Multi-mode sensing and disturbing integrated anti-unmanned aerial vehicle system

Abstract

The invention belongs to the technical field of anti-unmanned aerial vehicles, and discloses a multi-mode sensing, disturbing and destroying integrated anti-unmanned aerial vehicle system. Comprising an environment data acquisition preprocessing module, a multi-modal feature fusion module, a protocol fingerprint identification module, an identity intention identification module, a strategy control module, an efficiency evaluation and confrontation knowledge extraction module, a strategy dynamic optimization and re-confrontation module, and an intelligent terminal control verification module for generating multi-modal cross verification. When a target is confirmed to be subdued or a task is finished, all collected data are subjected to final fusion judgment, a group of structured reports with multi-modal intelligent terminal control verification parameters are generated, and enhanced target files, confrontation instruction sets, response data and efficiency evaluation results generated in the whole confrontation process are systematically associated and stored, so that the target is controlled and verified. And a reusable confrontation strategy knowledge base is continuously accumulated, and an authoritative and auditable effect verification certificate is provided for each confrontation action.

Description

Technical Field

[0001] This invention relates to the field of anti-drone technology, and more specifically, to a multimodal sensing and disruption integrated anti-drone system. Background Technology

[0002] Counter-drone technology is a comprehensive technical field encompassing the detection, identification, tracking, and disposal of unauthorized and potentially threatening drones. With the increasing prevalence and performance improvements of consumer and industrial drone technologies, the security risks they pose, such as intrusion into sensitive airspace, spying on key areas, illegal transportation, and swarm attacks, are becoming increasingly prominent. Therefore, developing efficient, reliable, and environmentally adaptable counter-drone systems has become an urgent need in defense, public safety, and critical infrastructure protection.

[0003] Existing anti-drone systems suffer from significant deficiencies in detection and identification technologies. These deficiencies stem from their reliance on single or limited sensor modes. Radar is highly susceptible to ground clutter when dealing with low-altitude, slow-moving, and small targets, leading to frequent missed detections. Radio spectrum detection is completely ineffective against silent or autonomously navigating drones and exhibits a high false alarm rate in complex electromagnetic environments. Optoelectronic systems (visible / infrared) are severely limited by visibility and lighting conditions, with performance plummeting at night and in rainy or foggy weather, and they struggle to accurately distinguish between drones and aerial objects such as birds. Acoustic detection has an extremely short range, and environmental noise interference limits its practical value.

[0004] In view of this, the present invention proposes a multimodal sensing and disruption integrated anti-drone system to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a multimodal sensing and jamming integrated anti-drone system, comprising: the multimodal sensing and jamming integrated anti-drone system comprising: The environmental data acquisition and preprocessing module is used to acquire visible light and infrared images of the corresponding airspace background based on radio frequency sensors at different spatial nodes, and to acquire temperature, humidity and air pressure data through meteorological units. All environmental data are normalized and finally output as a standardized multi-source data stream composed of time, space and frequency domains. The multimodal feature fusion module is used to extract spectral features such as signal strength, modulation type, and frequency modulation pattern from the standardized multi-source data streams acquired and transmitted, and filter potential threat targets with the preset rule base in the multimodal feature fusion module to generate a target tracking profile containing preliminary orientation, motion vector, and multimodal feature vectors. The protocol fingerprinting module is used to compare subtle features of the radio frequency feature vectors carried in the target tracking file to generate the communication protocol model used by the target and even the possible individual identity. The identity and intent recognition module is used to analyze the visual features and motion trajectory patterns of the generated target. The model comprehensively judges its UAV platform type, possible payload, and tactical intent, and generates an enhanced target profile with a high-confidence identity label and intent assessment. The strategy control module is used to receive the enhanced target profile and call the dynamic interference strategy generation algorithm to it. Combining the real-time electromagnetic environment map and system resource status, it calculates and generates a customized set of countermeasures instructions. The core of this set of instructions is to achieve stealth countermeasures that "minimize self-exposure". All sensing sensors are instructed to keep a close watch on the target and record its original multimodal response data after being disturbed. The effectiveness assessment and adversarial knowledge extraction module is used to evaluate the raw data of multimodal response, initiate the quantitative analysis process, calculate a comprehensive real-time adversarial effectiveness score, determine whether the interference effect is a successful takeover, partial disruption or failure, and form structured adversarial knowledge fragments through the determination. The strategy dynamic optimization and re-confrontation module associates the confrontation knowledge fragments with the enhanced target file to optimize the strategy knowledge base within the strategy control module. That is, it determines whether a new round of confrontation needs to be initiated immediately. If so, the strategy control unit will adjust the original strategy parameters or generate a completely new interference mode based on the new knowledge in a very short time. A multimodal cross-validation intelligent terminal control verification module is generated. When the target is confirmed to be controlled or the task ends, all data will be collected for final fusion and judgment, and a structured report with multimodal intelligent terminal control verification parameters will be generated.

[0006] Furthermore, standardizing multi-source data streams includes: The radio frequency sensor array, photoelectric sensor, and acoustic array deployed in the defense area are activated simultaneously to capture raw wireless signals; the photoelectric sensor collects visible light and infrared video streams at a rate of 30 frames per second, the acoustic array records ambient sound at a sampling rate of 48kHz, a shared high-precision clock source gives all data a uniform timestamp, and the spatial coordinates of each sensor are known. The captured raw wireless signal is sent to the central processing unit, which performs a fast Fourier transform on the wireless signal to generate a spectrum and uses a blind source separation algorithm to distinguish different signal sources. The photoelectric sensor sends visible light and infrared video streams to the central processing unit. The central processing unit applies a background subtraction algorithm to each frame of the image, detects moving pixel regions, and calculates the speed, direction, and trajectory of each moving region. The acoustic array transmits the recorded ambient sound waveforms to the central processing unit. The central processing unit applies a set of bandpass filters to the sound waveforms, covering a range from 200Hz to 12kHz, and calculates the acoustic energy spectrum. The central processing unit performs spatiotemporal alignment of the spectrograms, moving target trajectories, and voiceprint energy spectra within the same time period, marking them as candidate targets. It compares the known characteristics of civil aviation broadcast signals, cellular base station signals, and typical bird voiceprints with the characteristics of candidate targets, eliminates matching candidate targets, and creates a candidate target file for each candidate target that cannot be eliminated. The collection of all candidate target files is output as a "standardized multi-source data stream".

[0007] Furthermore, the constructed target tracking profile includes: The target tracking profile includes a protocol fingerprinting mechanism, a visual voiceprint determination mechanism, and a feature fusion and intent determination logic mechanism. The protocol fingerprinting mechanism demodulates the signal and analyzes the protocol frame structure, extracting macroscopic features such as frame length, frame interval, and synchronization header waveform, as well as hardware-related subtle features such as signal modulation error vector amplitude and carrier frequency offset. It then compares these features with the UAV protocol fingerprint library built into the model. The fingerprint library contains publicly available protocol standards from various manufacturers and non-public protocol features obtained through reverse engineering, generating an identity confidence score based on feature similarity. The visual judgment mechanism includes a target classifier based on a convolutional neural network and a voiceprint template library. The image is input into the target classifier, and the target classifier outputs the type recognition result of the target in the image. At the same time, Mel frequency cepstral coefficient features are extracted from the voiceprint segments, and the feature vector is matched with a voiceprint template library. The matching result gives the closest voiceprint template type and matching score. The target classifier has been trained on a large number of negative sample images of drones, birds, kites, etc. The voiceprint template library stores the typical noise spectrum features of different models of drones and their propellers. The feature fusion and intent determination logic mechanism is used to cross-validate the radio frequency protocol model output by the "protocol fingerprint recognition model", the platform type output by the visual classifier, and the noise type output by the voiceprint matching. If all three point to the same type of drone platform and the confidence level is higher than the preset threshold, the identity is successfully confirmed and a unique "identity identifier" is generated. Otherwise, the radio frequency protocol model is used as the main basis, with visual and voiceprint as auxiliary evidence, and the overall confidence level is reduced accordingly. Finally, an enhanced target identity profile is output for each candidate target.

[0008] Furthermore, the strategy control module includes a dynamic interference strategy generation algorithm and an integrated disruption platform: The enhanced target identity profile is received from the core analysis layer. The profile information, including the identity identifier, UAV platform model, intent tag, and currently observed communication status field, is read from the profile. The dynamic interference strategy generation algorithm is then invoked, with the profile information as the main input. The dynamic jamming strategy generation algorithm queries the preset strategy rule base and combines it with real-time updated electromagnetic environment monitoring data to calculate and determine specific jamming parameters. The jamming parameters, including jamming type, frequency, power, waveform, modulation content and timing, are encapsulated into a structured dynamic countermeasure instruction set. After the integrated jamming platform receives the dynamic countermeasure instruction set, the multi-functional radio frequency front-end in the integrated jamming platform loads the corresponding waveform file according to the instruction set, aligns the transmitting antenna beam with the latest location of the target in the enhanced target identity file, and transmits jamming signals at the specified power and frequency at the precise time specified in the instruction set. At the same time as the interference signal is transmitted, the strategy control module sends a signal to all sensing sensors to switch to tracking mode, and marks the continuously recorded raw observation data with timestamps and target identifiers to form multimodal response raw data.

[0009] Furthermore, the performance evaluation and adversarial knowledge extraction module includes a performance evaluation mechanism and an adversarial knowledge extraction mechanism: The performance evaluation mechanism will activate the radio frequency domain analysis thread, the visual domain analysis thread, and the acoustic domain analysis thread to analyze, fuse, and evaluate the raw multimodal response data, and determine the final interference effect category based on the specific combination of observation items; The adversarial knowledge extraction mechanism generates adversarial knowledge fragments from the input performance evaluation.

[0010] Furthermore, the strategy dynamic optimization and re-countermeasure module includes: a strategy adjustment system execution knowledge fusion and rule update mechanism; By receiving adversarial knowledge fragments from the performance evaluation mechanism, the strategy adjustment system first parses the adversarial knowledge fragments and extracts the identity identifiers, the summary of the dynamic adversarial instruction set executed, and the description of the target's behavioral response pattern. Based on the identity identifier, the system retrieves and associates the corresponding enhanced target identity profile and historical adversarial records stored in the knowledge base. The strategy adjustment system compares the current adversarial knowledge fragments with the historical records to analyze whether the target's behavioral reaction pattern is a newly emerging pattern or a re-verification of an existing pattern. If the adversarial knowledge fragment reveals a new and effective target anti-interference behavior pattern, the strategy adjustment system will add or update this pattern as a new response rule to the entry in the strategy rule base associated with the corresponding UAV platform model and interference type. Based on the received updated information, the policy control unit calls the dynamic interference policy generation algorithm again. The dynamic interference policy generation algorithm will regenerate an optimized dynamic adversarial instruction set based on the policy rule base containing the latest adversarial knowledge.

[0011] Furthermore, a multimodal cross-validation intelligent terminal control verification module is generated, including: When the preset mission termination conditions are met, such as the target leaving the defense zone, the target being determined to be permanently disabled, or a manual termination command being received, the final adjudication process is initiated. The final adjudication process first sends a data freeze command to the entire data processing link to stop the real-time interference and evaluation loop on the current target, and gathers complete closed-loop adversarial data. Retrieve and associate the full-process data records of the current target from the storage, and compare the radio frequency protocol identity output by the "protocol fingerprint recognition model", the platform physical model output by the visual classifier, and the noise type output by the voiceprint matching again through multimodal cross-validation logic to confirm whether these three identity features remain consistent throughout the entire adversarial process, or whether there have been feature changes due to deception or disguise, thereby determining its anti-interference capability level. Based on the cross-validation results and the final adversarial effect, the final adjudication logic generates multimodal intelligent terminal control verification parameters, and encapsulates the multimodal intelligent terminal control verification parameters and key process summaries into a structured final verification report.

[0012] The technical effects and advantages of the multimodal sensing and disruption integrated anti-drone system of this invention are as follows: 1. This invention integrates heterogeneous data from multiple sources, including radio frequency, optoelectronic, and acoustic data, and innovatively introduces a protocol fingerprinting mechanism. It not only extracts the macroscopic frame features of communication signals but also deeply analyzes subtle hardware-level features such as modulation errors and carrier offsets. Combined with cross-verification of visual and acoustic signatures, it achieves a high degree of confidence in determining the model of the UAV platform and even the identity of the individual. This capability effectively overcomes the problems of traditional single sensors being easily deceived and having a high false alarm rate. It achieves a leap from target detection to target identification, providing a precise intelligence foundation for subsequent targeted countermeasures. 2. Furthermore, by constructing a real-time closed-loop countermeasure loop, the strategy control module does not execute a fixed jamming program. Instead, it dynamically generates a customized jamming instruction set that minimizes its own exposure based on the target's identity and intent, combined with the real-time electromagnetic environment. After the jamming is implemented, the system synchronously monitors the target's response through multimodal sensors and quantifies the jamming effect through the performance evaluation module. The resulting structured countermeasure knowledge fragments are fed back to the strategy library in real time, driving the strategy to dynamically optimize and re-countermeasure within seconds. This closed-loop capability of learning, adapting, and evolving significantly improves the continuous countermeasure effectiveness against new and adaptive threats. 3. At the end of the mission, the system initiates the final decision process of multimodal cross-validation, reviews the data of the entire process, verifies the consistency of the target's identity characteristics in the confrontation, and evaluates its anti-interference level. Finally, it generates a structured report with multimodal verification parameters. This not only provides authoritative and auditable effect verification evidence for each confrontation action, but more importantly, the enhanced target profiles, confrontation instruction sets, response data and effectiveness evaluation results generated throughout the confrontation process are systematically linked and stored, and continuously accumulated into a reusable confrontation strategy knowledge base. Attached Figure Description

[0013] Figure 1 This is a system schematic diagram of the multimodal sensing and disruption integrated anti-drone system of the present invention; Figure 2 This is a schematic diagram of the process for obtaining and generating candidate targets according to the present invention; Figure 3 This is a schematic diagram of the process for obtaining and outputting enhanced target files according to the present invention; Figure 4 This is a schematic diagram of the overall data recording process for task termination in this invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example 1 Please see Figure 1 Figure 4 As shown, the multimodal sensing and disruption integrated anti-drone system described in this embodiment has the following main design contents, including: The environmental data acquisition and preprocessing module is used to acquire visible light and infrared images of the corresponding airspace background based on radio frequency sensors at different spatial nodes, and to acquire temperature, humidity and air pressure data through meteorological units. All environmental data are normalized and finally output as a standardized multi-source data stream composed of time, space and frequency domains. The multimodal feature fusion module is used to extract spectral features such as signal strength, modulation type, and frequency modulation pattern from the standardized multi-source data streams acquired and transmitted, and filter potential threat targets with the preset rule base in the multimodal feature fusion module to generate a target tracking profile containing preliminary orientation, motion vector, and multimodal feature vectors. The protocol fingerprinting module is used to compare subtle features of the radio frequency feature vectors carried in the target tracking file to generate the communication protocol model used by the target and even the possible individual identity. The identity and intent recognition module is used to analyze the visual features and motion trajectory patterns of the generated target. The model comprehensively judges its UAV platform type, possible payload, and tactical intent, and generates an enhanced target profile with a high-confidence identity label and intent assessment. The strategy control module is used to receive the enhanced target profile and call the dynamic interference strategy generation algorithm to it. Combining the real-time electromagnetic environment map and system resource status, it calculates and generates a customized set of countermeasures instructions. The core of this set of instructions is to achieve stealth countermeasures that "minimize self-exposure". All sensing sensors are instructed to keep a close watch on the target and record its original multimodal response data after being disturbed. The effectiveness assessment and adversarial knowledge extraction module is used to evaluate the raw data of multimodal response, initiate the quantitative analysis process, calculate a comprehensive real-time adversarial effectiveness score, determine whether the interference effect is a successful takeover, partial disruption or failure, and form structured adversarial knowledge fragments through the determination. The strategy dynamic optimization and re-confrontation module associates the confrontation knowledge fragments with the enhanced target file to optimize the strategy knowledge base within the strategy control module. That is, it determines whether a new round of confrontation needs to be initiated immediately. If so, the strategy control unit will adjust the original strategy parameters or generate a completely new interference mode based on the new knowledge in a very short time. A multimodal cross-validation intelligent terminal control verification module is generated. When the target is confirmed to be subdued or the task ends, all data will be collected for final fusion and judgment, and a set of structured reports with multimodal intelligent terminal control verification parameters will be generated. Standardized multi-source data streams, including: The radio frequency sensor array, photoelectric sensor, and acoustic array deployed in the defense area are activated simultaneously to capture raw wireless signals; the photoelectric sensor collects visible light and infrared video streams at a rate of 30 frames per second, the acoustic array records ambient sound at a sampling rate of 48kHz, a shared high-precision clock source gives all data a uniform timestamp, and the spatial coordinates of each sensor are known. The captured raw wireless signal is sent to the central processing unit, which performs a fast Fourier transform on the wireless signal to generate a spectrum and uses a blind source separation algorithm to distinguish different signal sources. The photoelectric sensor sends visible light and infrared video streams to the central processing unit. The central processing unit applies a background subtraction algorithm to each frame of the image, detects moving pixel regions, and calculates the speed, direction, and trajectory of each moving region. The acoustic array transmits the recorded ambient sound waveforms to the central processing unit. The central processing unit applies a set of bandpass filters to the sound waveforms, covering a range from 200Hz to 12kHz, and calculates the acoustic energy spectrum. The central processing unit performs spatiotemporal alignment of the spectrograms, moving target trajectories, and voiceprint energy spectra within the same time period, marks them as candidate targets, compares the known characteristics of civil aviation broadcast signals, cellular base station signals, and typical bird voiceprints with the characteristics of candidate targets, eliminates matching candidate targets, and creates a candidate target file for each candidate target that cannot be eliminated. The collection of all candidate target files is output as a standardized multi-source data stream. The constructed target tracking profile includes: The target tracking profile includes a protocol fingerprinting mechanism, a visual voiceprint determination mechanism, and a feature fusion and intent determination logic mechanism. The protocol fingerprinting mechanism demodulates the signal and analyzes the protocol frame structure, extracting macroscopic features such as frame length, frame interval, and synchronization header waveform, as well as hardware-related subtle features such as signal modulation error vector amplitude and carrier frequency offset. It then compares these features with the UAV protocol fingerprint library built into the model. The fingerprint library contains publicly available protocol standards from various manufacturers and non-public protocol features obtained through reverse engineering, generating an identity confidence score based on feature similarity. The visual judgment mechanism includes a target classifier based on a convolutional neural network and a voiceprint template library. The image is input into the target classifier, and the target classifier outputs the type recognition result of the target in the image. At the same time, Mel frequency cepstral coefficient features are extracted from the voiceprint segments, and the feature vector is matched with a voiceprint template library. The matching result gives the closest voiceprint template type and matching score. The target classifier has been trained on a large number of negative sample images of drones, birds, kites, etc. The voiceprint template library stores the typical noise spectrum features of different models of drones and their propellers. The feature fusion and intent determination logic mechanism is used to cross-validate the radio frequency protocol model output by the protocol fingerprint recognition model, the platform type output by the visual classifier, and the noise type output by the voiceprint matching. If all three point to the same type of drone platform and the confidence level is higher than the preset threshold, the identity is successfully confirmed and a unique identity is generated. Otherwise, the radio frequency protocol model is used as the main basis, with visual and voiceprint as auxiliary evidence, and the overall confidence level is reduced accordingly. Finally, an enhanced target identity profile is output for each candidate target. The profile must include the following fields: a unique identifier based on the fusion and rights confirmation results, the confirmed UAV platform model, the comprehensive confidence score, the current observed communication status (e.g., active, silent), and a preliminary intent label based on motion trajectory analysis. The strategy control module includes a dynamic interference strategy generation algorithm and an integrated disruption platform. The enhanced target identity profile is received from the core analysis layer. The profile information, including the identity identifier, UAV platform model, intent tag, and currently observed communication status field, is read from the profile. The dynamic interference strategy generation algorithm is then invoked, with the profile information as the main input. The dynamic jamming strategy generation algorithm queries a preset strategy rule base and combines it with real-time updated electromagnetic environment monitoring data. The strategy rule base is a mapping table that maps different UAV platform models, intent tags, and communication status combinations to a basic jamming template. The basic jamming template defines the physical layer type of jamming and calculates to determine specific jamming parameters. The jamming parameters, including jamming type, frequency, power, waveform, modulation content, and timing, are encapsulated into a structured dynamic countermeasure instruction set. After the integrated jamming platform receives the dynamic countermeasure instruction set, the multi-functional radio frequency front-end in the integrated jamming platform loads the corresponding waveform file according to the instruction set, aligns the transmitting antenna beam with the latest location of the target in the enhanced target identity file, and transmits jamming signals at the specified power and frequency at the precise time specified in the instruction set. At the same time as the interference signal is transmitted, the strategy control module sends a signal to all sensing sensors to switch to tracking mode, and marks the continuously recorded raw observation data with timestamps and target identifiers to form multimodal response raw data; The performance evaluation and adversarial knowledge extraction module includes a performance evaluation mechanism and an adversarial knowledge extraction mechanism. The performance evaluation mechanism will activate the radio frequency domain analysis thread, the visual domain analysis thread, and the acoustic domain analysis thread to analyze, fuse, and evaluate the raw multimodal response data, and determine the final interference effect category based on the specific combination of observation items; The adversarial knowledge extraction mechanism generates adversarial knowledge fragments from the input performance evaluation. An adversarial knowledge fragment is a structured record whose content must explicitly include: the identity identifier in the enhanced target identity profile that triggered this adversarial action, a summary of the dynamic adversarial instruction set executed, and most importantly, a description of the target's behavioral response pattern derived from the analysis of the original multimodal response data. The strategy dynamic optimization and re-countermeasure module includes: a strategy adjustment system execution knowledge fusion and rule update mechanism; By receiving adversarial knowledge fragments from the performance evaluation mechanism, the strategy adjustment system first parses the adversarial knowledge fragments and extracts the identity identifiers, the summary of the dynamic adversarial instruction set executed, and the description of the target's behavioral response pattern. Based on the identity identifier, the system retrieves and associates the corresponding enhanced target identity profile and historical adversarial records stored in the knowledge base. The strategy adjustment system compares the current adversarial knowledge fragments with the historical records to analyze whether the target's behavioral reaction pattern is a newly emerging pattern or a re-verification of an existing pattern. If the adversarial knowledge fragment reveals a new and effective target anti-interference behavior pattern, the strategy adjustment system will add or update this pattern as a new response rule to the entry in the strategy rule base associated with the corresponding UAV platform model and interference type. Based on the received update information, the strategy control unit calls the "dynamic interference strategy generation algorithm" again. The dynamic interference strategy generation algorithm will regenerate an optimized dynamic adversarial instruction set based on the strategy rule base containing the latest adversarial knowledge. A smart terminal control verification module for generating multimodal cross-validation includes: When the preset mission termination conditions are met, such as the target leaving the defense zone, the target being determined to be permanently disabled, or a manual termination command being received, the final adjudication process is initiated. The final adjudication process first sends a data freeze command to the entire data processing link to stop the real-time interference and evaluation loop on the current target, and gathers complete closed-loop adversarial data. Retrieve and associate the full-process data records of the current target from the storage, and compare the radio frequency protocol identity output by the "protocol fingerprint recognition model", the platform physical model output by the visual classifier, and the noise type output by the voiceprint matching again through multimodal cross-validation logic to confirm whether these three identity features remain consistent throughout the entire adversarial process, or whether there have been feature changes due to deception or disguise, thereby determining its anti-interference capability level. Based on the cross-validation results and the final adversarial effect, the final adjudication logic generates multimodal intelligent terminal control verification parameters, and encapsulates the multimodal intelligent terminal control verification parameters and key process summaries into a structured final verification report.

[0016] In this embodiment, the present invention establishes an initial digital twin entity based on real materials and geometric properties, ensuring a high-fidelity virtual image that can dynamically reflect the real aging state of power devices under a specific usage history, thus laying a near-realistic simulation foundation for subsequent evaluation. By performing reverse calibration on the digital twin entity based on the BMS historical operation log, parameters that are prone to physical aging, such as the thermal conductivity of the solder layer, are corrected, reducing the prediction accuracy deviation caused by ignoring the service aging of power devices, and significantly improving the fidelity of the digital twin entity to the real physical state. By constructing a closed-loop iterative mechanism that includes virtual damage attacks and physical compliance verification, an adversarial boundary extreme condition sequence is generated using a genetic algorithm within the feasible space of operating parameters. This automatically detects extreme combinations of operating conditions that are difficult to cover in traditional standard tests and are most likely to induce device failure, effectively avoiding blind spots in safety design caused by incomplete operating condition coverage. By integrating full life-cycle reliability indicators and cost-efficiency evaluation indicators, and by constructing a comprehensive evaluation index system that includes multiple dimensions such as failure count, safety margin, and cost ratio, the system maximizes engineering efficiency while ensuring reliability under extreme operating conditions and taking into account economic cost and space volume. This provides quantitative and traceable decision support for the scientific selection of power devices for new energy vehicles.

[0017] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0018] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0019] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0020] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal sensing and disruption integrated anti-drone system, characterized in that, The multimodal sensing and disruption integrated anti-drone system includes: The environmental data acquisition and preprocessing module is used to acquire visible light and infrared images of the corresponding airspace background based on radio frequency sensors at different spatial nodes, and to acquire temperature, humidity and air pressure data through meteorological units. All environmental data are normalized and finally output as a standardized multi-source data stream composed of time, space and frequency domains. The multimodal feature fusion module is used to extract spectral features such as signal strength, modulation type, and frequency modulation pattern from the standardized multi-source data streams acquired and transmitted, and filter potential threat targets with the preset rule base in the multimodal feature fusion module to generate a target tracking profile containing preliminary orientation, motion vector, and multimodal feature vectors. The protocol fingerprinting module is used to compare subtle features of the radio frequency feature vectors carried in the target tracking file to generate the communication protocol model used by the target and even the possible individual identity. The identity and intent recognition module is used to analyze the visual features and motion trajectory patterns of the generated target. The model comprehensively judges its UAV platform type, possible payload, and tactical intent, and generates an enhanced target profile with a high-confidence identity label and intent assessment. The strategy control module is used to receive the enhanced target file and call the dynamic interference strategy generation algorithm to it. Combining the real-time electromagnetic environment map and system resource status, it calculates and generates a customized set of countermeasures instructions. The core of this set of instructions is to achieve stealth countermeasures that "minimize self-exposure". All sensing sensors are instructed to keep a close watch on the target and record its original multimodal response data after being disturbed. The effectiveness assessment and adversarial knowledge extraction module is used to evaluate the raw data of multimodal response, initiate the quantitative analysis process, calculate a comprehensive real-time adversarial effectiveness score, determine whether the interference effect is a successful takeover, partial disruption or failure, and form structured adversarial knowledge fragments through the determination. The strategy dynamic optimization and re-confrontation module associates the confrontation knowledge fragments with the enhanced target file to optimize the strategy knowledge base within the strategy control module. That is, it determines whether a new round of confrontation needs to be initiated immediately. If so, the strategy control unit will adjust the original strategy parameters or generate a completely new interference mode based on the new knowledge in a very short time. A multimodal cross-validation intelligent terminal control verification module is generated. When the target is confirmed to be controlled or the task ends, all data will be collected for final fusion and judgment, and a structured report with multimodal intelligent terminal control verification parameters will be generated.

2. The multimodal sensing and disruption integrated anti-drone system according to claim 1, characterized in that, The standardized multi-source data stream includes: The radio frequency sensor array, photoelectric sensor, and acoustic array deployed in the defense area are activated simultaneously to capture raw wireless signals; the photoelectric sensor collects visible light and infrared video streams at a rate of 30 frames per second, the acoustic array records ambient sound at a sampling rate of 48kHz, a shared high-precision clock source gives all data a uniform timestamp, and the spatial coordinates of each sensor are known. The captured raw wireless signal is sent to the central processing unit, which performs a fast Fourier transform on the wireless signal to generate a spectrum and uses a blind source separation algorithm to distinguish different signal sources. The photoelectric sensor sends visible light and infrared video streams to the central processing unit. The central processing unit applies a background subtraction algorithm to each frame of the image, detects moving pixel regions, and calculates the speed, direction, and trajectory of each moving region. The acoustic array transmits the recorded ambient sound waveforms to the central processing unit. The central processing unit applies a set of bandpass filters to the sound waveforms, covering a range from 200Hz to 12kHz, and calculates the acoustic energy spectrum. The central processing unit performs spatiotemporal alignment of the spectrograms, moving target trajectories, and voiceprint energy spectra within the same time period, marking them as candidate targets. It compares the known characteristics of civil aviation broadcast signals, cellular base station signals, and typical bird voiceprints with the characteristics of candidate targets, eliminates matching candidate targets, and creates a candidate target file for each candidate target that cannot be eliminated. The collection of all candidate target files is output as a standardized multi-source data stream.

3. The multimodal sensing and disruption integrated anti-drone system according to claim 2, characterized in that, The constructed target tracking profile includes: The target tracking profile includes a protocol fingerprinting mechanism, a visual voiceprint determination mechanism, and a feature fusion and intent determination logic mechanism. The protocol fingerprinting mechanism demodulates the signal and analyzes the protocol frame structure, extracting macroscopic features such as frame length, frame interval, and synchronization header waveform, as well as hardware-related subtle features such as signal modulation error vector amplitude and carrier frequency offset. It then compares these features with the UAV protocol fingerprint library built into the model. The fingerprint library contains publicly available protocol standards from various manufacturers and non-public protocol features obtained through reverse engineering, generating an identity confidence score based on feature similarity. The visual judgment mechanism includes a target classifier based on a convolutional neural network and a voiceprint template library. The image is input into the target classifier, and the target classifier outputs the type recognition result of the target in the image. At the same time, Mel frequency cepstral coefficient features are extracted from the voiceprint segments, and the feature vector is matched with a voiceprint template library. The matching result gives the closest voiceprint template type and matching score. The target classifier has been trained on a large number of negative sample images of drones, birds, kites, etc. The voiceprint template library stores the typical noise spectrum features of different models of drones and their propellers. The feature fusion and intent determination logic mechanism is used to cross-validate the radio frequency protocol model output by the protocol fingerprint recognition model, the platform type output by the visual classifier, and the noise type output by the voiceprint matching. If all three point to the same type of drone platform and the confidence level is higher than the preset threshold, the identity is successfully confirmed and a unique identity is generated. Otherwise, the radio frequency protocol model is used as the main basis, with visual and voiceprint as auxiliary evidence, and the overall confidence level is reduced accordingly. Finally, an enhanced target identity profile is output for each candidate target.

4. The multimodal sensing and disruption integrated anti-drone system according to claim 3, characterized in that, The strategy control module includes a dynamic interference strategy generation algorithm and an integrated disruption platform: The enhanced target identity profile is received from the core analysis layer. The profile information, including the identity identifier, UAV platform model, intent tag, and currently observed communication status field, is read from the profile. The dynamic interference strategy generation algorithm is then invoked, with the profile information as the main input. The dynamic jamming strategy generation algorithm queries the preset strategy rule base and combines it with real-time updated electromagnetic environment monitoring data to calculate and determine specific jamming parameters. The jamming parameters, including jamming type, frequency, power, waveform, modulation content and timing, are encapsulated into a structured dynamic countermeasure instruction set. After the integrated jamming platform receives the dynamic countermeasure instruction set, the multi-functional radio frequency front-end in the integrated jamming platform loads the corresponding waveform file according to the instruction set, aligns the transmitting antenna beam with the latest location of the target in the enhanced target identity file, and transmits jamming signals at the specified power and frequency at the precise time specified in the instruction set. At the same time as the interference signal is transmitted, the strategy control module sends a signal to all sensing sensors to switch to tracking mode, and marks the continuously recorded raw observation data with timestamps and target identifiers to form multimodal response raw data.

5. The multimodal sensing and disruption integrated anti-drone system according to claim 4, characterized in that, The performance evaluation and adversarial knowledge extraction module includes a performance evaluation mechanism and an adversarial knowledge extraction mechanism: The performance evaluation mechanism will activate the radio frequency domain analysis thread, the visual domain analysis thread, and the acoustic domain analysis thread to analyze, fuse, and evaluate the raw multimodal response data, and determine the final interference effect category based on the specific combination of observation items; The adversarial knowledge extraction mechanism generates adversarial knowledge fragments from the input performance evaluation.

6. The multimodal sensing and disruption integrated anti-drone system according to claim 5, characterized in that, The strategy dynamic optimization and re-countermeasure module includes: a strategy adjustment system execution knowledge fusion and rule update mechanism; By receiving adversarial knowledge fragments from the performance evaluation mechanism, the strategy adjustment system first parses the adversarial knowledge fragments and extracts the identity identifiers, the summary of the dynamic adversarial instruction set executed, and the description of the target's behavioral response pattern. Based on the identity identifier, the system retrieves and associates the corresponding enhanced target identity profile and historical adversarial records stored in the knowledge base. The strategy adjustment system compares the current adversarial knowledge fragments with the historical records to analyze whether the target's behavioral reaction pattern is a newly emerging pattern or a re-verification of an existing pattern. If the adversarial knowledge fragment reveals a new and effective target anti-interference behavior pattern, the strategy adjustment system will add or update this pattern as a new response rule to the entry in the strategy rule base associated with the corresponding UAV platform model and interference type. Based on the received update information, the policy control unit calls the "dynamic interference policy generation algorithm" again. The dynamic interference policy generation algorithm will regenerate an optimized dynamic adversarial instruction set based on the policy rule base containing the latest adversarial knowledge.

7. The multimodal sensing and disruption integrated anti-drone system according to claim 6, characterized in that, The intelligent terminal control verification module for generating multimodal cross-validation includes: When the preset mission termination conditions are met, such as the target leaving the defense zone, the target being determined to be permanently disabled, or a manual termination command being received, the final adjudication process is initiated. The final adjudication process first sends a data freeze command to the entire data processing link to stop the real-time interference and evaluation loop on the current target, and gathers complete closed-loop adversarial data. Retrieve and associate the full-process data records of the current target from the storage, and compare the radio frequency protocol identity output by the "protocol fingerprint recognition model", the platform physical model output by the visual classifier, and the noise type output by the voiceprint matching through multimodal cross-validation logic to confirm whether these three identity features remain consistent throughout the entire adversarial process, or whether there have been feature changes due to deception or disguise, thereby determining its anti-interference capability level. Based on the cross-validation results and the final adversarial effect, the final adjudication logic generates multimodal intelligent terminal control verification parameters, and encapsulates the multimodal intelligent terminal control verification parameters and key process summaries into a structured final verification report.

Images