Data processing method and system for light weapon shooting training intelligent target range

Abstract

The application discloses a light weapon shooting training intelligent shooting range data processing method and system, relates to the technical field of shooting range data processing, and comprises the following steps: constructing a nanosecond unified time baseline, synchronously collecting multi-source sensing signals of a shooting training shooting range by using a high-resolution optical array and an acoustic sensing matrix, performing time synchronization and delay compensation on each sensing signal through a phase alignment mechanism, and forming a unified sensing base surface for multi-target identification. The application realizes high-precision synchronous collection and interference feature stripping by constructing a nanosecond unified time baseline and fusing multi-source sensing, significantly improves target differentiation and identification accuracy in a complex background, simultaneously generates a dynamic reliability atlas and implements phase conjugate regulation in combination with a time-space coupling identification and counterfactual playback mechanism, forms an identification and correction closed loop, effectively reduces misjudgment and drift, and ensures stable tracking and continuous identification in a multi-target scene.

Description

Technical Field

[0001] This invention relates to the field of range data processing technology, specifically to a method and system for processing data from a smart range for light weapons shooting training. Background Technology

[0002] Smart range data processing for light weapons shooting training refers to the real-time collection, transmission, processing, and analysis of various data generated during shooting training by integrating modern sensors, artificial intelligence, big data analytics, and cloud computing technologies. This optimizes training effectiveness and improves shooting accuracy. Specifically, it involves the efficient collection of shooting data (such as hit location, shooting accuracy, and reaction time) and analysis using data mining and intelligent algorithms to generate training feedback and improvement suggestions. Simultaneously, a big data platform allows for the storage, analysis, and comparison of large amounts of training data, helping to analyze the impact of different environmental conditions, shooting postures, or individual shooter characteristics on shooting results. This provides a scientific basis for personalized training, tactical improvements, and overall training efficiency enhancement.

[0003] The existing technology has the following shortcomings:

[0004] In multi-target interference scenarios, especially when the target background is complex and the distance between targets is small, existing target recognition algorithms may fail to accurately distinguish between individual targets, leading to multiple targets being misidentified as a single target or mistracking. This technical problem is particularly prominent in dynamic training environments, especially in shooting training under high-density, complex backgrounds. If such misidentification or mistracking occurs, the shooter's target locking and engagement accuracy will be severely affected, preventing the shooter from effectively and accurately engaging multiple targets, thus impacting the accuracy and reliability of training results. More seriously, if this problem is not detected and corrected in a timely manner, it may prevent the training objectives from being achieved, thereby affecting the effectiveness of tactical assessments and subsequent combat decisions.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a data processing method and system for intelligent firing ranges for light weapons shooting training, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a data processing method for a smart firing range for light weapons shooting training, comprising the following steps:

[0008] Step 1: Construct a nanosecond-level unified time baseline. Use a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range. Use a phase alignment mechanism to synchronize and compensate for the delay of each sensing signal to form a unified sensing baseline for multi-target recognition, which will be used for the unified processing of target features in the future.

[0009] Step 2: Based on the unified sensing baseline, interference feature stripping is performed on the collected multi-source sensing signals. An interference feature stripping model is established, and time-frequency decomposition and energy suppression are performed on the pseudo-feature components in the complex background. Independent feature clusters corresponding to each target are extracted, and the independent feature clusters are used as the steady-state reference for subsequent target identification.

[0010] Step 3: Based on independent feature clusters, construct a spatiotemporal coupled recognition network, map the independent feature clusters into dynamic convolutional tensors, enhance the discriminativeness of target boundaries through a phase weight self-calibration mechanism, achieve accurate recognition of each target in multi-target interference scenarios, and output recognition results containing target boundaries and temporal information.

[0011] Step 4: Based on the recognition results of the spatiotemporal coupled recognition network, construct a counterfactual replay chain, perform time reversal replay of the recognition path, detect and correct misjudgments and boundary drifts in the recognition results, and generate a dynamic credibility map reflecting temporal consistency for reliability assessment of the recognition results.

[0012] Step 5: Based on the time-series information of the dynamic credibility map, implement phase conjugate dynamic control to perform rhythmic traction, reverse diffusion correction, and amplitude limiting write-back on the tracking signal during the target recognition process, so as to achieve closed-loop control of recognition and correction, thereby obtaining stable target differentiation and accurate tracking results in complex, multi-target interference scenarios.

[0013] Preferably, the steps for constructing a nanosecond-level unified time baseline include the following sub-steps:

[0014] High-resolution optical arrays and high-sensitivity acoustic sensing arrays are deployed at the shooting training range to collect ballistic trajectory, target status, shooting action and impact sound information, respectively, and to establish a time sampling reference benchmark with nanosecond-level accuracy, so that all sensing devices can perform synchronous sampling under a unified clock framework.

[0015] After the device deployment and time reference setting are completed, the multi-source sensing signals of the actual shooting event are synchronously collected according to the unified time baseline, and a nanosecond-level timestamp is added to each data collection unit to maintain the time correlation of the multi-source signals.

[0016] After the data acquisition is completed, the multi-source sensing signals are synchronized and delayed by time synchronization and delay compensation based on the difference in the propagation speed of light and sound waves, and the signals from different channels are re-aligned to a unified timing sequence.

[0017] The synchronized and compensated multi-source sensing data is mapped onto a unified spatiotemporal recognition framework to form a unified sensing base surface for multi-target recognition, which is then used for unified processing of target features.

[0018] Preferably, in the step of mapping the synchronized and compensated multi-source sensing data to a unified spatiotemporal recognition framework, the spatial distribution relationship of the data collected by the optical array and the acoustic sensing array is jointly calibrated, and the spatiotemporal coordinates of each sensing point are matched accordingly based on the nanosecond-level time baseline, so as to ensure both time synchronization accuracy and spatial positioning consistency in the process of forming a unified sensing base surface.

[0019] Preferably, the steps for removing interference features from the multi-source sensing signals acquired based on a unified sensing base include:

[0020] The signals from each sensing device are reviewed in time and space, and continuous sensing trajectories are extracted, marking potentially high-energy interference areas;

[0021] Identify anomalous components within the marked high-energy interference region and determine their time-frequency distribution, duration, and spatial extension characteristics.

[0022] The identified anomalous components are subjected to targeted energy suppression and multidimensional feature stripping, and real target features with consistency in temporal rhythm, spatial distribution and frequency characteristics are preserved according to the temporal and spatial mapping of the unified sensing base surface.

[0023] The remaining features after suppression and stripping are aggregated according to their temporal evolution trajectory and spatial response location to form independent feature clusters corresponding to each target. These independent feature clusters serve as steady-state references for subsequent target identification.

[0024] Preferably, in the step of performing directional energy suppression and multidimensional feature stripping on the anomalous components, the temporal reference and spatial mapping boundary of the unified sensing base surface are used as references to perform directional suppression on the energy peak of the interference components, and transition processing is performed in the boundary region to avoid the erroneous stripping of the real target features or interference residues, thereby ensuring that the formed independent feature clusters have temporal continuity and spatial boundary stability.

[0025] Preferably, the steps for constructing a spatiotemporally coupled identification network based on independent feature clusters include:

[0026] The temporal series features, spatial response features, and multi-channel sensing features of each independent feature cluster are uniformly reconstructed to establish a feature expression framework with spatiotemporal consistency.

[0027] The feature representation framework is mapped to a continuously deformable dynamic representation structure, in which the temporal continuity, spatial hierarchicality and channel response consistency of each feature cluster are maintained;

[0028] A phase weight self-correction mechanism is introduced into the dynamic expression structure to analyze the phase difference of multiple targets at the same time node, and to generate enhanced feature response peaks at the target boundary positions to enhance boundary discrimination.

[0029] Based on the feature representation results enhanced by phase weight, the spatial boundary and temporal evolution trajectory of the target are extracted to form an identification result containing target boundary and temporal information, which is used for subsequent target tracking and correction processing.

[0030] Preferably, in the phase weight self-calibration mechanism, the phase jump point at the target boundary is extracted by comparing the phase response offset of adjacent time nodes, and the phase weight distribution is dynamically adjusted according to the jump intensity to form a boundary response enhancement region in the spatial mapping, thereby further improving the boundary clarity and recognition accuracy between close targets.

[0031] Preferably, the step of constructing a counterfactual playback chain based on the recognition results of the spatiotemporal coupling identification network includes:

[0032] Extract the time series response and spatial boundary evolution path of each target in the target recognition results output by the spatiotemporal coupling recognition network to form a set of recognition trajectories with reversible reconstruction capability;

[0033] Based on the set of identified trajectories, a counterfactual playback chain is constructed to perform time reversal and replay of the identified path, and compare it with the original identification results to identify abnormal nodes and boundary drift.

[0034] Based on the abnormal nodes and drift segments identified during the inversion process, the identification results are smoothed at the boundary and corrected for misjudgments, and then aligned consistently by combining the steady-state reference of independent feature clusters.

[0035] A dynamic confidence map is constructed based on the stable performance of the corrected identification path during the time inversion process. The identification confidence level is recorded with time as the horizontal axis and target identification as the vertical axis for subsequent reliability assessment.

[0036] Preferably, the steps for implementing phase conjugate dynamic control based on the time-series information of the dynamic confidence map include:

[0037] Extract the temporal information of each target identification path in the dynamic credibility map, analyze the trend of credibility level change and construct the target rhythm change profile to identify rhythm drift and abnormal response segments;

[0038] Based on the rhythm change profile, a rhythm traction operation is performed on the rhythm abnormal section in the identification path to adjust the time density and displacement velocity of the abnormal response node to the range of the rhythm reference curve.

[0039] For trajectory segments that still exhibit spatial boundary diffusion or target fusion after rhythmic traction, inverse correction is implemented by combining historical records of the dynamic credibility map, so that the target boundary is recovered along the historical path and tends to converge.

[0040] After completing rhythmic traction and reverse diffusion correction, the recognition response range of each target is limited and written back, and the response amplitude and spatial boundary threshold are set to form a stable target tracking output structure, thereby realizing closed-loop control of recognition and correction.

[0041] The data processing system for the intelligent firing range for light weapons shooting training includes a multi-source sensing synchronous acquisition module, an interference feature stripping processing module, a spatiotemporal coupling identification and enhancement module, a counterfactual playback correction module, and a phase conjugate dynamic control module.

[0042] The multi-source sensing synchronous acquisition module constructs a nanosecond-level unified time baseline, uses a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range, and performs time synchronization and delay compensation for each sensing signal through a phase alignment mechanism to form a unified sensing baseline for multi-target recognition.

[0043] The interference feature stripping module, based on a unified sensing baseline, performs interference feature stripping on the acquired multi-source sensing signals, establishes an interference feature stripping model, performs time-frequency decomposition and energy suppression on pseudo-feature components in complex backgrounds, and extracts independent feature clusters corresponding to each target.

[0044] The spatiotemporal coupling identification enhancement module constructs a spatiotemporal coupling identification network based on independent feature clusters, maps independent feature clusters into dynamic convolutional tensors, and enhances the discriminativeness of target boundaries through a phase weight self-calibration mechanism, thereby achieving accurate identification of each target in multi-target interference scenarios and outputting identification results containing target boundaries and temporal information.

[0045] The counterfactual replay correction module constructs a counterfactual replay chain based on the recognition results of the spatiotemporal coupling recognition network, performs time reversal replay of the recognition path, detects and corrects misjudgments and boundary drifts in the recognition results, and generates a dynamic credibility map that reflects temporal consistency.

[0046] The phase conjugate dynamic control module, based on the time-series information of the dynamic credibility map, implements phase conjugate dynamic control to perform rhythmic traction, reverse diffusion correction, and amplitude limiting write-back on the tracking signal during the target recognition process, thereby achieving closed-loop control of recognition and correction.

[0047] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0048] This invention establishes a multi-source sensing synchronization mechanism covering all temporal and spatial dimensions by constructing a nanosecond-level unified time baseline and integrating high-resolution optical and acoustic sensing methods. This enables high-precision, low-latency acquisition of target information during the acquisition phase. Phase alignment and latency compensation effectively address the time-axis offset problem of multi-channel data, providing a stable and consistent sensing foundation for subsequent recognition processing. Furthermore, by combining interference feature stripping and independent feature cluster extraction, it can accurately separate false features from real target features under complex background conditions, significantly improving target discrimination and providing stable input for accurate recognition in multi-target scenarios. The entire process achieves high temporal consistency and spatial decoupling from the sensing starting point, significantly improving recognition accuracy and system robustness under conditions of densely distributed multi-targets.

[0049] This invention introduces a spatiotemporal coupling identification mechanism and a counterfactual playback chain construction process. Through time reversal and boundary correction of the target identification path, it achieves dynamic verification and continuity assessment of the identification results in the post-identification stage, generating a dynamic credibility map with temporal credibility levels. Based on this map, a phase conjugate modulation strategy is further implemented to perform rhythmic traction and amplitude-limiting write-back operations on the tracking signal, thereby forming a stable identification-correction closed loop at the target tracking level, enabling continuous target identification and high-precision dynamic tracking in complex environments. The overall solution not only effectively reduces problems such as misjudgment, boundary drift, and multi-target fusion, but also significantly improves the continuity of the target trajectory and the system's adaptability in dynamic task environments. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a flowchart of the data processing method for the intelligent firing range of light weapons shooting training according to the present invention.

[0052] Figure 2 This is a schematic diagram of the modules of the intelligent target range data processing system for light weapons shooting training of the present invention. Detailed Implementation

[0053] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0054] This invention provides, for example Figure 1The data processing method for a smart firing range for light weapons shooting training, as shown, includes the following steps:

[0055] Step 1: Construct a nanosecond-level unified time baseline. Use a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range. Use a phase alignment mechanism to synchronize and compensate for the delay of each sensing signal to form a unified sensing baseline for multi-target recognition, which will be used for the unified processing of target features in the future.

[0056] The specific steps to achieve this process are as follows:

[0057] High-resolution optical arrays and high-sensitivity acoustic sensing arrays are deployed at the shooting training range to capture various types of information, including ballistic trajectories, target states, firing actions, and impact sounds. The optical arrays should possess high spatial resolution, capable of acquiring richly detailed images from different perspectives; while the acoustic sensing arrays should have broad sensing capabilities covering the entire target area and the ability to accurately model sound wave propagation delays. After the various sensing devices are deployed, a precise time sampling reference is established, ensuring that all sensing devices perform synchronous sampling within a unified clock framework. This fundamentally avoids data misalignment or analytical bias caused by inconsistent sampling times. This time reference operates with nanosecond-level precision, maintaining high temporal consistency among multiple sensing points even in high-density target training scenarios, forming a unified time starting point at the sensing level.

[0058] After completing the deployment of the sensing devices and setting the time reference baseline, the synchronous acquisition of multi-source sensing signals for actual shooting events occurring at the firing range began. In this phase, all optical and acoustic sensing arrays acquired corresponding raw sensing data according to a unified time baseline. Due to the different spatial layouts of the sensing devices, the arrival times of the sensing signals for the same event in different devices would inevitably have slight differences. To control these differences within an effective recognition range, each data unit acquired in this phase was accompanied by a timestamp with nanosecond precision, used to identify the acquisition time of the data under the unified time baseline, thus ensuring a strict temporal correlation between the multi-source data. Simultaneously, through the coordinated operation of optical and acoustic sampling, the dynamic changes generated by each shooting action could be recorded simultaneously in multiple sensing channels, maintaining comparability in spatial location and temporal order.

[0059] After completing the synchronous acquisition, precise time synchronization and delay compensation processing is performed on the aforementioned multi-source sensing signals. Because light waves and sound waves propagate at different speeds in the target range, and because the distances between sensing devices vary, subtle delay differences may occur in the recording time points of the same shooting event across different sensing channels. Therefore, at this stage, a nanosecond-level time baseline is used as the standard for aligning all data. By precisely comparing the timestamps attached to the acquired data, the time offset caused by differences in the propagation medium and distance along the actual propagation path is identified for each channel's data. After identifying the time offset, delay compensation is performed on each sensing signal according to the synchronization framework of the unified time baseline. This involves repositioning the reference time point of the data to realign data from the same event in different channels onto the same timeline, ensuring that the temporal consistency of the data is not disrupted by sensing errors during subsequent target state identification, position determination, and trajectory reconstruction.

[0060] After completing time synchronization and latency compensation for all collected data, the aligned multi-source sensing data is mapped onto a unified spatiotemporal recognition framework, forming a complete multi-target recognition sensing baseline. This sensing baseline is a highly unified data representation platform with high spatiotemporal consistency. Its core advantage lies in its ability to maintain the spatial distribution characteristics of data while strictly ensuring the temporal synchronization between data points, thereby constructing a target recognition reference structure with high spatial accuracy and temporal coherence. Within this sensing baseline, the sensing data corresponding to each shooting target can be clearly distinguished and accurately located, regardless of the complexity of its background or the closeness of its distance from other targets. Subsequent feature extraction, target separation, and tracking analysis can be performed within the unified reference framework. Simultaneously, this sensing baseline can also serve as a stable input source for subsequent interference stripping, feature clustering, and dynamic identification processes, ensuring the accuracy and continuity of subsequent processing flows. Through this approach, clear, stable, and sustainable unified sensing processing of multiple targets is ultimately achieved in high-density, highly interfered shooting training scenarios, laying a high-temporal-sequence, high-consistency sensing foundation for the data analysis and recognition process of the entire smart firing range.

[0061] This implementation constructs a nanosecond-level unified time baseline, combines a high-resolution optical array with an acoustic sensing array to conduct synchronous acquisition of multi-source signals, and then performs time synchronization and delay compensation processing. Finally, a target recognition sensing base surface is formed under a unified spatiotemporal reference, which not only effectively improves the multi-target resolution capability in complex scenes, but also provides a stable and reliable foundation for subsequent feature stripping and identification processing.

[0062] Step 2: Based on the unified sensing baseline, interference feature stripping is performed on the collected multi-source sensing signals. An interference feature stripping model is established, and time-frequency decomposition and energy suppression are performed on the pseudo-feature components in the complex background. Independent feature clusters corresponding to each target are extracted, and the independent feature clusters are used as the steady-state reference for subsequent target identification.

[0063] The specific steps to achieve this process are as follows:

[0064] Based on the established unified sensing baseline, the temporal and spatial consistency of each sensing signal is verified to ensure high synchronization and spatial alignment accuracy of the data involved in subsequent processing. In this stage, signals from different sensing devices are uniformly sorted in the temporal domain and mapped in the spatial domain to form a complete, dual-domain continuous data structure. This data structure possesses strong temporal dependence and spatial response coherence, providing a structured input foundation for interference stripping. Based on this, continuous sensing trajectories representing target activity are extracted, and potentially high-energy interference background regions and their corresponding signal segments are marked, laying the groundwork for subsequent identification of false feature components.

[0065] After completing the temporal and spatial verification, an in-depth interference component identification operation is performed on the sensing signals contained in the unified sensing datum. Specifically, within the potentially high-interference area identified in the previous step, the changing trends, response amplitudes, and spatial distribution characteristics of each sensing signal are analyzed. Combined with their fluctuation patterns in the time series, anomalous components inconsistent with the target activity are identified. These anomalous components typically exhibit rhythmicity, response persistence, or spatial diffusion patterns different from the target characteristics, and may originate from external factors such as mechanical vibrations, echo interference, lighting changes, obstruction by debris, or reflections from training facilities in the target range background. After identifying these interference components, their distribution patterns in the time and frequency domains are modeled, and the dominant response frequency bands, durations, and spatial extension directions of the interference components are extracted, providing clear targets for subsequent energy suppression and feature purification operations.

[0066] After clarifying the response characteristics of the interference components, energy suppression and multi-dimensional feature stripping are performed. In this stage, using a unified sensing baseline as a reference, all sensing signals are mapped into a data representation framework with three dimensions: time, frequency, and spatial domains. Within this framework, the energy peaks corresponding to the identified interference components are targeted for suppression, reducing their influence on the overall sensing data. Simultaneously, through multi-dimensional slicing of the interference components in terms of frequency variation, response intensity, and spatial diffusion direction, they are systematically stripped from the original sensing data, preserving those true target features that are consistent in temporal rhythm, stable in spatial distribution, and exhibit consistency in frequency variation. To ensure the accuracy of the stripping process, the boundary regions are also transitionally processed based on the previously constructed time synchronization reference and spatial mapping boundary, avoiding the mis-cutting of effective features or the retention of interference components during the stripping process.

[0067] After suppressing the energy and stripping the features of the interference components, the remaining sensing data undergoes feature aggregation to form independent feature clusters corresponding to each real target. These independent feature clusters should possess high stability, consistency, and distinguishability, maintaining continuous temporal response, exhibiting clear spatial boundaries, and demonstrating significant differences across sensing dimensions, thus serving as a fundamental reference for subsequent target identification operations. In this stage, the data within each feature cluster is further integrated according to the temporal evolution trajectory and spatial response location of the sensing signal, forming a structured set of target feature representations. Each target feature cluster corresponds one-to-one with a specific target entity, possessing clear temporal start and end points, spatial distribution range, and response intensity characteristics, providing an accurate reference benchmark for the next stage of target identification, tracking, and correction operations.

[0068] This implementation method, based on a unified sensing baseline, sequentially completes the spatiotemporal verification of sensing data, identification of interference components, energy suppression, and feature stripping, ultimately extracting a clearly structured independent target feature cluster. This solves the problem of pseudo-feature interference in complex backgrounds and significantly improves the robustness and accuracy of subsequent multi-target recognition and tracking processes.

[0069] Step 3: Based on independent feature clusters, construct a spatiotemporal coupled recognition network, map the independent feature clusters into dynamic convolutional tensors, enhance the discriminativeness of target boundaries through a phase weight self-calibration mechanism, achieve accurate recognition of each target in multi-target interference scenarios, and output recognition results containing target boundaries and temporal information.

[0070] The specific steps to achieve this process are as follows:

[0071] After obtaining multiple independent feature clusters, the temporal series features, spatial response features, and multi-channel sensing features contained in each feature cluster are reconstructed in a unified structure to build an expression framework with spatiotemporal consistency. This framework uses a unified sensing baseline as a reference, combining and integrating the temporal evolution process of each feature cluster with its spatial diffusion boundary to form a continuous, deformable feature mapping sequence with physical temporal logic. During this process, the complete state evolution trajectory of each feature cluster between the initial and final time points must be preserved to ensure that the dynamic trends of each feature over time are clearly expressed. Simultaneously, in the spatial dimension, the spatial coverage of each feature cluster is finely characterized, especially when multiple targets are close together or there are blurred boundary regions. The spatial positioning information retained in the previous feature stripping stage ensures that the feature clusters maintain clear spatial boundary transitions after reconstruction. This step transforms the original sensing features from discrete fragments into a unified and continuous spatiotemporal expression, laying the structural foundation for subsequent dynamic mapping and phase modulation.

[0072] After constructing the spatiotemporal representation framework of the feature clusters, all feature information within this framework is mapped into a continuously deformable dynamic representation structure. This structure uses continuous time as one axis and distributed spatial coordinates as another axis, forming a multi-dimensional feature hierarchy. During this mapping process, the differences in spatial distribution and temporal variation of each feature cluster from the previous step must be considered. A corresponding response density distribution is assigned according to its specific evolutionary form, ensuring a quantifiable intensity contrast between the feature responses of different targets. Simultaneously, the perceptual channel information contained in the original feature clusters is retained in the feature mapping structure. For example, brightness change trajectories obtained through optical responses and sound pressure evolution characteristics captured through acoustic responses are stored in a nested manner at corresponding spatial locations and temporal nodes during the mapping process, ensuring the consistency of feature responses across multiple channels. Through this approach, a dynamic representation structure is formed that is continuously progressive in the temporal dimension, has clear boundary hierarchies in the spatial dimension, and exhibits complementary responses in the perceptual dimension. This structure not only enhances the visibility of target features but also provides a clear feature hierarchy foundation for subsequent boundary enhancement operations.

[0073] After constructing the aforementioned dynamic representation structure, a recognition enhancement strategy based on a phase weight self-calibration mechanism is introduced to further strengthen the ability to identify boundaries between multiple targets. This strategy takes the phase difference in temporal evolution as its starting point, analyzing the phase response shift of each target feature cluster at the same time node. Due to differences in motion paths, response rhythms, or feature densities among multiple targets, their phase trajectories formed on the continuous time axis also exhibit discernible feature changes. At this stage, by identifying the phase fluctuation features appearing between each feature cluster at adjacent time nodes, representative boundary transition points are extracted. Then, by dynamically adjusting the weight distribution of the phase response, enhanced feature response peaks are generated at the target boundary positions. This intentionally amplifies and emphasizes the ambiguous boundaries between similar targets structurally, making subsequent interpretation clearer. This self-calibration mechanism, through proactive analysis of temporal responses, effectively avoids erroneous fusion between multiple targets due to boundary overlap or feature similarity, making the expression of target features in the spatiotemporal structure clearer and more distinguishable.

[0074] After phase weight self-calibration and enhancement, the recognition boundary and its complete temporal evolution trajectory for each target are extracted based on the enhanced feature representation results, forming the final recognition output structure. This structure should contain information in two main dimensions: first, the accurate spatial location boundary of the target, which must be continuous and closed to ensure that each target region can be completely delineated in a high-density target environment; second, the temporal response sequence of the target, which must continue from the initial appearance of the target to the end of the target state, reflecting the dynamic changes and positional drift of the target during the training process. The recognition results are encapsulated in a structured manner for subsequent stages of access and analysis, and can also serve as input for confidence assessment of the recognition process, used to determine whether there are problems such as misidentification or boundary misalignment during target recognition. Thus, through the complete process from unified spatiotemporal mapping of feature clusters to the construction of dynamic representation structures, and then to phase weight enhancement and recognition result generation, high-precision recognition of multiple targets in complex shooting training scenarios is finally achieved, ensuring clear boundaries and coherent responses between targets, providing a highly reliable basic input for subsequent tracking, correction, and feedback evaluation.

[0075] This implementation achieves accurate identification of various targets in a high-density target environment through the above steps, namely, spatiotemporal structure reconstruction of feature clusters, dynamic expression structure mapping, phase weight self-correction enhancement and recognition result generation.

[0076] Step 4: Based on the recognition results of the spatiotemporal coupled recognition network, construct a counterfactual replay chain, perform time reversal replay of the recognition path, detect and correct misjudgments and boundary drifts in the recognition results, and generate a dynamic credibility map reflecting temporal consistency for reliability assessment of the recognition results.

[0077] The specific steps to achieve this process are as follows:

[0078] After obtaining the target recognition results output by the spatiotemporally coupled recognition network, the temporal sequence response and spatial boundary evolution path of each target in the recognition results are extracted to form a set of reconstructable recognition trajectories. This set of recognition trajectories includes not only the spatial position change process of the target, but also a complete record of its response amplitude, boundary fluctuations, positional offsets, and occurrence timing throughout the recognition process. In this stage, these trajectory sets are formatted and archived to give them reversible expressive capabilities; that is, each trajectory can not only evolve forward in time, but also be reconstructed backward at the original temporal resolution. For this purpose, the timestamps and spatial markers of the recognition results need to be accurately preserved and correlated, thus providing a complete basic data framework for the subsequent construction of a time-reversal-based playback process. The extraction of these recognition trajectories aims to restore the evolutionary state of each frame or moment in the recognition path, making it the source of dynamic replay and error tracing.

[0079] After extracting the set of identified trajectories, a counterfactual playback chain is constructed to perform temporal reversal replay of the identified paths. Temporal reversal replay refers to reconstructing the identification evolution process time-by-time, starting from the target's current final identified state, until the target was initially detected or responded to. This playback process is not simply playing the identified frames in reverse order, but rather combining the interaction logic between time, space, and perception intensity to reconstruct the target's movement trend and boundary expansion path during the identification process. During counterfactual playback, the replayed state at each moment is compared and analyzed with the original forward identification result to identify the offset between the two at spatial boundaries, response intensity, or time points, thereby determining whether there is an abnormal identification phenomenon at that node. For example, if the target boundary at a certain time point exhibits nonlinear jumps or irregular contractions during the reversal process, it may indicate boundary drift or target fusion problems during forward identification; if a node shows a valid target response during the forward process but cannot trace a clear trajectory continuation during the reversal, it may be a misjudged target.

[0080] After completing the time reversal and anomaly node identification, the identification results are further corrected based on the error characteristics exposed in the identified trajectory. The correction is based on two main aspects: First, by identifying boundary drift segments during the reversal process, the abnormal fluctuation range of the target on the time axis is determined, and the spatial continuity and response consistency of the boundary positions within this range are recalculated to smooth out abnormal changes and reconstruct boundary coherence. Second, by performing contextual analysis on the temporal behavior of misjudged nodes and combining the target response state within the preceding and following time periods, it is determined whether the node has the rationality to be identified as a target. If not, the identification label corresponding to the node is removed from the trajectory set to eliminate the interference caused by false responses to subsequent target tracking and strike judgment. In this correction stage, the corrected trajectory also needs to be aligned with the steady-state feature references initially extracted from independent feature clusters to ensure that the corrected trajectory retains the original target's characteristic features while conforming to the spatiotemporal logical relationship constructed by the unified perception base in its overall structure. Such fine-tuning significantly enhances the stability and consistency of the recognition results in both time and space dimensions, providing a reliable input basis for subsequent confidence modeling.

[0081] After correcting the identification results and ensuring their structural coherence, a dynamic confidence map is constructed based on the stable performance of the identification path during the time-reversal process. This map uses time as the horizontal axis and target identification as the vertical axis, recording the confidence level of target identification at each moment. This level is comprehensively evaluated based on the following factors: whether there are boundary jumps, whether there are missing responses, whether the reversed path is continuous, whether it has undergone multiple corrections, and whether it highly matches the original feature cluster. During the construction process, target identification segments exhibiting high consistency and coherence in the time-reversal are assigned a high confidence level, while identification segments with abnormal changes or that have undergone corrections are marked with a low confidence level. These segments are distinguished in the map using color depth, transparency, or intensity steps, giving the entire identification path a hierarchical confidence structure in its visual representation. This dynamic confidence map not only allows for quantitative analysis of the reliability of the current identification results but also provides important reference for subsequent target hit judgments, training feedback confidence assessments, and instructor decision support.

[0082] This implementation method achieves comprehensive diagnosis and dynamic compensation for potential problems in the recognition process through steps such as trajectory extraction of recognition results, time reversal and replay, recognition error correction and dynamic credibility map generation. It effectively improves the credibility and structural integrity of target recognition results in complex multi-target scenarios.

[0083] Step 5: Based on the time-series information of the dynamic credibility map, implement phase conjugate dynamic control to perform rhythmic traction, inverse diffusion correction and amplitude limiting write-back on the tracking signal during the target recognition process, so as to achieve closed-loop control of recognition and correction, thereby obtaining stable target differentiation and accurate tracking results in complex, multi-target interference scenarios.

[0084] The specific steps to achieve this process are as follows:

[0085] After constructing the dynamic credibility map, the temporal information of each target recognition path recorded in the map is extracted, and the rhythmic fluctuations of the target during recognition are analyzed based on the trend of credibility level changes at time nodes. In this stage, by retrieving the start and end times, key transition points, and confidence value abrupt changes of the target recognition trajectory in the map, a rhythmic change profile of the target on the time axis is constructed, serving as the basis for judging the stability of target behavior and the credibility of dynamic response. Simultaneously, these rhythmic change patterns are mapped back to the original trajectory of the recognition result, using time sequence as an index to mark segments in the recognition process that may exhibit rhythm drift, discontinuity fluctuations, or frequency anomalies. The main purpose of this operation is to clarify whether the target recognition process conforms to the expected response rhythm, whether there are abnormal response segments that are too fast or too slow, and whether subsequent dynamic rhythmic guidance is needed, thereby achieving a priori understanding of the temporal rhythm of the tracking signal.

[0086] After constructing the rhythmic change profile, rhythmic traction is performed on the corresponding tracking signals based on time segments in the identification path where rhythmic drift or response abrupt changes occur. Specifically, within the segments marked as rhythmic anomalies in the confidence map, a set of time-symmetric rhythmic reference sequences is introduced as a reference rhythm for the target under normal tracking conditions, and this reference rhythm is compared with the current abnormal response segment time-slice by time. For response nodes that are detected to deviate from the reference rhythm, their response density and displacement velocity on the time axis are gradually adjusted to bring them closer to the rhythmic reference curve, thereby providing flexible traction and dynamic adjustment to the temporal behavior of the target during the identification process. The core of this step is to construct a temporal self-stabilizing mechanism, that is, if the target response deviates for a short time during the identification process, it is guided back through external rhythmic traction, thereby avoiding tracking drift or target misjudgment caused by rhythm disorder, laying the foundation for subsequent spatial morphological stability.

[0087] After guiding the recognition rhythm, targeted reverse correction is performed to address the spatial path anomaly diffusion problem in the target trajectory. This step mainly addresses phenomena such as target trajectory spread, boundary expansion, or multi-target fusion caused by boundary ambiguity, signal overlap, or background interference during the recognition process. In this stage, combining historical records of boundary stability in the dynamic confidence map, the spatial expansion trend and response coverage of time points marked as high-risk for boundary drift are re-analyzed. For trajectory segments exhibiting rapid boundary diffusion or contour deformation, a reverse convergence path backtracking method is used to spatially trace back along the historical boundary morphology of the recognition path, and amplitude constraints and response compression are applied to the current state, thereby achieving convergence and normalization of the target trajectory boundary. This method effectively suppresses the boundary generalization problem caused by recognition ambiguity in areas with multiple targets approaching or high-density target overlap, ensuring that each target maintains a clear, separate, and traceable state structure in the spatial dimension, enhancing the stability of spatial recognition.

[0088] After completing rhythmic traction and anti-diffusion correction, a further amplitude-limiting write-back operation is performed on all recognition results to form a final stable tracking output structure. Amplitude-limiting write-back refers to setting upper and lower thresholds for the recognition response range of each target after combining rhythm adjustment and boundary compression, to control potential response abrupt changes, signal jumps, or structural variations during subsequent tracking. In this stage, based on the comprehensive performance of the target recognition path in the dynamic credibility map, a corresponding credibility encapsulation interval is assigned to each target. This interval covers the upper limit of the time response rate, the maximum amplitude of the spatial expansion boundary, and the stable interval of the sensing intensity. During amplitude-limiting control, response data exceeding the boundary thresholds are restricted, and the final processed recognition trajectory is written back to the target tracking record, ensuring that the target data referenced in subsequent tracking, strike judgment, and feedback generation processes remains structurally stable, rhythmically reasonable, and spatially clear. Through amplitude-limiting write-back, closed-loop control of the recognition link is achieved, constructing a closed-loop structure from feature extraction, target recognition, result correction to output stabilization, significantly improving the robustness and response consistency of the entire recognition and tracking process.

[0089] This implementation method, through the above processing steps—rhythm information extraction, rhythmic traction adjustment, spatial back-diffusion correction, and amplitude-limiting write-back integration—deeply modulates the tracking signal in the recognition result based on the temporal evolution information of the dynamic credibility map, constructing a stable, continuous, and controllable target recognition and tracking feedback closed loop. When facing challenging scenarios in shooting training, such as dense multi-target formations, strong interference, and blurred boundaries, this modulation method demonstrates strong adaptability and structural stability, ensuring that the state of each target is credible, the response is clear, and the path is continuous during the recognition output process. This provides support for evaluating shooting training effectiveness, generating tactical judgment criteria, and optimizing intelligent feedback mechanisms.

[0090] This invention establishes a multi-source sensing synchronization mechanism covering all temporal and spatial dimensions by constructing a nanosecond-level unified time baseline and integrating high-resolution optical and acoustic sensing methods. This enables high-precision, low-latency acquisition of target information during the acquisition phase. Phase alignment and latency compensation effectively address the time-axis offset problem of multi-channel data, providing a stable and consistent sensing foundation for subsequent recognition processing. Furthermore, by combining interference feature stripping and independent feature cluster extraction, it can accurately separate false features from real target features under complex background conditions, significantly improving target discrimination and providing stable input for accurate recognition in multi-target scenarios. The entire process achieves high temporal consistency and spatial decoupling from the sensing starting point, significantly improving recognition accuracy and system robustness under conditions of densely distributed multi-targets.

[0091] This invention introduces a spatiotemporal coupling identification mechanism and a counterfactual playback chain construction process. Through time reversal and boundary correction of the target identification path, it achieves dynamic verification and continuity assessment of the identification results in the post-identification stage, generating a dynamic credibility map with temporal credibility levels. Based on this map, a phase conjugate modulation strategy is further implemented to perform rhythmic traction and amplitude-limiting write-back operations on the tracking signal, thereby forming a stable identification-correction closed loop at the target tracking level, enabling continuous target identification and high-precision dynamic tracking in complex environments. The overall solution not only effectively reduces problems such as misjudgment, boundary drift, and multi-target fusion, but also significantly improves the continuity of the target trajectory and the system's adaptability in dynamic task environments.

[0092] This invention provides, for example Figure 2 The data processing system for the smart firing range for light weapons shooting training shown includes a multi-source sensing synchronous acquisition module, an interference feature stripping processing module, a spatiotemporal coupling identification enhancement module, a counterfactual playback correction module, and a phase conjugate dynamic control module.

[0093] The multi-source sensing synchronous acquisition module constructs a nanosecond-level unified time baseline, uses a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range, and performs time synchronization and delay compensation for each sensing signal through a phase alignment mechanism to form a unified sensing baseline for multi-target recognition.

[0094] The interference feature stripping module, based on a unified sensing baseline, performs interference feature stripping on the acquired multi-source sensing signals, establishes an interference feature stripping model, performs time-frequency decomposition and energy suppression on pseudo-feature components in complex backgrounds, and extracts independent feature clusters corresponding to each target.

[0095] The spatiotemporal coupling identification enhancement module constructs a spatiotemporal coupling identification network based on independent feature clusters, maps independent feature clusters into dynamic convolutional tensors, and enhances the discriminativeness of target boundaries through a phase weight self-calibration mechanism, thereby achieving accurate identification of each target in multi-target interference scenarios and outputting identification results containing target boundaries and temporal information.

[0096] The counterfactual replay correction module constructs a counterfactual replay chain based on the recognition results of the spatiotemporal coupling recognition network, performs time reversal replay of the recognition path, detects and corrects misjudgments and boundary drifts in the recognition results, and generates a dynamic credibility map that reflects temporal consistency.

[0097] The phase conjugate dynamic control module, based on the time-series information of the dynamic credibility map, implements phase conjugate dynamic control to perform rhythmic traction, reverse diffusion correction, and amplitude limiting write-back on the tracking signal during the target recognition process, thereby achieving closed-loop control of recognition and correction.

[0098] The data processing method for smart firing ranges for light weapons shooting training provided in this embodiment of the invention is implemented through the aforementioned data processing system for smart firing ranges for light weapons shooting training. For details of the specific methods and processes of the data processing system for smart firing ranges for light weapons shooting training, please refer to the embodiments of the aforementioned data processing method for smart firing ranges for light weapons shooting training, which will not be repeated here.

[0099] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A data processing method for a smart firing range for light weapons shooting training, characterized in that, Includes the following steps: Step 1: Construct a nanosecond-level unified time baseline. Use a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range. Use a phase alignment mechanism to synchronize and compensate for the delay of each sensing signal to form a unified sensing baseline for multi-target recognition. Step 2: Based on the unified sensing baseline, interference feature stripping is performed on the collected multi-source sensing signals. An interference feature stripping model is established, and time-frequency decomposition and energy suppression are performed on the pseudo-feature components in the complex background to extract the independent feature clusters corresponding to each target. Step 3: Based on independent feature clusters, construct a spatiotemporal coupled recognition network, map the independent feature clusters into dynamic convolutional tensors, enhance the discriminativeness of target boundaries through a phase weight self-calibration mechanism, achieve accurate recognition of each target in multi-target interference scenarios, and output recognition results containing target boundaries and temporal information. The steps for constructing a spatiotemporally coupled identification network based on independent feature clusters include: The temporal series features, spatial response features, and multi-channel sensing features of each independent feature cluster are uniformly reconstructed to establish a feature expression framework with spatiotemporal consistency. The feature representation framework is mapped to a continuously deformable dynamic representation structure, in which the temporal continuity, spatial hierarchicality and channel response consistency of each feature cluster are maintained; A phase weight self-correction mechanism is introduced into the dynamic expression structure to analyze the phase difference of multiple targets at the same time node, and to generate enhanced feature response peaks at the target boundary positions to strengthen boundary discrimination. Based on the feature representation results enhanced by phase weights, the spatial boundary and temporal evolution trajectory of the target are extracted to form a recognition result containing target boundary and temporal information; In the phase weight self-calibration mechanism, the phase jump point at the target boundary is extracted by comparing the phase response offset of adjacent time nodes, and the phase weight distribution is dynamically adjusted according to the jump intensity to form a boundary response enhancement region in the spatial mapping. Step 4: Based on the recognition results of the spatiotemporal coupling recognition network, construct the counterfactual playback chain, perform time reversal replay of the recognition path, detect and correct misjudgments and boundary drifts in the recognition results, and generate a dynamic credibility map that reflects temporal consistency. Step 5: Based on the time-series information of the dynamic credibility map, implement phase conjugate dynamic control to perform rhythmic traction, reverse diffusion correction, and amplitude limiting write-back on the tracking signal during the target recognition process, thereby achieving closed-loop control of recognition and correction.

2. The data processing method for a smart firing range for light weapons shooting training according to claim 1, characterized in that, The steps to construct a nanosecond-level unified time baseline include the following sub-steps: High-resolution optical arrays and high-sensitivity acoustic sensing arrays are deployed at the shooting training range to collect ballistic trajectory, target status, shooting action and impact sound information, respectively, and to establish a time sampling reference benchmark with nanosecond-level accuracy, so that all sensing devices can perform synchronous sampling under a unified clock framework. After the device deployment and time reference setting are completed, the multi-source sensing signals of the actual shooting event are synchronously collected according to the unified time baseline, and a nanosecond-level timestamp is added to each data collection unit to maintain the time correlation of the multi-source signals. After the data acquisition is completed, the multi-source sensing signals are synchronized and delayed by time synchronization and delay compensation based on the difference in the propagation speed of light and sound waves, and the signals from different channels are re-aligned to a unified timing sequence. The synchronized and compensated multi-source sensing data is mapped onto a unified spatiotemporal recognition framework to form a unified sensing basis for multi-target recognition.

3. The data processing method for a smart firing range for light weapons shooting training according to claim 2, characterized in that, In the step of mapping the synchronized and compensated multi-source sensing data to a unified spatiotemporal recognition framework, the spatial distribution relationship of the data collected by the optical array and the acoustic sensing array is jointly calibrated, and the spatiotemporal coordinates of each sensing point are matched accordingly based on the nanosecond-level time baseline.

4. The data processing method for a smart firing range for light weapons shooting training according to claim 1, characterized in that, The steps for removing interference features from multi-source sensing signals based on a unified sensing base include: The signals from each sensing device are reviewed in time and space, and continuous sensing trajectories are extracted to mark potential high-energy interference areas. Identify anomalous components within the marked high-energy interference region and determine their time-frequency distribution, duration, and spatial extension characteristics. The identified anomalous components are subjected to targeted energy suppression and multidimensional feature stripping, and real target features with consistency in temporal rhythm, spatial distribution and frequency characteristics are preserved according to the temporal and spatial mapping of the unified sensing base surface. The remaining features after suppression and stripping are aggregated according to their temporal evolution trajectory and spatial response location to form independent feature clusters corresponding to each target.

5. The data processing method for a smart firing range for light weapons shooting training according to claim 4, characterized in that, In the steps of implementing directional energy suppression and multidimensional feature stripping for anomalous components, the temporal reference and spatial mapping boundary of the unified sensing datum are used as references to perform directional suppression of the energy peak of the interference components, and transition processing is performed in the boundary region to avoid erroneous stripping of real target features or interference residue.

6. The data processing method for a smart firing range for light weapons shooting training according to claim 1, characterized in that, The steps for constructing a counterfactual playback chain based on the recognition results of the spatiotemporal coupled identification network include: Extract the time series response and spatial boundary evolution path of each target in the target recognition results output by the spatiotemporal coupling recognition network to form a set of recognition trajectories with reversible reconstruction capability; Based on the set of identified trajectories, a counterfactual playback chain is constructed to perform time reversal and replay of the identified path, and compare it with the original identification results to identify abnormal nodes and boundary drift. Based on the abnormal nodes and drift segments identified during the inversion process, the identification results are smoothed at the boundary and corrected for misjudgments, and then aligned consistently by combining the steady-state reference of independent feature clusters. A dynamic confidence map is constructed based on the stable performance of the corrected identification path during the time inversion process, and the identification confidence level is recorded with time as the horizontal axis and target identifier as the vertical axis.

7. The data processing method for a smart firing range for light weapons shooting training according to claim 6, characterized in that, The steps for implementing phase conjugate dynamic control based on the time-series information of dynamic reliability maps include: Extract the temporal information of each target identification path in the dynamic credibility map, analyze the trend of credibility level change and construct the target rhythm change profile, and identify rhythm drift and response anomaly segments; Based on the rhythm change profile, a rhythm traction operation is performed on the rhythm abnormal section in the identification path to adjust the time density and displacement velocity of the abnormal response node to the range of the rhythm reference curve. For trajectory segments that still exhibit spatial boundary diffusion or target fusion after rhythmic traction, inverse correction is implemented by combining historical records of the dynamic credibility map, so that the target boundary is recovered along the historical path and tends to converge. After completing rhythmic traction and reverse diffusion correction, the recognition response range of each target is subjected to amplitude limiting and write-back processing, and the response amplitude and spatial boundary threshold are set to form the target tracking output structure.

8. A smart firing range data processing system for light weapons shooting training, used to implement the smart firing range data processing method for light weapons shooting training as described in any one of claims 1-7, characterized in that, It includes a multi-source sensing synchronous acquisition module, an interference feature stripping and processing module, a spatiotemporal coupling identification and enhancement module, a counterfactual playback correction module, and a phase conjugate dynamic control module; The multi-source sensing synchronous acquisition module constructs a nanosecond-level unified time baseline, uses a high-resolution optical array and acoustic sensing matrix to synchronously acquire multi-source sensing signals from the shooting training range, and performs time synchronization and delay compensation for each sensing signal through a phase alignment mechanism to form a unified sensing baseline for multi-target recognition. The interference feature stripping module, based on a unified sensing baseline, performs interference feature stripping on the acquired multi-source sensing signals, establishes an interference feature stripping model, performs time-frequency decomposition and energy suppression on pseudo-feature components in complex backgrounds, and extracts independent feature clusters corresponding to each target. The spatiotemporal coupling identification enhancement module constructs a spatiotemporal coupling identification network based on independent feature clusters, maps independent feature clusters into dynamic convolutional tensors, and enhances the discriminativeness of target boundaries through a phase weight self-calibration mechanism, thereby achieving accurate identification of each target in multi-target interference scenarios and outputting identification results containing target boundaries and temporal information. The counterfactual replay correction module constructs a counterfactual replay chain based on the recognition results of the spatiotemporal coupling recognition network, performs time reversal replay of the recognition path, detects and corrects misjudgments and boundary drifts in the recognition results, and generates a dynamic credibility map that reflects temporal consistency. The phase conjugate dynamic control module, based on the time-series information of the dynamic credibility map, implements phase conjugate dynamic control to perform rhythmic traction, reverse diffusion correction, and amplitude limiting write-back on the tracking signal during the target recognition process, thereby achieving closed-loop control of recognition and correction.

Images