Intelligent simulation weapon target damage judgment method and system based on multi-source fusion

Abstract

The invention belongs to the technical field of intelligent training instruments, and particularly relates to an intelligent simulation weapon target damage judgment method and system based on multi-source fusion, and the method comprises the steps: enabling an intelligent simulation weapon to collect attack data through an integrated inertial measurement unit, an AI visual sensor and a contact sensor, an attack action and a target key are recognized in real time at a weapon end by using an AI model, and an attack event is generated; meanwhile, a coding signal is transmitted to the target training harness through an infrared transmitter; the weapon and the training harness report an attack event and an attacked event to the background guidance and control evaluation system respectively, and the background guidance and control evaluation system confirms attack validity and calculates a damage value based on a damage model to realize damage judgment. Through multi-source data fusion and closed-loop verification, objective and accurate quantification and real-time feedback of the near-warfare confrontation behaviors are achieved, and the authenticity and effectiveness of simulation training are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent training equipment, and in particular relates to a method and system for determining target damage of intelligent simulated weapons based on multi-source fusion. Background Art

[0002] In traditional military simulation and close combat training, training equipment generally lacks intelligence. Trainees, especially those being attacked, often need to wear bulky protective gear that can compromise the authenticity of their movements. Existing training systems struggle to accurately quantify attack effects and damage effectiveness. Results often rely on the manual experience of referees or instructors, lacking objectivity and consistency. Furthermore, the training process often lacks immediate, data-driven feedback, which limits the accuracy of training evaluations and hinders effective guidance for trainees' improvement.

[0003] For example, some systems may only determine whether a hit has occurred based on simple contact, and are unable to distinguish the force and angle of the attack, whether the specific part of the attack is vital, and are unable to comprehensively consider the attacker's movement continuity and attack intention. Other systems that rely on pure visual solutions will have a significant impact on the accuracy and real-time performance of target recognition under fast movement or complex lighting conditions, and are prone to missed or misjudgment. Solutions that rely solely on inertial sensors have difficulty accurately identifying the target part of the attack. Therefore, how to effectively integrate information from multiple sensors to achieve accurate monitoring of the training process, fine-grained identification of attack movements, accurate determination of target vitals, and quantitative evaluation of damage effects, and provide real-time feedback, is a technical problem that needs to be solved in this field. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to address the deficiencies of the existing technology and provide an intelligent simulated weapon target damage determination method and system based on multi-source fusion, so as to overcome the problems in the existing technology such as low intelligence level of training equipment, insufficient evaluation quantification and accuracy, and lack of real-time and accurate feedback.

[0005] Technical solution: The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to the present invention comprises the following steps: S1: An intelligent simulated weapon that integrates an inertial measurement unit, an AI vision sensor, and a contact sensor collects weapon motion data, target-oriented visual data, and physical contact signals during the attack. S2: The main control unit of the intelligent simulated weapon performs the following processing: identifying the training action type based on the motion data, using a preset AI visual model to identify the predicted hit location in the visual data and the attack timestamp detected based on the physical contact signal, and generating an attack event data packet including the attack action type, the predicted hit location, and the attack timestamp; S3: The intelligent simulated weapon transmits a coded attack signal to the training equipment configured on the target party. After receiving the signal, the signal receiver of the training equipment on the target party generates an attack event data packet including the attacked part and the receiving timestamp; S4: The backend guidance and control evaluation system receives the attack event data packet and the attacked event data packet and performs collaborative verification; S5: When the collaborative verification passes, the background control and evaluation system calculates the target damage value based on the preset damage model and comprehensively considers the attack action type, predicted hit location and attack timestamp, and issues a damage instruction to the target training equipment to update its health status.

[0006] To further improve the above technical solution, a one-dimensional convolutional neural network model is used to process the three-axis acceleration and / or angular velocity time series data collected by the inertial measurement unit to identify the type of attack action; the action type includes chopping, slashing or stabbing; the identification is achieved through the following criteria: identifying the stabbing action by collecting single-peak acceleration pulses and low angular velocity characteristics; identifying the chopping action by collecting periodic acceleration oscillations and high angular velocity characteristics; and identifying the cutting action by collecting continuous median acceleration and sawtooth angular velocity characteristics.

[0007] Furthermore, by running a lightweight target detection model based on the YOLO series of algorithms, the images collected by the AI ​​vision sensor are processed to identify and locate one or more predicted hit parts on the target body in real time; and when it is detected that the image motion blur exceeds a preset threshold, the several frames of images captured before the blur occurs are reversely analyzed frame by frame based on the motion trajectory recorded by the inertial measurement unit to correct or confirm the actual target part hit at the moment of blur.

[0008] Furthermore, the preset damage determination model uses the following formula to calculate the target damage value D: Where: is the action energy coefficient, is the kinetic energy term, for anatomical weights; is the contact pressure; The target armor penetration threshold; This is the speed enhancement benchmark value. is the relative velocity at the time of contact.

[0009] Furthermore, the collaborative verification includes: comparing the attack timestamp of the attack event data packet with the receiving timestamp of the attacked event data packet to determine whether the time difference is within a preset range; and comparing the predicted hit part in the attack event data packet with the body part corresponding to the triggered signal receiver in the attacked event data packet to determine whether the two are spatially consistent.

[0010] Furthermore, a grip pressure sensor is integrated on the grip of the intelligent simulation weapon, and before calculating the damage value, the pressure timing curve collected by the grip pressure sensor is analyzed to identify the attack intention for weighted calculation of the damage value, wherein: a charged attack is determined by identifying the rise-steady-release characteristics in the pressure curve, and a rapid combo is determined by identifying the specified pulse interval and decay rate in the pressure curve.

[0011] Furthermore, it also includes a visual-inertial tightly coupled positioning step: combining the pre-integration result of the inertial measurement unit with the visual feature point tracking result of the visual sensor to correct the motion trajectory of the intelligent simulated weapon; and when the image blurriness collected by the AI ​​visual sensor exceeds a preset threshold, pure inertial measurement unit data is used to infer the motion trajectory within a preset time window.

[0012] Furthermore, it also includes an adaptive weight allocation step: according to the ambient light intensity value, the contribution weight of the visual data in determining the hit part is dynamically calculated by the following formula: , Where, is the visual weight, Lux is the current ambient light intensity, k is the attenuation coefficient, It is the center point of the light comfort zone; and adjust the inertial measurement unit weights accordingly , to rely more on inertial data under adverse visual conditions.

[0013] The intelligent simulated weapon target damage determination system for implementing the above method includes: An intelligent simulated weapon, equipped with an inertial measurement unit, an AI vision sensor, a contact sensor, an infrared transmitter, and a grip sensor, is used to collect weapon motion data, target-oriented visual data, and physical contact signals during the attack process, generate the attack event data packet, and transmit it via the infrared transmitter; The training device is equipped with an infrared receiver, which is in communication with the infrared transmitter of the intelligent simulated weapon and is used to receive attack event data packets and generate attacked event data packets; A background control and evaluation system is communicatively connected to the training equipment; wherein the background control and evaluation system is configured to perform the collaborative verification and damage value calculation.

[0014] Furthermore, the intelligent simulation weapon includes: a weapon body, a main control board arranged in the weapon body, an infrared transmitter for transmitting coded attack signals, an inertial measurement unit, an AI visual sensor, and a contact sensor connected to the main control board, for respectively collecting weapon motion data, target-oriented visual data, and physical contact signals during the attack process; wherein the main control unit is configured to: integrate data from the inertial measurement unit, AI visual sensor, and contact sensor to generate attack event data representing attack behavior, and control the intelligent simulation weapon to communicate with an external system.

[0015] The training equipment includes: an infrared receiving array distributed at vital parts of the body for receiving the coded attack signal; a positioning device for obtaining real-time geographic location information; a central processing unit for processing received data and managing health status; and a wireless communication module for communicating data with the background guidance and control evaluation system.

[0016] Beneficial effects: Compared with the prior art, the advantages of the present invention are: By integrating an inertial measurement unit, AI vision sensors, and contact sensors into intelligent simulated weapons, combined with a background damage model, this invention fundamentally addresses the technical challenge of traditional training equipment's inability to objectively and quantitatively evaluate combat maneuvers such as hacking, slashing, and stabbing. This transforms the subjective, fuzzy training process into a multi-dimensional data-driven, quantitative analysis accurate to every strike. Furthermore, damage commands can be issued in real time through a background system, providing trainees with immediate and accurate combat feedback, significantly enhancing the immersion and effectiveness of training.

[0017] This invention creatively integrates inertial data, visual data, contact signals, and infrared beacons in four dimensions. Compared to solutions relying on a single sensor, this invention effectively overcomes the recognition failures of purely visual solutions due to motion blur during rapid motion, through IMU-assisted visual recognition. Through an infrared collaborative verification mechanism, it fundamentally addresses the ambiguity inherent in pure algorithms in confirming physical hits at very close range or under complex occlusion. This cross-validation and complementarity of multi-source data enables the system to maintain stable operation in complex environments such as changing lighting and high-speed confrontations. Testing has shown an overall damage determination accuracy of at least 95%, demonstrating exceptionally high technical robustness.

[0018] The present invention deploys core AI algorithms such as motion recognition and target vital point recognition on the main control unit of the intelligent simulated weapon for edge computing. This design avoids the huge network delay caused by uploading large amounts of raw video data to the cloud for processing, ensuring a millisecond response from the occurrence of the action to the completion of recognition, meeting the strict real-time requirements of high-intensity, fast-paced close combat confrontation training, and providing trainees with a smooth experience without delay.

[0019] By integrating a grip pressure sensor on the grip, the present invention incorporates the factor of "attack intention" into the physical damage assessment model; the system can distinguish different tactical intentions such as "charged heavy blow" and "rapid combo", and use them as weighting factors for damage calculation, so that damage assessment is no longer limited to simple physical collisions, but is closer to the complex combination of strength and skills in actual combat, significantly improving the authenticity of training and the precision of assessment.

[0020] Through its unique multi-source data fusion architecture, edge AI processing capabilities, and closed-loop collaborative verification mechanism, this invention successfully solves the core pain points of existing technologies in training quantification, real-time feedback, judgment accuracy, and scenario adaptability, providing a set of technologically advanced and effective intelligent training and evaluation solutions for military, security, competitive sports, and even rehabilitation medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the appearance and structure of the intelligent simulation weapon of the present invention; Figure 2 is a cross-sectional view of the intelligent simulation weapon of the present invention; Figure 3 It is a schematic diagram of the decomposition of the intelligent simulation weapon in the present invention.

[0022] In the figure, 1-19 represent the following components respectively: 1 is the blade body; 2 is the blade body slot; 3 is the fixing plate; 4 is the blade body resistance switch; 5 is the AI ​​camera lens high-transmittance protection plate; 6 is the blade body movable shaft; 7 is the AI ​​camera lens; 8 is the connector; 9 is the sealing ring; 10 is the AI ​​camera core board; 11 is the grip; 12 is the anti-slip film; 13 and 16 are the grip pressure sensors; 14 is the main control board; 15 is the dagger grip cover; 17 is the lithium battery; 18 is the battery cover; 19 is the power switch. DETAILED DESCRIPTION

[0023] The technical solution of the present invention is described in detail below with reference to the accompanying drawings, but the protection scope of the present invention is not limited to the embodiments.

[0024] Example 1: Figures 1 to 3 As shown, intelligent simulation weapons provide the basic form of weapons (such as Figure 1The training device is in the form of a dagger, but the present invention is not limited to this form and can be other training weapons such as a stick). Figure 2 、 Figure 3 As shown, the blade body slot 2 is a movable setting for installing the blade body 1; the fixing plate 3 is used to connect the blade body 1 and the handle 11; the blade body contact switch 4 serves as a contact sensor for detecting physical contact with the target; the AI ​​camera lens high-transmittance protection plate 5 is used to protect the AI ​​camera lens; after the blade body 1 is installed, the blade body contact switch 4 can be triggered with the blade body movable shaft 6 as the fulcrum; the connecting piece 8 is used to fix the blade body contact switch 4; 9 is a sealing ring; the AI ​​camera core board 10 is connected to the AI ​​camera lens 7 and the main control board 14, and the main control board 14 serves as the core processor of the system. The dagger grip cover 15 is installed after fixing the main control board 14, and the lithium battery 17 is installed in the handle 11 for powering the main control board 14 and the AI ​​camera core board 10; a non-slip film 12 is provided on the handle 11; a grip pressure sensor 13 and a grip pressure sensor 16 are installed at the top and bottom of the handle 11 respectively; the power switch 19 is installed at the tail of the handle 11 and a battery cover 18 is installed.

[0025] Through the internal integrated sensors, the intelligent simulated weapon realizes the following functional modules, including: Inertial Measurement Unit (IMU): Located on the main control board 14, it is used to collect real-time acceleration and angular velocity data of the weapon in three-dimensional space. This data is the basis for identifying training movements such as chopping, slashing, and stabbing; AI visual sensor: Consists of an AI camera lens 7 and an AI camera core board 10, and is located at the front end of the weapon or at a suitable location to capture images or video streams of training targets; Contact sensor: This is the blade contact switch 4, which can be a piezoelectric sensor or other type of pressure / contact sensor. It is installed at the effective attack part of the weapon (such as the blade head or the end of the stick) to detect the precise moment when the weapon physically contacts the target and the contact pressure. Infrared transmitter: integrated into the front end of the weapon, used to transmit coded infrared signals to the target device; The grip sensor, i.e., the pressure sensors 13 and 16 integrated in the grip, is used to detect the way and strength with which the user holds the weapon, and can be used to assist in determining attack intention (such as whether it is a full-strength attack) or grip stability.

[0026] The main control board 14 (MCU / DSP) inside the intelligent simulation weapon is integrated with a processor and can also be configured to be connected to an external computer or server of the system to process data from various sensors, execute algorithms and decision logic.

[0027] To ensure the accuracy of data fusion, the processor implements a strict synchronization mechanism based on the multimodal sensor synchronization module (trigger circuit), including time synchronization and spatial synchronization.

[0028] Time synchronization: The processor has a built-in high-precision hardware clock. At startup, a hardware trigger circuit or a unified high-precision clock timing mechanism ensures that data from the IMU, AI vision sensor, and contact sensor are precisely aligned on the time base. For example, a unified timing signal from a wearable device can be received via wireless connection to achieve subsystem time synchronization.

[0029] Spatial synchronization: Mechanical design ensures that the AI ​​vision sensor 7, AI camera core board 10, infrared transmitter, and dagger main control board 14 are rigidly connected to the weapon body, maintaining a fixed relative position and orientation. During calibration, the center of the AI ​​vision field of view, the center of the infrared transmitter space, and the IMU's attack vector direction (e.g., the Z axis) are aligned to the same reference coordinate system (e.g., with the tip of the dagger as the origin), ensuring spatial consistency of multi-source data.

[0030] Example 2: This example provides a complete intelligent simulated weapon target damage determination system, which consists of three parts: training equipment, background guidance and control evaluation system, and the intelligent simulated weapon in Example 1. Through collaborative work, accurate damage determination and situation feedback are achieved.

[0031] Training gear is worn by both the attacker and the target, and is usually a vest or full-body suit. Its key components include: Infrared receiving array: Multiple infrared receivers are distributed in key parts of the body (such as the left / right chest, abdomen, head, and limbs) to receive attack signals from intelligent simulated weapons; Positioning device (such as GPS / Beidou): used to obtain the real-time geographic location of trainees; Central Processing Unit: Responsible for processing the equipment's data, managing health status, and interacting with the wireless communication module; Wireless communication module: As the hub of data interaction, it is responsible for short-range wireless connection with intelligent simulation weapons and long-distance data transmission with the background guidance, control and evaluation system.

[0032] Backstage guidance, control and evaluation system: Usually a software system deployed on a server, it is the "brain" of the entire training, responsible for receiving data from all trainees, making final damage judgments, battlefield situation calculations, and data recording and analysis.

[0033] The data flow and decision logic of the entire system follow a rigorous closed-loop process, ensuring the authority and real-time nature of the decisions: Before the training begins, each trainee's intelligent simulated weapon is linked to the training equipment they wear via short-range wireless communication; at the same time, all training equipment is connected to the background guidance, control and evaluation system through communication radios.

[0034] During training, each set of training gear regularly reports its status information to the backend guidance and evaluation system. This information includes at least the user's ID, current location coordinates, and current health status (for example, initially at 100%). The backend guidance and evaluation system uses this information to construct and update a battlefield situation map in real time.

[0035] When attacker A uses an intelligent simulated weapon to attack target B: Weapon-side data collection: Multi-source sensors on intelligent simulated weapons instantly collect attack data (IMU action, AI vision location, contact signal, grip strength); Infrared beacon transmission: The infrared transmitter at the front end of the intelligent simulated weapon transmits a coded infrared signal to target B. The signal contains attacker A's ID, weapon type, attack action type, predicted hit location, and precise timestamp; Local data transmission and integration: The intelligent simulation weapon sends the complete attack data packet to the training equipment worn by attacker A via short-range wireless. Data upload to the backend: After receiving the data, attacker A's training equipment immediately uploads the event data package containing all attack details to the backend guidance and control assessment system through its communication radio, requesting damage determination. The background control and assessment system receives the attack event data packet from the attacker A, combines more information such as the position and posture of both parties, and calculates the final damage value D to the target party B based on the damage model.

[0036] After the judgment is completed, the background guidance and control evaluation system will send the damage command (including the damage value D) to the training equipment of the target party B.

[0037] Target B's training device receives the damage command from the backend. Its built-in central processing unit updates its locally stored health value based on the received damage value D. After the health value changes, Target B's training device triggers immediate feedback and reports the updated health value to the backend guidance and control evaluation system in its next scheduled status report.

[0038] Example 3: A damage calculation and assessment method using the system provided in Example 2 includes: Step S1: Training action type recognition When a user trains with an intelligent simulated weapon, its built-in IMU collects real-time time-series data on three-axis acceleration and angular velocity during its motion. The processor receives this IMU data and processes it using a model based on a one-dimensional convolutional neural network (1D-CNN). The 1D-CNN model trains using a large amount of labeled IMU motion data. During the real-time recognition phase, the 1D-CNN model outputs a probability distribution for each pre-set training motion type corresponding to the current IMU data segment. The type with the highest probability is considered the currently executed motion.

[0039] Specific recognition criteria: The model accurately classifies actions by learning the signal characteristics of different actions. Its criteria include: stabbing: identifying single-peak pulse features in acceleration data, while the angular velocity value is relatively low; chopping: identifying periodic oscillation features in acceleration data, accompanied by high angular velocity values; cutting: identifying continuous medium-amplitude features in acceleration data, while the angular velocity presents a sawtooth waveform.

[0040] Motion feature extraction: def classify_action(imu_data): # Use 1D-CNN to process three-axis acceleration / angular velocity timing signals # Output probability distribution of stabbing / slashing / cutting actions return action_type Damage effect calculation: Damage = f(action_type, target_region, contact_pressure, relative_velocity) Step S2: Identification of target vital parts The AI ​​vision sensor on the intelligent simulated weapon continuously captures images of the target ahead. The AI ​​camera core board 10 runs a lightweight target detection model based on an improved YOLO series algorithm to identify pre-defined vital parts of the human body. This model is trained using a customized dataset targeting 14 or more vital parts of the human body (such as the eyes, throat, and heart). This dataset uses 133 key points from common datasets such as COCO-WholeBody to generate the 14 vital parts through mapping relationships.

[0041] Motion blur compensation module (dynamically adjusted based on IMU data): To cope with rapid movements, the system uses an IMU-assisted compensation mechanism. When the processor determines that the image blur exceeds a preset threshold by analyzing image clarity or IMU data, the system triggers the compensation logic: based on the motion trajectory accurately recorded by the IMU, the system performs a reverse frame-by-frame analysis of several frames of images captured before the blur occurs, thereby correcting or confirming the actual target part that was hit at the moment of blur.

[0042] Adaptive weight distribution: The system dynamically adjusts the contribution weights of vision and IMU data based on the ambient light intensity. The weight distribution follows the following formula: in, is the visual weight, Lux is the current ambient light intensity, k is the attenuation coefficient (typical value is 0.08), It is the center point of the light comfort zone (typical value is 50k Lux). When the light is too strong or too weak, Reduce the weight of IMU The accuracy is improved accordingly, making the system more dependent on IMU data under adverse visual conditions.

[0043] Step S3: Physical contact moment detection A contact sensor 4 (e.g., a piezoelectric blade contact sensor) mounted on the weapon's front end or effective striking surface generates a signal (e.g., a voltage pulse or pressure change) when the weapon makes physical contact with the target. The processor detects this signal and records the precise moment of physical contact, tc, which serves as a critical time anchor for subsequent damage assessment.

[0044] Step S4: Infrared data interaction and collaborative verification Attacker Channel: When an attack occurs, the attacker's intelligent simulated weapon transmits complete attack event data (including IMU, AI vision, timestamp, etc.) to the backend system via the radio on its training equipment. This represents the "attacker's claim."

[0045] Target Channel: Simultaneously, the weapon's infrared transmitter transmits a coded signal (including the attacker's ID and timestamp) to the target. Upon receiving this signal, the infrared receiver in the corresponding part of the target's training gear (e.g., the chest) immediately generates an "attacked" event (including the attacker's ID, the part hit, and the time of reception) and reports it to the backend system via its radio. This represents the "target's perception."

[0046] The backend guidance and assessment system receives reports from two different sources within a preset time window. At this point, the backend system begins to perform collaborative verification: Event correlation: The system correlates the attack events reported by the attacker and the attacked events reported by the target using the attacker ID and timestamp.

[0047] Logical comparison: Temporal consistency: Check whether the timestamps of the two events are within the preset error range (for example, the difference is less than 50ms); Spatial consistency: Check whether the "hit location" reported by the target (for example, the infrared receiver on the chest is triggered) roughly matches the hit location predicted by the attacker's AI vision; Causal consistency: Confirm that the attacker's ID is exactly the same as the attack source ID reported by the target.

[0048] Validity Confirmation: Only when all of the aforementioned comparison logic is met will the backend guidance and control evaluation system determine the attack as a "valid hit." This effectively filters out false infrared triggers due to environmental interference, AI vision misjudgments, or situations where the weapon moves but does not actually make contact due to excessive distance.

[0049] Step S5: Calculation of target damage value Grip intention analysis: Before calculation, the system will analyze the data from grip pressure sensors 13 and 16 to determine the attack intention. Among them, for a charged attack, the pressure curve shows a clear three-stage feature of "rising - stabilizing - releasing"; for a rapid combo, the pressure curve appears as a sequence with pulse intervals and possibly decreasing amplitude.

[0050] Damage model calculation: When the physical contact moment tc is detected and a valid attack is confirmed, the processor calls the preset damage judgment model, comprehensively considers the following factors and uses a mathematical model to calculate the damage value D: Kinematic characteristics (from S1): action type (the energy coefficients of stabbing, chopping, and slashing are different), speed at contact and accumulated kinetic energy; criticality of the attack (from S2 and S4): whether the hit part is a vital part (for example, the heart area has a higher weight than the limbs); target protection information (from S4): the protection level carried in the target's infrared signal code; attack intention (from the grip sensor): for example, "charged heavy attack" will obtain a higher damage bonus.

[0051] The final damage value D is calculated using the following core formula: Where: is the action energy coefficient, which is determined by the action type of S1 (thrust = 1.2, chop = 0.8, slice = 0.5); is the anatomical weight (heart region = 1.0, liver = 0.7, limbs = 0.3); is the contact pressure; The target armor penetration threshold (obtained through infrared communication); is the speed enhancement benchmark value (5 m / s, experimental calibration), is the relative velocity at the time of contact; is the kinetic energy term, , is the equivalent mass of the holding mode, which is dynamically calculated based on the holding posture: Where: The physical quality of the dagger; Penetration coefficient Depends on contact pressure Armor penetration threshold with the target : when Treated as full penetration (100% damage transferred).

[0052] The target damage value D finally calculated can be used for real-time display, recording training logs, updating the target's virtual health value, etc.

[0053] As above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. An intelligent simulation weapon target damage determination method based on multi-source fusion, characterized by: The following steps are involved: S1: An intelligent simulated weapon that integrates an inertial measurement unit, an AI vision sensor, and a contact sensor collects weapon motion data, target-oriented visual data, and physical contact signals during the attack. S2: The main control unit of the intelligent simulated weapon performs the following processing: identifying the training action type based on the motion data, using a preset AI visual model to identify the predicted hit location in the visual data and the attack timestamp detected based on the physical contact signal, and generating an attack event data packet including the attack action type, the predicted hit location, and the attack timestamp; S3: The intelligent simulated weapon transmits a coded attack signal to the training equipment configured on the target party. After receiving the signal, the signal receiver of the training equipment on the target party generates an attack event data packet including the attacked part and the receiving timestamp; S4: The backend guidance and control evaluation system receives the attack event data packet and the attacked event data packet and performs collaborative verification; S5: When the collaborative verification passes, the background control and evaluation system calculates the target damage value based on the preset damage model and comprehensively considers the attack action type, predicted hit location and attack timestamp, and issues a damage instruction to the target training equipment to update its health status.

2. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: A one-dimensional convolutional neural network model is used to process the three-axis acceleration and / or angular velocity time series data collected by the inertial measurement unit to identify the type of attack action; the action type includes chopping, slashing or stabbing; the identification is achieved through the following criteria: identifying the stabbing action by collecting single-peak acceleration pulses and low angular velocity characteristics; identifying the chopping action by collecting periodic acceleration oscillations and high angular velocity characteristics; and identifying the cutting action by collecting continuous median acceleration and sawtooth angular velocity characteristics.

3. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: By running a lightweight target detection model based on the YOLO series of algorithms, the images collected by the AI ​​vision sensor are processed to identify and locate one or more predicted hit parts on the target body in real time; and when it is detected that the image motion blur exceeds a preset threshold, the several frames of images captured before the blur occurs are reversely analyzed frame by frame based on the motion trajectory recorded by the inertial measurement unit to correct or confirm the actual target part hit at the moment of blur.

4. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: The preset damage determination model uses the following formula to calculate the target damage value D: Where: is the action energy coefficient, is the kinetic energy term, for anatomical weights; is the contact pressure; The target armor penetration threshold; This is the speed enhancement benchmark value. is the relative velocity at the time of contact.

5. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: The collaborative verification includes: comparing the attack timestamp of the attack event data packet with the reception timestamp of the attacked event data packet to determine whether the time difference is within a preset range; and comparing the predicted hit part in the attack event data packet with the body part corresponding to the triggered signal receiver in the attacked event data packet to determine whether the two are spatially consistent.

6. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: The grip of the intelligent simulation weapon is also integrated with a grip pressure sensor. Before calculating the damage value, the pressure timing curve collected by the grip pressure sensor is analyzed to identify the attack intention for weighted calculation of the damage value, wherein: a charged attack is determined by identifying the rise-steady-release characteristics in the pressure curve, and a rapid combo is determined by identifying the specified pulse interval and decay rate in the pressure curve.

7. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: It also includes a visual-inertial tightly coupled positioning step: combining the pre-integration result of the inertial measurement unit with the visual feature point tracking result of the visual sensor to correct the motion trajectory of the intelligent simulated weapon; and when the image blurriness collected by the AI ​​visual sensor exceeds a preset threshold, pure inertial measurement unit data is used to infer the motion trajectory within a preset time window.

8. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: It also includes an adaptive weight allocation step: according to the ambient light intensity value, the contribution weight of the visual data in determining the hit part is dynamically calculated by the following formula , Where, is the visual weight, Lux is the current ambient light intensity, k is the attenuation coefficient, It is the center point of the light comfort zone; and adjust the inertial measurement unit weights accordingly , to rely more on inertial data under adverse visual conditions.

9. An intelligent simulated weapon target damage determination system, characterized in that: include: An intelligent simulated weapon, equipped with an inertial measurement unit, an AI vision sensor, a contact sensor, an infrared transmitter, and a grip sensor, is used to collect weapon motion data, target-oriented visual data, and physical contact signals during the attack process, generate the attack event data packet, and transmit it via the infrared transmitter; The training device is equipped with an infrared receiver, which is in communication with the infrared transmitter of the intelligent simulated weapon and is used to receive attack event data packets and generate attacked event data packets; A background control and evaluation system is communicatively connected to the training equipment; wherein the background control and evaluation system is configured to perform the collaborative verification and damage value calculation.

10. The method for determining target damage of intelligent simulated weapons based on multi-source fusion according to claim 1 is characterized in that: The intelligent simulated weapon includes: a weapon body, a main control board disposed within the weapon body, an infrared transmitter for transmitting coded attack signals, an inertial measurement unit, an AI vision sensor, and a contact sensor connected to the main control board, for respectively collecting weapon motion data, target-oriented visual data, and physical contact signals during an attack; wherein the main control unit is configured to: integrate data from the inertial measurement unit, the AI ​​vision sensor, and the contact sensor to generate attack event data representing attack behavior, and control the intelligent simulated weapon to communicate with an external system; The training equipment includes: an infrared receiving array distributed at vital parts of the body for receiving the coded attack signal; a positioning device for obtaining real-time geographic location information; a central processing unit for processing received data and managing health status; and a wireless communication module for communicating data with the background guidance and control evaluation system.

Images