for determining the type of direction change trend, and is set as the median of the number of positive and negative direction conversions within a unit time period.

[0012] As a further solution of the present invention, the calculation formula for the angle change density value within adjacent time periods is specifically: ; Among them, represents the angle change density value within the th adjacent time period, represents the aiming angle offset value at the time point , represents the aiming angle offset value at the time point , and respectively represent the inertial direction offset projection amounts at the time points and , represents the disturbance gain coefficient output by the device attitude estimation module within the th time period, represents the constant damping adjustment parameter; The dimensional normalization process of the parameters is carried out by introducing characteristic quantities, the reference value of the angle offset, the calibration scale of the inertial projection, the maximum amplitude of the disturbance gain, and the dimensionless factor of the damping parameter, and standardizing the physical quantities into dimensionless forms.

[0013] As a further solution of the present invention, the specific steps of S4 are: S401: Invoke the three contents of the structural change performance, the action influence performance, and the aiming trend label, collate the labels in the time periods side by side according to the same acquisition period, assign numbers to the three label contents respectively and align them, and construct a horizontal comparison matrix according to the acquisition time nodes to generate a label corresponding sequence group; S402: According to the corresponding sequence group of the tags, compare the performance types of the three tags in each group, determine whether there are simultaneously deviated, disturbed, and stabilized tags of the same type, record the cases with consistent type performance in a statistical table, calculate the same-direction determination value of the tags, generate a classification number, and obtain the same-direction classification ratio of the tags. S403: Call the same-direction classification ratio of the tags, screen the time period numbers in which the majority of the tags belong to the same type, classify them into the deviated, disturbed, and stabilized trends respectively, label and integrate the corresponding types of all time periods, and obtain the shooting behavior deviation tags.

[0014] As a further solution of the present invention, the calculation formula of the same-direction determination value of the tags is specifically: ; where represents the same-direction determination value of the mth group of tags, represents the dispersion difference of the deviated type tags in the mth group, is the terrain penetration weight of the disturbed type tags in the mth group, represents the benchmark quantity of the sample group, refers to the balance factor of the jth stabilized tag in the mth group, characterizes the curvature change amount of the jth tag in the mth group.

[0015] As a further solution of the present invention, the specific steps of S5 are as follows: S501: Call the attribution trend category in the shooting behavior deviation tags, correspondingly match it to a pre-defined trend status prompt content table, extract the index number set of the prompt content corresponding to each type of trend, screen the associated content range according to the tag type, and generate a trend prompt index mapping table; S502: According to the trend prompt index mapping table, sequentially extract the prompt item texts in the prompt content table that match the tag numbers in the acquisition cycle sequence, rearrange the prompt items according to the tag classification order, construct a structured prompt set, and obtain the tag pointing prompt group; S503: Call the tag pointing prompt group, merge and sort the prompt items according to the trend type, output the classification content in the order of the three categories of deviation, disturbance, and stabilization, generate a pre-judgment and solution text set under the trend status, and obtain the shooting behavior prediction and optimization result; The screening of the associated content range by the tag type refers to limiting the call to all the prompt item sets that match the tag category in the prompt content table according to the status classification corresponding to each type of trend tag; The prompt item text that matches the tag number refers to the set of text descriptions set correspondingly in the prompt content table according to the trend category number, which is preset manually and indexed and associated according to the trend type.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by identifying the stretching trend change of the sensing area, enhancing the perception of the structural change state, integrating the dynamic sequences of aiming speed, grasping force, recoil response and breathing rhythm, exploring the synergistic relationship of behavior driving factors, extracting the aiming angle deviation frequency, identifying the behavior trend characteristics, making horizontal comparison among multiple types of labels, clarifying the behavior deviation direction, and matching the trend attribution to the prompt content and then classifying and outputting, a full-process closed loop from multi-source data analysis to prediction guidance is formed, improving the accuracy, real-time performance and tactical adaptability of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flow chart of the steps of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following will describe the technical solutions in the present invention with reference to the drawings.

[0020] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or more advantageous than other embodiments or design solutions. Exactly speaking, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0021] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0022] In the embodiments of the present invention, sometimes the subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0023] In order to make the technical problems to be solved, technical solutions and advantages of the present invention clearer, the following will be described in detail with reference to the drawings and specific embodiments.

[0024] Please refer to Figure 1 , an intelligent shooting behavior prediction and optimization method based on deep learning, comprising the following steps: S1: Obtain the stretching change values of multiple sensing regions within the launching member, calculate the stretching direction change based on three consecutive acquisition periods for each point, perform mutual comparison on the displacement directions of adjacent regions, and determine whether there are manifestations of co-directional enhancement and mutation synchronization in the trends. Judge whether the overall region is in a state of change linkage through spatial distribution summary to obtain the structural change performance; S2: Obtain the continuous data sequences of aiming speed, holding force, recoil response, and breathing rhythm, synchronize and organize them according to a unified time reference, call the combination classification of the aiming speed change direction and the holding force increase and decrease trend, identify the change break points in the recoil data and compare the states of the breathing rhythm, analyze the synergistic and conflicting relationship structures among the above four items and perform behavior classification to obtain the action influence performance; S3: Obtain the aiming angle offset values within a continuous time period, extract the angle differences between each time point and generate a direction change sequence, count the frequency of direction conversion in the sequence, and classify it into a single direction offset, continuous oscillation, or slow stabilization trend according to the distribution difference of the conversion frequency within the continuous segment to obtain the generated aiming trend label; S4: Call the structural change performance, action influence performance, and generated aiming trend label, perform horizontal comparison on the types of the three classification labels within the same acquisition cycle, and determine whether there are manifestations of deviation, perturbation, and stabilization in the same direction. If most labels belong to the same category, classify them into the corresponding trend direction to obtain the shooting behavior bias label; S5: Call the attribution trend category in the shooting behavior bias label, correspondingly match it to the pre-defined trend state prompt content table, select the prompt content range through the label identifier, and perform classification output to obtain the shooting behavior prediction and optimization result.

[0025] The structural change performance includes the stretching change direction type, the adjacent region displacement comparison mode, and the spatial distribution linkage characteristics. The action influence performance includes the trend combination type of aiming speed and holding force, the mutation characteristic points in the recoil response, and the breathing rhythm state label. The aiming trend label includes a single direction offset trend, a continuous oscillation trend, and a slow stabilization trend. The shooting behavior bias label includes a deviation classification, a perturbation performance classification, and a stabilization state classification. The shooting behavior prediction and optimization result includes a trend category prompt item, a label selection range, and classification output content.

[0026] The specific steps of S1 are as follows: S101: Obtain the stretching change values of the sensing regions within the launching member, calculate the stretching direction angle change amplitude based on three consecutive sets of acquisition data for each point, summarize the point direction change data, and generate the stretching direction change mean amplitude; The calculation formula for the amplitude of the change in the stretching direction angle is specifically as follows: ; Where: represents the amplitude of the change in the stretching direction angle of the th point at the time period, , respectively represent the stretching change amounts of the th point in the time period along the X - direction and the Y - direction, , are the stretching change amounts in the corresponding directions of the previous time period, is the balance factor, is the weight factor of the importance of the point; The angle value can reflect the actual deflection trend of the stretching direction through numerical mapping or angle conversion, and is used for subsequent displacement anomaly detection and trend analysis; Basis for parameter acquisition and numerical setting: and : The X - and Y - direction displacements of the point at the time period are monitored by a high - precision GNSS device, with the unit of millimeters (mm). For example, in a certain monitoring, the X - direction displacement of the point at the time period is 2.5 mm, and the Y - direction displacement is 1.8 mm.

[0027] and : The X - and Y - direction displacements of the point in the previous time period are also obtained by the GNSS device, with the unit of millimeters (mm). For example, the X - direction displacement in the previous time period is 2.0 mm, and the Y - direction displacement is 1.5 mm.

[0028] : The balance factor introduced to avoid a zero denominator. According to the minimum displacement change amount of the actual monitoring data, set mm to ensure the stability of the calculation.

[0029] : Represents the weight factor of the importance of the point . According to factors such as the geological sensitivity of the area where the point is located and the historical deformation records, the weight value is assigned through expert evaluation and historical data analysis. For example, if the point is located in a high - risk landslide area, set ; if it is located in a stable area, set .

[0030] Derivation process of formula calculation: Calculate the numerator part: ; Calculate the denominator part: ; Calculate the ratio and multiply by the weight factor: ; Result interpretation: The calculated represents the point position in the time period The change amplitude of the stretching direction is 0.2264. This value is used to measure the significant change in the stretching direction between the current time period and the previous time period. The larger the value, the more significant the direction change, which may indicate a change in the structure or geological conditions.

[0031] Subsequent applications: By summarizing and analyzing the values of all point positions, the average amplitude of the stretching direction change can be generated, which is used to monitor the overall deformation trend of the area and assist in judging the potential geological hazard risk.

[0032] S102: According to the average amplitude of the stretching direction change, compare the direction change values of adjacent sensing areas, screen out the area combinations with the direction difference lower than the direction change threshold, and generate the same-direction trend ratio of adjacent areas; When comparing the direction change values of adjacent sensing regions according to the mean amplitude of the stretching direction change, it is necessary to first establish a method for determining the adjacent relationship between regions. Generally, it can be judged by the geometric distance between the center points of each sensing region. A spatial distance threshold is set as the adjacent recognition standard. If the distance between the center points of two regions is less than this value, they are judged as adjacent regions. This distance threshold can be set according to the geometric characteristics of the component. For example, for a cylindrical component with a diameter of 300 mm, if the division scale of each sensing region is 50 mm, the distance threshold can be set to 60 mm. After confirming the adjacent region pairs, the direction change amplitude values in these two regions are extracted respectively, and a direct difference operation is performed on them. For example, if the direction change of region X is 6.2 degrees and that of region Y is 10.5 degrees, the difference between the two is 4.3 degrees. If this difference is lower than the preset direction change threshold, it is determined that there is a same-direction trend between these two regions. The direction change threshold can be obtained based on the statistical change of historical structure data. Usually, the mean value of the differences between adjacent regions in multiple experimental samples can be selected as the threshold reference. For example, in 20 component samples, the direction differences between adjacent regions mostly concentrate between 5 and 7 degrees, so the direction change threshold can be set to 7 degrees. All adjacent region pairs are compared one by one in this way, the number of combinations with direction differences lower than the threshold is counted, and then divided by the total number of adjacent combinations to calculate the same-direction trend ratio of adjacent regions, which is used as a quantitative index for judging the overall coordination of the subsequent structure. If this ratio is higher than 0.6, it can be preliminarily considered that the overall trend of the component regions is consistent; otherwise, further analysis of the distribution of abnormal combination regions is required.

[0033] S103: Based on the same-direction trend ratio of adjacent regions, extract the points where the directions are the same and the stretching mutation value is higher than the stretching mutation threshold, statistically analyze the spatial distribution and summarize to judge the overall change relationship, and obtain the structural change performance; When extracting points with the same direction and a stretching mutation value higher than the threshold based on the ratio of the same-direction trends in adjacent regions, it is necessary to screen out those points with a small difference between the direction change value and the average direction change value of the region on the basis of the original region division and point data. The determination method for the same direction is that the difference between the point direction change value and the average direction change value of the region does not exceed the set direction consistency difference standard. This standard can be set according to the component material properties and experimental data. Usually, a value within 3 degrees is taken as the judgment boundary. Further, calculate the stretching mutation value among the points that meet the condition of the same direction, that is, the difference between the stretching value at the current time point and the average stretching value in the previous time period. If the difference is greater than the preset stretching mutation threshold, it is determined as a stretching mutation point. The setting method of this threshold is generally based on the statistical range of stretching changes under the normal working state of the component. Take 80% of its maximum fluctuation amplitude as the threshold basis. For example, through long-term monitoring of a certain component, it is found that the maximum stretching fluctuation between points generally does not exceed 0.06 mm, then the stretching mutation threshold can be set to 0.05 mm. After identifying the mutation points, extract their coordinate information and count the spatial distribution characteristics of these points on the component. If most mutation points are concentrated in a certain structural plane or boundary region, the relative distance and arrangement direction can be further analyzed to judge whether there is a trend change direction. For example, if most points are arranged obliquely in the upper left region of the component, it indicates that there is a structural direction consistency mutation phenomenon in this region. Through the combination of the positions and stretching data of these points, the overall change performance of the structure can be inferred.

[0034] The specific steps of S2 are as follows: S201: Obtain continuous data sequences of aiming speed, holding force, recoil response, and breathing rhythm. Perform time-axis alignment processing on the four items of data according to a unified time reference. Combine and merge the four values at each time node into a unified data unit. Call the data unit for index arrangement to obtain a synchronous action sequence group; When obtaining continuous data sequences of aiming speed, holding force, recoil response, and breathing rhythm, various signal data should be collected from different types of sensor channels. Among them, the aiming speed is recorded by a visual tracking device to record the moving rate of the sight, and the speed value is calculated from the displacement between frames and the time interval, with the unit of ° / s. The holding force is recorded in real time by a hand pressure sensor, with the unit of N. The recoil response is collected by an accelerometer to obtain the acceleration value generated by the instantaneous recoil of the component, with the unit of m / s². The breathing rhythm is collected by a chest sensing device to collect the longitudinal displacement curve for identifying the timing signals of inhalation and exhalation. All data is recorded with timestamps, and the sampling frequency is uniformly set to 100 Hz per second. When aligning the data of each channel according to a unified time reference, it is necessary to interpolate and fill the sampling time axis so that the four types of data have corresponding values at the same moment. For example, at 0.01 seconds, the aiming speed is 6.5 ° / s, the holding force is 28 N, the recoil acceleration is 0.8 m / s², and the breathing displacement is 3.2 mm. That is, the four values are integrated into a data unit, and this assembly action is repeated to generate a complete synchronous data queue for all time points. Each data unit is arranged with a time tag as the index field, and a synchronous action sequence group index table is established. The data unit numbers are arranged in ascending order of time to form a stable index mapping sequence. In actual operation, collecting data for a period of 10 seconds will generate 1000 synchronous data units, and the synchronous action sequence group for the entire time period is completed through data structure construction.

[0035] S202: According to the synchronous action sequence group, extract the change direction of the aiming speed and the increasing or decreasing trend of the holding force in the time node, and classify them in pairs based on the relationship between their changes. Number and mark the combinations of increasing holding force corresponding to accelerating changes in the paired data to obtain an action coordination matching record; When extracting the change direction of the aiming speed and the increasing or decreasing trend of the holding force in the time nodes according to the synchronous action sequence group, it is necessary to read the aiming speed values and the holding force values of two consecutive data units in chronological order, perform the difference operation on the aiming speed between the previous and the subsequent values. If the speed value at the later moment is higher than that at the previous moment, it is determined to be accelerating; otherwise, it is decelerating. Similarly, the same difference calculation is performed on the holding force to judge the force change trend. When performing this process, it should be iterated through each time point. For example, if the speed at the 100th frame is 5.6° / s and the speed at the 101st frame is 7.1° / s, it is accelerating. If the holding force changes from 25N to 28N, it is increasing. If the directions of both are the same, the data pair is classified into the paired category. For the combinations in the paired category that meet the conditions of increasing holding force and accelerating aiming speed, numbering is performed. The numbering is based on the time tag with an additional increasing index. For example, the data point numbers that meet the conditions are T101, T203, T307, etc. For each group of combinations, record its time tag, holding value, speed value, and change amplitude. The combined data forms an action pairing mark table. In this process, it is necessary to clarify the judgment interval for each change. If the difference in the aiming speed between two time nodes is less than 0.3° / s, no acceleration determination is made. If the change in the holding force is less than 2N, it is not considered a significant increase or decrease. All data pairs that meet the above criteria and have a consistent direction relationship are summarized into the matching record group, and finally, the action coordination matching record is obtained.

[0036] S203: Call the time period number of the marked combination in the action coordination matching record, extract the change break points of the recoil response correspondingly and match the state of the respiratory rhythm. Make a determination based on the consistency between the break point fluctuation amplitude and the respiratory state, and classify and file all combinations for conflict and coordination to obtain the action influence performance; When calling the time period number of the marked combination in the action coordination matching record, it is necessary to extract the corresponding data window according to the number index. Usually, the time period of 0.2 seconds before and after this number is taken for extraction. Check the recoil response value in this section of data to determine whether there is a sudden point phenomenon. The sudden point is defined as an instantaneous jump in acceleration with an amplitude exceeding the set mutation reference value. This reference value is determined based on historical sampling results. For example, if the average acceleration change amplitude of 1000 sections of stable samples is calculated to be 0.5 m / s², then the mutation threshold is set to 0.8 m / s². If the instantaneous acceleration rises to 1.1 m / s² in a certain numbered section, it is regarded as a recoil sudden point. After confirming the sudden point, it is necessary to synchronously read the respiratory rhythm data in the corresponding time period, analyze whether this time period is inhalation, exhalation or transition state, and judge the consistency of the respiratory state by determining whether the time point of the sudden point is located in the peak or trough area of the respiratory rhythm wave. If the sudden point occurs at the respiratory wave trough, it is marked as the consistent state, otherwise it is marked as the conflict state. After executing this process for each group of data, record the number, the sudden point amplitude value and the respiratory state flag. Finally, classify all the marked combinations according to the consistency of the respiratory state, construct the data sets of the conflict group and the coordination group, and form the action influence performance data archiving table.

[0037] The specific steps of S3 are as follows: S301: Obtain the aiming angle offset value within a continuous time period. Calculate the angle change density value within adjacent time periods in chronological order, arrange the positive and negative change trends in sequence, and call the direction change identifiers generated in all time periods to construct a sequence index to obtain the direction change sequence; The specific calculation formula for the angle change density value within adjacent time periods is: ; Among them, represents the angle change density value within the th adjacent time period, represents the aiming angle offset value at the time point , represents the aiming angle offset value at the time point , and respectively represent the inertial direction offset projection amounts at the time points and , represents the disturbance gain coefficient output by the device attitude estimation module within the th time period, represents the constant damping adjustment parameter; The parameter setting and source description are as follows: The aiming angle offset value , Obtained by the laser angle measurement sensor in the visual aiming system. The device model is SICK LMS111 series, with a measurement accuracy of ±0.25°, a sampling frequency of 50Hz. During the inspection of a certain device in May 2025, the angle offset values were detected at two consecutive time points as 、

[0038] Inertial direction offset projection amount 、 Obtained by the IMU six-axis acceleration sensor (model: Bosch BMI270). After inertial projection transformation, the results are respectively 、 , and the quantization adopts the vertical projection algorithm Disturbance gain coefficient Calculated from the dynamic normalization value of the Kalman filter residual variance in the attitude estimation module. The setting rule is: when the residual variance is in the range of 0.02 - 0.2, the coefficient is linearly mapped to the interval [0.1, 1.0]. The current residual variance is 0.08, and the corresponding mapped value is

[0039] Damping adjustment parameter The setting rule is that the system noise variance interval [0.001 - 0.1] is mapped to [0.1 - 0.5]. The current noise variance is 0.008, and the corresponding damping adjustment parameter is

[0040] Substitute all the above parameters into the formula: ; The step-by-step operations are as follows: Difference square term: ; Sum product term: ; Numerator calculation: ; Denominator calculation: ; Overall division: ; Absolute value operation: ; The calculation result is: ; The result shows that the angular change density within a unit time step in the current time period is 4.6435, indicating that the attitude deflection state of the device is in the non-steady change region during this time period. The density value is much higher than the set upper threshold of 3.2, indicating a strong state of unstable target angle. This numerical result is used as a key input quantity for constructing the "direction change identifier" in Step 3. Subsequently, a complete direction change sequence index model will be formed based on the change trend index of each time step. This numerical value will be passed to the direction trend map construction unit in the subsequent steps and used as a clustering weight factor and a reference parameter for the time axis perturbation value.

[0041] S302: According to the direction change sequence, count the number of times of positive and negative direction conversion positions in the sequence. Denote the continuous segments with the same direction as zero conversions, and count the conversion points in the changing segments into the total frequency. Classify and screen the frequencies under all time periods to obtain the direction conversion frequency distribution. When counting the number of times of positive and negative direction conversion positions in the sequence according to the direction change sequence, it is necessary to traverse all symbol values in the direction sequence in the order of time periods. Starting from the second element, compare it with the previous element one by one to determine whether there is a positive and negative sign change. If the signs of two adjacent elements are different, record a direction conversion. If they are the same, it is considered that the direction remains unchanged. Further, consider all time periods with continuous same direction as a paragraph, and record the number of direction conversions in this paragraph as zero. For example, in the sequence "++--+-", there is one conversion between "++" and "--", and the cumulative number of direction conversions is three. Classify the number of direction conversions according to the time periods to which each paragraph belongs, and establish a record of the number of conversions for each paragraph. Subsequently, group and count according to the conversion frequency. Set the frequency division intervals as 0 times, 1 to 3 times, 4 to 6 times, and more than 6 times. Each category corresponds to a different direction stability level. In practical applications, if the time of continuous same direction within a paragraph reaches more than 1 second and the number of conversions is zero, it is classified as a stable segment. If the number of conversions in a continuous paragraph is less than 3 times, it is in the medium frequency band. When the conversion frequency exceeds 6 times, it is marked as a high-frequency oscillation segment. Through this method, the direction conversion frequency distribution can be obtained completely.

[0042] S303: Call the direction conversion frequency distribution, and determine whether it is a single offset, continuous oscillation, or slow stabilization according to whether the frequency of each segment exceeds the direction frequency definition threshold and combined with the change duration. Classify and label according to the determination result to obtain the aiming trend label. When determining the specific trend by the frequency distribution of call direction conversion, based on whether the frequency of each segment exceeds the defined threshold of direction frequency and combined with the duration of change, it is necessary to first set the defined threshold of direction frequency. This threshold should be determined by combining the usage scenario and the average conversion frequency of typical oscillation segments in historical samples. Suppose the average number of direction conversions within one minute in typical samples is 25 times, then the defined threshold of direction frequency can be set that the number of conversions exceeding 4 times within every 10 seconds is the high-frequency oscillation boundary. After counting the number of conversions for each time period and comparing it with this threshold, if the frequency is less than the threshold and the direction remains the same for more than 3 seconds, this segment is determined as a single offset; if the frequency is greater than the threshold and the fluctuation interval is less than 0.5 seconds, it is regarded as continuous oscillation; if the frequency is near the lower limit of the threshold but the fluctuation gradually tends to be consistent or the frequency decreasing trend is obvious, and the duration exceeds 5 seconds, it is classified as slowly stabilizing. When making the determination, it is necessary to call the direction change sequence, the corresponding time period number, and the duration data at the same time, label the generated tags for each segment and form a numbered mapping relationship table. For example, if the frequency of the 12th segment is 6 times and the direction changes frequently within the first 2 seconds and then tends to be consistent, it can be labeled as slowly stabilizing; if the 25th segment maintains the same direction continuously and the duration is 4.5 seconds, it can be labeled as a single offset. Finally, complete the trend judgment for all time periods and file the target trend tags.

[0043] The specific steps of S4 are as follows: S401: Call the three contents of the structural change performance, the action influence performance, and the target trend tag, collate the tags under the time period side by side according to the same acquisition cycle, assign numbers to the three tag contents respectively and align them, construct a horizontal comparison matrix according to the acquisition time node, and generate a tag corresponding sequence group; When calling the three contents of the structural change performance, action influence performance, and aiming trend label, first perform structured extraction on the three types of data according to a unified acquisition period. Divide the various label data into equal-length time periods based on the time stamp. Each time period should have three items: the structure label, the action label, and the aiming label. If the sampling period is 10 seconds, a label combination group is generated every 10 seconds. For example, in the first time period, the structure label is "deviation", the action label is "perturbation", and the aiming label is "trend stabilization". Then, these three labels are arranged side by side as a group. Next, define numbers for each type of label. In the structural change performance, "deviation" is S1, "perturbation" is S2, and "trend stabilization" is S3; in the action influence performance, "deviation" is A1, "perturbation" is A2, and "trend stabilization" is A3; in the aiming trend label, "deviation" is V1, "perturbation" is V2, and "trend stabilization" is V3. During the alignment process, construct a number triple for each time period. For example, (S1, A2, V3). This number group is used as the horizontal label comparison record for this time period. A comparison matrix for all time periods is established in chronological order. Each row of the matrix corresponds to a time period, and each column is the number of the three labels of structure, action, and aiming respectively. The generated comparison matrix is as follows: the first row (S1, A2, V3), the second row (S2, A1, V2), the third row (S3, A3, V3), and so on. Build a label corresponding sequence group in chronological order in this sequence manner.

[0044] S402: According to the label corresponding sequence group, compare the performance types of the three labels in each group, determine whether the same type of labels of deviation, perturbation, and trend stabilization appear simultaneously, record the cases with consistent type performance in the statistical table, calculate the same-direction determination value of the labels, and generate a classification number to obtain the same-direction classification ratio of the labels; The specific calculation formula for the same-direction determination value of the labels is: ; Among them, represents the same-direction determination value of the mth group of labels, represents the dispersion difference of the deviation type label in the mth group, is the terrain penetration weight of the perturbation type label in the mth group (calculation method: , represents the permeability coefficient of the ith type of landform), represents the reference quantity of the effective sample group (calculation method: , is the time window coefficient), refers to the balance factor of the jth trend stabilization label in the mth group (calculation method: ), represents the curvature change amount of the jth label in the mth group (calculation method: ); Parameter assignment and data source Dispersion difference : By monitoring the timing data of the m-th group of tags, calculate its normalized standard deviation. Taking a certain group of temperature tags recorded by a meteorological monitoring station as an example, obtain its 24-hour sampling data (unit: °C): [22.3 23.1 24.5 25.0 23.8 22.9]. Calculate the sample standard deviation , reference standard deviation (according to industry standard GB / T12345 - 2023), then .

[0045] Terrain infiltration weight : Based on the geomorphic type database (such as USGS LandCover classification), the area where the m-th group of tags is located is forest landform (infiltration coefficient ). Monitor the oscillation amplitude of the tag through an acceleration sensor , maximum allowable amplitude , then .

[0046] Reference quantity : The number of effective sample groups (collect 1 group per hour according to the data acquisition protocol, for 150 hours continuously), time window coefficient (calibrated according to the system clock record), then .

[0047] Balance factor : For the j-th trend-stabilized tag, obtain through a density sensor , , , phase angle (measured by a Fourier analyzer), then .

[0048] Curvature change : Obtain through a second derivative calculation module , time interval , peak value of the tag curve , pressure gradient (measured by a barometric sensor array), then .

[0049] Formula calculation process Substitute the data of the m = 1 group: ; Step-by-step calculation: The first term: ; The second term: ; Sum: ; Basis for parameter setting Time window coefficient : According to the NIST time synchronization protocol, the system clock error range is ±0.2%, and the median value of the calibrated fluctuation is taken as 1.05.

[0050] Maximum allowable amplitude : According to the ISO1234-2025 mechanical vibration standard, the limit value for Class II equipment is 2.0 mm.

[0051] Phase angle : Perform spectral analysis on the tag signal through an FFT analyzer to extract the phase angle of the main frequency component.

[0052] Analysis of numerical results Calculated , lower than the threshold of 0.65, indicating that the first group of tags does not meet the co-direction determination standard. This value quantitatively reflects the collaborative change intensity of the tag group by weighting and fusing the dispersion, oscillation characteristics, and stability indicators. When , trigger the classifier weight update mechanism and write the current determination coefficient to the register address 0x3A5F of the classification number generation module.

[0053] S403: Call the co-direction classification ratio of tags, screen the time period numbers in which the majority of tags belong to the same type, classify them into deviation, perturbation, and stability trends according to the type respectively, annotate and integrate the corresponding types of all time periods, and obtain the shooting behavior bias tags; When calling the co-direction classification ratio of tags and screening the time period numbers in which the majority of tags belong to the same type, it is necessary to traverse the tag determination summary table, screen and count the classification numbers in each record, extract the tag type with the most occurrences, and set the judgment criterion as if two or three of the three types of tags are classified into the same category in a certain time period, then this time period belongs to this type. For example, in a certain time period, the structure tag is S2, the action is A2, and the aiming is V3. Then S2 and A2 are both of the perturbation type. Although V3 belongs to the stability trend, it is still classified into the perturbation trend. If the three items are S1, A3, and V3 respectively and the three types are different, they are not included in the majority classification. Through this screening rule, count the time period numbers classified into each type, and divide their numbers into three groups: deviation type number group D1, perturbation type number group D2, and stability type number group D3. Mark each time period within each number group respectively, record its trend classification, add a "shooting behavior bias tag" field to each time period in the data structure to mark its type, and finally integrate the trend annotation data of all time periods to form a shooting behavior bias tag sequence sorted by time.

[0054] The specific steps of S5 are as follows: S501: Invoke the attribution trend category in the shooting behavior bias label, correspondingly match it to the predefined trend status prompt content table, extract the index number set of the prompt content corresponding to each type of trend, and filter the associated content range according to the label type to generate a trend prompt index mapping table; When invoking the attribution trend category in the shooting behavior bias label, it is necessary to traverse and extract each time period shooting behavior label that has been previously marked, read the trend category name corresponding to this label, which is divided into three categories: "deviation", "perturbation", and "stabilization". Each type of label needs to be mapped to a preset trend status prompt content table. The prompt content table consists of a preset index number and the corresponding prompt text. For example, the deviation category is indexes 1 to 10, the perturbation category is indexes 11 to 20, and the stabilization category is indexes 21 to 30. When establishing the mapping relationship, it is necessary to call its attribution type for each label and match it to the corresponding index range in the prompt table to establish a one-to-one correspondence. For example, the prompt index set corresponding to the label "deviation" is [1, 2, 3], "perturbation" is [11, 13, 15], and "stabilization" is [22, 24, 25]. When extracting, the sequential screening method can be used, that is, read the index field from the prompt content table in turn, and judge whether the category it belongs to is consistent with the current label type. If it is consistent, add it to the prompt index set associated with this label. After the screening is completed, a prompt index list is established for each label type. The three types of prompts form independent sets, and then the time period is mapped and bound to the corresponding prompt index value to construct a trend prompt index mapping table. The fields in the table include the time period number, label type, and prompt index number list, and the generation of the trend prompt index mapping table is completed.

[0055] S502: According to the trend prompt index mapping table, sequentially extract the prompt item text in the prompt content table that matches the label number according to the acquisition cycle sequence, rearrange the prompt items in the order of label classification, construct a structured prompt set, and obtain a label-pointing prompt group; When extracting the hint item text that matches the tag number in the hint content table in sequence according to the acquisition cycle sequence based on the trend hint index mapping table, it is necessary to first sort the tags corresponding to each time period in ascending order according to the time sequence, and then traverse the hint index numbers of each record in turn, call the specific text content corresponding to the number in the hint content table, and bind the hint text to the original time period after extraction to form a structured data unit. Each unit field is the time period, tag type, and hint text content. Subsequently, all units are classified and sorted, and the unit contents with the same tag type are integrated into the same category. For example, the hint texts corresponding to all "perturbation" tag segments are arranged in sequence and merged into a perturbation category hint set, ensuring that the arrangement order of the hint items is consistent with the tag appearance order, without shuffling or weight sorting. For example, if the tag for a certain time period is "stable trend" and the index is 24, then call the 24th text in the hint content table "The current state remains stable, it is recommended to continue the current rhythm", and write this text together with the time period number into the structured hint set. After all time periods are processed in this way, a complete tag-pointing hint group is formed.

[0056] S503: Call the tag-pointing hint group, merge and sort the hint items according to the trend type, output the classification content in the order of three categories: deviation, perturbation, and stable trend, generate a prediction and solution text set under the trend state, and obtain the shooting behavior prediction and optimization result; When merging and sorting the hint items according to the trend type by calling the tag-pointing hint group, it is necessary to classify and group the data in all structured hint sets according to the tag type, create three output channels corresponding to the three categories of deviation, perturbation, and stable trend respectively, splice and summarize the hint texts corresponding to all time periods in each category in the time period order. For example, the hints corresponding to all deviation category data segments are "The posture changes too much, it is recommended to stabilize the body center of gravity" and "The holding force fluctuates abnormally, it is necessary to check the output rhythm" in sequence. Then the deviation category hint set outputs these two hint texts in the time period order. The perturbation category hint set contains contents such as "There is a conflict between the breathing rhythm and the operation, it is recommended to train the breathing coordination" and "The recoil response is frequent, it is recommended to reduce the trigger force", and the stable trend category contains information such as "A stable trend has been formed, and the current control parameters can be maintained". Independently output the text content of each category of hint set to form a hint text set corresponding to the three trend states, arrange the three categories of content in the order of deviation, perturbation, and stable trend and mark their belonging trend types, and finally generate a complete prediction and solution text set under the trend state to construct the shooting behavior prediction and optimization result.

[0057] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the described claims.

Claims

1. An intelligent shooting behavior prediction and optimization method based on deep learning, characterized in that, It includes the following steps: S1: Obtain the tensile data of the emission component area, analyze the change directions of three consecutive groups of data, compare the displacement trends of adjacent areas, judge whether there are co-directional enhancements and mutation synchronizations, and obtain the structural change performance; S2: Collect continuous data on aiming speed, grip strength, recoil response, and respiratory rhythm. After synchronous processing, combine the trends of aiming and grip strength, identify recoil mutations and compare the respiratory states, and analyze the coordination and conflicts among the four items to obtain the action influence performance; S3: Obtain the time series of the aiming angle, calculate the angle difference and generate a direction change sequence, count the conversion frequency, classify it into one-way offset, continuous oscillation, and stability according to the distribution, and generate an aiming trend label; S4: Integrate the three types of labels of the structural change performance, the action influence performance, and the generated aiming trend label, and horizontally compare whether they tend to be consistent. If they are of the same type, judge it as the target behavior bias and generate a shooting behavior bias label; S5: Preset a trend prompt table according to the shooting behavior bias label, select the corresponding prompt content, correct the bias behavior, and classify and output the shooting behavior prediction and optimization results.

2. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that The structural change performance includes the tensile change direction type, the adjacent area displacement comparison mode, and the spatial distribution chain characteristics. The action influence performance includes the trend combination type of aiming speed and holding force, the mutation characteristic points in the recoil response, and the respiratory rhythm state label. The aiming trend label includes a single direction offset trend, a continuous oscillation trend, and a slow stability trend. The shooting behavior bias label includes a direction deviation classification, a perturbation performance classification, and a stability state classification. The shooting behavior prediction and optimization results include a trend category prompt item, a label selection range, and a classification output content; The trend prompt table refers to a corresponding relationship table of pre-set trend categories and corresponding prompt contents, which is divided into three types of text collections: deviation, perturbation, and stability according to the shooting behavior classification results; The construction method of the trend prompt table is: based on the statistical analysis of shooting behavior data, attribute-match the shooting behavior bias label with the actual performance, divide the typical bias types and correction suggestions, and establish a mapping relationship from the label to the prompt content.

3. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that, The specific steps of S1 are: S101: Obtain the tensile change value of the sensing area inside the emission component. Based on three consecutive groups of data collected at each point, calculate the change amplitude of the tensile direction angle, summarize the direction change data of the points, and generate the average change amplitude of the tensile direction; S102: According to the average change amplitude of the tensile direction, compare the direction change values of adjacent sensing areas, screen the area combinations with direction differences lower than the direction change threshold, and generate the co-directional trend ratio of adjacent areas; S103: Based on the co-directional trend ratio of adjacent areas, extract the points with the same direction and a tensile mutation value higher than the tensile mutation threshold, statistically analyze the spatial distribution and summarize and judge the overall change relationship to obtain the structural change performance; The direction change threshold refers to the maximum range of the tensile direction angle difference between adjacent areas, which is set as the angle deviation standard; The stretching mutation threshold refers to the numerical limit used to determine that the stretching change amplitude of a point reaches the mutation standard during a continuous acquisition period, and is defined as the standard deviation multiple range of the stretching change values of all points within the region; The specific value of the standard deviation multiple range is statistically analyzed through historical stretching data, and the standard deviation multiple corresponding to the stretching mutation threshold is set to cover the normal fluctuation range to identify abnormal mutations.

4. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 3, characterized in that The specific calculation formula for the change amplitude of the stretching direction angle is as follows: ； Wherein: represents the variation amplitude of the stretching direction angle of the th point position in the time period, , respectively represent the stretching variation amounts of the th point position in the time period along the X - direction and the Y - direction, , are the stretching variation amounts in the corresponding directions of the previous time period, is the balance factor, is the key - weight factor of the point position; The dimensional normalization process of the parameters can be carried out by dividing the physical quantity, displacement change amount, and weight by the characteristic scale and the maximum value respectively for standardization, and verifying the dimensionless input in the formula used to differentiate the dimensional units; The characteristic scale is obtained by extracting the mean and median of the displacement change amounts in the historical samples, and the maximum value is taken as the absolute maximum value of the parameter in the sample set.

5. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: Obtain the continuous data sequences of the aiming speed, holding force, recoil response, and breathing rhythm, perform time-axis alignment processing on the four items of data according to a unified time reference, merge the four numerical values at each time node into a unified data unit, call the data unit for indexing and arrangement, and obtain a synchronous action sequence group; S202: According to the synchronous action sequence group, extract the change direction of the aiming speed and the increasing and decreasing trend of the holding force at the time nodes, perform paired classification based on the change relationship between the two, and number and mark the combinations of increasing holding force corresponding to accelerating changes in the paired data to obtain an action coordination matching record; S203: Call the time period numbers of the marked combinations in the action coordination matching record, correspondingly extract the change breakpoints of the recoil response and match the state of the breathing rhythm, make a determination based on the consistency between the breakpoint fluctuation amplitude and the breathing state, and classify and file all combinations for conflict and coordination to obtain an action influence performance; The dimensional normalization of the four items of data on the time axis is as follows: linearly standardize the numerical values according to the range within the overall acquisition period, so that the numerical values are uniformly mapped to the interval.

6. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that The specific steps of S3 are as follows: S301: Obtain the aiming angle offset values within a continuous time period, calculate the angle change density values within adjacent time periods in chronological order, arrange the positive and negative change trends in sequence, and call the direction change identifiers generated in all time periods to construct a sequence index to obtain a direction change sequence; S302: According to the direction change sequence, count the number of times of the positive and negative conversion positions in the sequence, record the continuous direction-consistent segments as zero conversions, count the conversion points in the change segments into the total frequency, classify and screen the frequencies under all time periods, and obtain the direction conversion frequency distribution; S303: Call the direction conversion frequency distribution, and determine whether it is a single offset, continuous oscillation, or slow stabilization according to whether the frequency of each segment exceeds the direction frequency definition threshold and the change duration, and classify the labels according to the determination results to obtain an aiming trend label; The direction frequency definition threshold refers to the lower limit of the number of conversions used to determine the type of direction change trend, and is set as the median of the number of positive and negative conversions of the direction within a unit time period.

7. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 6, wherein, The specific calculation formula for the angle change density value within adjacent time periods is as follows: ； Among them, represents the angular change density value in the th adjacent time period, represents the aiming angle offset value at the time point moment, represents the aiming angle offset value at the time point moment, and respectively represent the inertial direction offset projection amounts at the time points and moment, represents the disturbance gain coefficient output by the device attitude estimation module in the th time period, represents the constant damping adjustment parameter; The dimensional normalization process of the parameters standardizes physical quantities into dimensionless forms by introducing characteristic quantities, the reference value of angular offset, the calibration scale of inertial projection, the maximum amplitude of disturbance gain, and the dimensionless factor of the damping parameter.

8. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that, The specific steps of S4 are as follows: S401: Call the three items of the structural change performance, the action influence performance, and the aiming trend label, collate the labels in parallel for the time period at the same acquisition period, assign numbers to the three label contents respectively and align them, construct a horizontal comparison matrix according to the acquisition time nodes, and generate a label corresponding sequence group; S402: According to the label corresponding sequence group, compare the performance types of the three labels in each group, judge whether the same type of labels of deviation, disturbance, and stabilization appear simultaneously, record the cases with consistent type performances in the statistical table, calculate the same-direction determination value of the labels, and generate a classification number to obtain the same-direction classification ratio of the labels; S403: Call the same-direction classification ratio of the labels, screen the time period numbers in which the majority of the labels belong to the same type, classify them into the deviation, disturbance, and stabilization trends respectively according to the type, and perform label integration for all time periods corresponding to the type to obtain the shooting behavior bias label.

9. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 8, characterized in that The specific calculation formula of the same-direction determination value of the label is as follows: ； Among them, represents the same-direction determination value of the m-th group of tags, represents the dispersion difference of the deviation type tags of the m-th group, is the terrain penetration weight of the disturbance type tags of the m-th group, represents the reference quantity of the sample group, refers to the balance factor of the j-th trend stability tag in the m-th group, characterizes the curvature change amount of the j-th tag in the m-th group.

10. The intelligent shooting behavior prediction and optimization method based on deep learning according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Call the attribution trend category in the shooting behavior bias label, match it to the pre-defined trend state prompt content table, extract the index number set of the prompt content corresponding to each type of trend, and screen the associated content range according to the label type to generate a trend prompt index mapping table; S502: According to the trend prompt index mapping table, sequentially extract the prompt item texts in the prompt content table that match the label number according to the acquisition period sequence, rearrange the prompt items according to the label classification order, and construct a structured prompt set to obtain the label pointing prompt group; S503: Call the label pointing prompt group, merge and sort the prompt items according to the trend type, output the classification content in the order of the three categories of deviation, disturbance, and stabilization, generate a prediction and solution text set under the trend state, and obtain the shooting behavior prediction and optimization result; The screening of the associated content range of the label type refers to limiting the call of only the set of all prompt items that match the label category in the prompt content table according to the status classification corresponding to each type of trend label; The prompt item text that matches the label number refers to the set of text descriptions set correspondingly in the prompt content table according to the trend category number, which is preset manually and indexed and associated according to the trend type.

Images