Construction method of unmanned aerial vehicle real-time data analysis model based on block chain

Abstract

The invention discloses an unmanned aerial vehicle real-time data analysis model construction method based on a block chain, and relates to the technical field of unmanned aerial vehicle communication, and the method comprises the steps: an unmanned aerial vehicle builds an ad-hoc mesh network through a 5G NR-U frequency band, a ground control station carries out the validity verification, generates an initial network topological graph, builds a private block chain network based on a Hyperledger Fabric framework, and generates a creation block; collecting video stream data, recording attitude data by an inertial measurement unit, and generating a compressed data packet; the unmanned aerial vehicle scans a fuselage metasurface pattern through a terahertz emitter to obtain a scattering feature matrix, generates a composite hash value and adds the composite hash value to the compressed data packet to form a to-be-transmitted data packet; and measuring the refractive index structure constant of the current area to obtain the turbulence intensity grade. According to the method, physical fingerprint features are obtained through terahertz scanning of a fuselage metasurface, an anti-interference hash value is generated in combination with space-time parameters, hardware-level identity authentication is achieved, and the risk of unmanned aerial vehicle identity cloning is fundamentally eradicated.

Description

Technical Field

[0001] The present invention relates to the field of drone communication technology, and in particular to a method for constructing a real-time data analysis model for drones based on blockchain. Background Art

[0002] With the rapid development of drone technology in recent years, to ensure efficient collaboration and data security for drone swarms in complex environments, the trusted storage and sharing of drone data has been achieved through the construction of distributed ledgers. This technology utilizes smart contracts to authenticate and authorize drone data and employs traditional erasure coding to ensure reliable data transmission. However, this solution lacks physical layer security at the data collection end. Drone authentication relies on traditional digital certificates, making it difficult to defend against hardware cloning attacks. The fixed error correction coding strategy used at the data transmission layer is unable to adapt to dynamic channel conditions such as atmospheric turbulence, making it difficult to ensure data integrity in high bit error rate scenarios.

[0003] Existing blockchain-based drone communication technology lacks a coordinated guarantee mechanism for the drone's physical identity and data transmission reliability, and its adaptive coding and real-time analysis capabilities in dynamic environments are insufficient. This invention mainly solves the problems of trusted networking, anti-interference transmission and real-time analysis of drone clusters in complex environments. By integrating metasurface terahertz fingerprint authentication, turbulence adaptive coding and edge intelligent analysis, it achieves full-stack security and efficiency optimization from the physical layer to the application layer. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for constructing a real-time data analysis model for drones based on blockchain, which solves the technical problem that drone physical identity authentication and dynamic channel adaptive transmission cannot be coordinated and optimized in the existing technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for constructing a blockchain-based real-time data analysis model for drones, which includes: establishing an ad-hoc mesh network using the 5G NR-U frequency band with a drone, performing legitimacy verification by a ground control station, generating an initial network topology, building a private blockchain network based on the Hyperledger Fabric framework, and generating a genesis block;

[0008] Collect video stream data, the inertial measurement unit records attitude data, and generates compressed data packets;

[0009] The drone uses a terahertz transmitter to scan the metasurface pattern on its fuselage to obtain a scattering feature matrix, and generates a composite hash value that is appended to the compressed data packet to form a data packet to be transmitted;

[0010] Measure the refractive index structure constant of the current area to obtain the turbulence intensity level, add the turbulence intensity level to the data packet to be transmitted, split the data packet into data segments according to the initial network topology map, and select relay nodes for routing and forwarding using the Kademlia algorithm;

[0011] Redundant encoding is performed on the data packets to be transmitted based on the turbulence intensity level to generate transmission units with error correction codes. The endorsing nodes of the private blockchain network verify the transmission units with error correction codes through the Byzantine fault-tolerant consensus mechanism, sort the service nodes in the private blockchain network to generate a new block, and transmit the verified transmission units with error correction codes to the new block.

[0012] The smart contract triggers the edge computing node to perform Reed-Solomon decoding on the transmission unit with the error correction code to restore it to the data packet to be transmitted, use the TensorFlow Lite inference engine to perform target detection analysis, and generate a format analysis report.

[0013] As a preferred solution of the method for constructing a blockchain-based drone real-time data analysis model described in the present invention, the drone establishes an ad-hoc mesh network through the 5G NR-U frequency band, the ground control station performs legitimacy verification, generates an initial network topology map, builds a private blockchain network based on the Hyperledger Fabric framework, and generates a genesis block, including the following steps:

[0014] The baseband chip scans the 5925-7125MHz frequency band, measures the frequency signal-to-noise ratio, and generates a spectrum analysis report. The ground control station receives the spectrum analysis report, executes the frequency algorithm to obtain the center frequency, and then derives the TDMA time slot allocation plan based on the center frequency to obtain the time slot configuration table.

[0015] The drone sends a beacon frame according to the time slot configuration table, establishes a connection node and generates a neighbor list. The ground control station sends a PUF challenge code to each drone to obtain a PUF feature vector set. The control station sends a maneuver command and generates a behavior authentication result through IMU data verification.

[0016] Combine the behavior authentication results with the PUF feature vector set, perform threshold judgment, and output a list of legal drones;

[0017] Based on the neighbor list and the legal drone list, the Dijkstra algorithm is used to generate the initial topology map;

[0018] According to the list of legal drones, use the cryptogen tool to generate a node certificate set, configure the node certificate set, and obtain the genesis block.

[0019] As a preferred solution of the method for constructing a real-time data analysis model of a UAV based on blockchain described in the present invention, wherein: collecting video stream data, an inertial measurement unit records attitude data, and generating a compressed data packet, the following steps are included:

[0020] The drone aligns the sensor clock through the GPS PPS signal to obtain a timestamp sequence, and dynamically adjusts the frame rate based on the timestamp sequence and IMU speed data to obtain video stream data;

[0021] The inertial measurement unit uses the timestamp sequence IMU velocity data to trigger differential encoding to obtain attitude data;

[0022] The video stream data and posture data are spliced ​​together to obtain a compressed data packet.

[0023] As a preferred solution of the method for constructing a real-time data analysis model of a UAV based on blockchain described in the present invention, the UAV scans the metasurface pattern of the fuselage through a terahertz transmitter to obtain a scattering feature matrix, generates a composite hash value and appends it to the compressed data packet to form a data packet to be transmitted, including the following steps:

[0024] The drone uses a 0.3THz transmitting module to scan the fuselage metasurface in a spiral trajectory to obtain raw scattered signal data. It then performs a short-time Fourier transform on the raw scattered signal data to obtain a time-frequency matrix.

[0025] The ResNet-18 model is trained based on the sample data set in the time-frequency matrix. The time-frequency matrix is ​​input into the trained ResNet-18 model to obtain the unique scattering fingerprint feature vector.

[0026] Read the drone's current grid code, fuselage temperature, and unique scattering fingerprint feature vector to obtain anti-interference data;

[0027] Perform SHA3-256 hash operation on the interference data to obtain a composite hash value;

[0028] The composite hash value is concatenated with the compressed data packet to obtain the data packet to be transmitted.

[0029] As a preferred solution of the method for constructing a blockchain-based drone real-time data analysis model described in the present invention, the method includes: measuring the refractive index structure constant of the current area to obtain the turbulence intensity level, adding the turbulence intensity level to the data packet to be transmitted, splitting the data packet to be transmitted into data segments according to the initial network topology map, and selecting relay nodes for routing and forwarding through the Kademlia algorithm, including the following steps:

[0030] By collecting and measuring temperature, humidity and air pressure in real time, the atmospheric data are summarized and the refractive index structure constant is calculated based on the atmospheric data;

[0031] The turbulence intensity level is obtained based on the current altitude of the UAV and the refractive index structure constant;

[0032] The turbulence intensity level is added to the header flag of the data packet to be transmitted to obtain a turbulence-marked data packet. The fragment size is calculated based on the initial network topology map and the turbulence-marked data packet, and the turbulence-marked data packet is split into a sequence of data fragments.

[0033] The Kademlia algorithm is used to select relay nodes from the initial network topology to obtain a node list, and the data slice sequence is sent to the target node according to the node list.

[0034] As a preferred solution of the method for constructing a blockchain-based drone real-time data analysis model described in the present invention, wherein: redundant encoding is performed on the data packets to be transmitted according to the turbulence intensity level to generate a transmission unit with an error correction code, and the endorsing node of the private blockchain network verifies the transmission unit with the error correction code through the Byzantine fault-tolerant consensus mechanism, including the following steps:

[0035] Selecting a Reed-Solomon coding scheme according to the turbulence intensity level to perform redundant coding on the data packet to be transmitted, obtaining a configuration instruction, performing byte block division on the compressed data packet to obtain a data fragment group of a sequence of data blocks to be encoded, encoding the data fragment group of the sequence of data blocks to be encoded to obtain an original coding unit, adding a header to the original coding unit, and generating a final transmission unit with an error correction code;

[0036] Perform a cyclic redundancy check on the transmission unit with the error correction code according to the endorsing node of the private blockchain network to obtain the initial verification state. Perform Reed-Solomon decoding on the initial verification state to obtain the decoding state. Check the difference between the original data shard timestamp and the current block height to obtain the time-space verification state. Verify the digital signature of the transmission unit with the error correction code in the Intel SGX enclave to obtain the security authentication state.

[0037] Generate local verification results based on the initial verification state, decoding state, time and space verification state and security authentication state;

[0038] The local verification results are verified through the Byzantine fault-tolerant consensus mechanism to obtain the verified transmission unit with error correction code.

[0039] As a preferred solution of the method for constructing a blockchain-based drone real-time data analysis model according to the present invention, the service nodes in the private blockchain network are sorted to generate a new block, and the verified transmission units with error correction codes are transmitted to the new block, including the following steps:

[0040] Based on the CPU load, network latency, and storage IOPS of the service nodes in the private blockchain network, a priority queue is generated. Verified transmission units with error correction codes are selected from the memory pool, sorted by reception timestamp, and a block structure template is generated. The block structure template is hashed with SHA3-256 to obtain a block hash value. The block hash value is then ECDSA-signed with the private key to obtain a block signature.

[0041] Combine the block structure template, block hash value and block signature to obtain a new block;

[0042] Through the gRPC protocol, the verified transmission unit with error correction code is delivered to the new block.

[0043] As a preferred solution of the method for constructing a blockchain-based drone real-time data analysis model described in the present invention, the smart contract triggers the edge computing node to perform Reed-Solomon decoding on the transmission unit with the error correction code to restore it to the data packet to be transmitted, uses the TensorFlow Lite inference engine to perform target detection analysis, and generates an analysis report, including the following steps:

[0044] The video stream data and posture data in the private blockchain network are counted through smart contracts to obtain dynamic thresholds;

[0045] The smart contract monitors and verifies that the transmission units with error correction codes have reached the dynamic threshold, sends a trigger instruction to the edge node, and obtains the transmission unit ID list;

[0046] According to the transmission unit ID list, the transmission unit with Reed-Solomon error correction code is obtained from IPFS and decoded to obtain the data packet to be transmitted;

[0047] Decode the data packets to be transmitted, input them into the TensorFlow Lite model, and generate an analysis report.

[0048] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for constructing a blockchain-based drone real-time data analysis model as described in the first aspect of the present invention is implemented.

[0049] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for constructing a blockchain-based drone real-time data analysis model as described in the first aspect of the present invention is implemented.

[0050] The beneficial effects of the present invention are: obtaining physical fingerprint features through terahertz scanning of the fuselage metasurface, generating anti-interference hash values ​​in combination with spatiotemporal parameters, realizing hardware-level identity authentication, fundamentally eliminating the risk of drone identity cloning, dynamically adjusting the RS coding scheme and routing strategy based on real-time turbulence intensity, accurately quantifying the channel state through the refractive index structure constant, improving coding efficiency, and ensuring the transmission success rate in a strong interference environment. Physical fingerprint authentication ensures the credibility of the data source, dynamic coding ensures the reliability of the transmission process, and together with the blockchain consensus mechanism, constructs a full-stack security system from the physical layer to the application layer. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 Flowchart of the method for building a blockchain-based drone real-time data analysis model.

[0053] Figure 2 This is a schematic diagram of the initial topology.

[0054] Figure 3 Schematic diagram of the analysis report.

[0055] Figure 4 Schematic diagram of data collection and processing. DETAILED DESCRIPTION

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0058] Second, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, characteristic, or combination of features and / or characteristics described herein that can be included in at least one implementation of the present application. The various appearances of "in one embodiment" or "an embodiment" are not necessarily all referring to the same embodiment, although they can. Rather, they are each referring to one of a possible number of alternative embodiments that can be implemented separately or in any combination.

[0059] Referring to Figures 1 to 4 For one embodiment of the present application, the embodiment provides a method for constructing a real-time data analysis model of a UAV based on a blockchain, comprising the following steps:

[0060] S1, the UAV establishes an ad-hoc mesh network through a 5G NR-U frequency band, the ground control station performs legality verification, generates an initial network topology graph, builds a private blockchain network based on a Hyperledger Fabric framework, and generates a genesis block.

[0061] S1.1, scan the 5925-7125MHz frequency band through the baseband chip, measure the signal-to-noise ratio of each frequency point to generate a spectrum analysis report, the ground control station receives the spectrum analysis report, executes a frequency point algorithm to obtain a center frequency point, and obtains a TDMA time slot allocation scheme according to the center frequency point to obtain a time slot configuration table.

[0062] Specifically, the expression is,

[0063]

[0064] Where f optimal is the center frequency point, P interf (i) is the interference power of the ith sending node, SNR(i) is the signal-to-interference-and-noise ratio of the ith sending node, CQI(f) is the channel quality score of the frequency point f, f is the frequency point index, b is the channel quality weight coefficient, i is the sending node index, and N is the total number of sending nodes.

[0065] It should be noted that the baseband chip scans the 5925-7125MHz frequency band, measures the signal-to-noise ratio of each frequency point, and generates a spectrum analysis report. After the ground control station receives the spectrum analysis report, it executes a frequency point selection algorithm to calculate the optimal center frequency point, for example, when the signal-to-noise ratio of the frequency point 6100MHz is the highest and the interference power is the lowest, it is selected as the center frequency point. According to the center frequency point, a time division multiple access time slot allocation scheme is calculated, and a time slot configuration table containing the mapping relationship between the time slot number and the UAV number is output, for example, when the center frequency point is 6100MHz, an 8-time slot cyclic allocation configuration table is generated. The spectrum analysis report contains the signal-to-noise ratio, interference power and channel quality score of each frequency point.

[0066] S1.2. The drone sends a beacon frame according to the time slot configuration table, establishes a connection node and generates a neighbor list. The ground control station sends a PUF challenge code to each drone to obtain a PUF feature vector set. The control station sends a maneuver command and generates a behavior authentication result through IMU data verification.

[0067] Furthermore, the drone sends a beacon frame containing the device ID and location information in the specified time slot according to the time slot configuration table, and the nodes that successfully establish a communication connection form a neighbor list; the ground control station sends a physical unclonable function challenge code to each drone, and the drone returns a hardware fingerprint response to generate a physical unclonable function feature vector set; the control station sends an instruction containing a preset maneuver trajectory, and the drone records the actual motion data through the inertial measurement unit, calculates the correlation coefficient between the preset trajectory and the actual trajectory, and generates a behavior authentication pass result when the correlation coefficient exceeds 0.9.

[0068] S1.3. Combine the behavior authentication results and the PUF feature vector set, perform threshold judgment, and output a list of legal drones.

[0069] Furthermore, the behavioral authentication results are combined with the physical unclonable function feature vector set for verification. When the motion trajectory correlation coefficient in the behavioral authentication result is greater than 0.9 and the detection distance of the physical unclonable function feature vector set is less than 0.15, it is determined to be a legal drone and output to the legal drone list. During the verification process, the motion trajectory correlation coefficient in the behavioral authentication result is calculated using the three-axis acceleration and angular velocity data collected by the inertial measurement unit, and the Hamming distance of the physical unclonable function feature vector set is compared with the pre-stored benchmark value to obtain the legal drone list.

[0070] S1.4. Based on the neighbor list and the legal drone list, the Dijkstra algorithm is used to generate the initial topology map.

[0071] Furthermore, based on the communication connection status in the neighbor list and the device verification information in the legal drone list, the Dijkstra algorithm is used to calculate the optimal path to generate an initial topology graph. The neighbor list contains the device IDs and signal strength data of drones that have established connections, and the legal drone list contains the device IDs of drones that have passed physical unclonable functions and behavioral authentication. The Dijkstra algorithm uses communication delay and signal strength as edge weights. When the signal strength between two nodes is greater than -85dBm and the communication delay is less than 20ms, a connection edge is established. The final output is an adjacency matrix representation containing all legal drone nodes and their optimal paths. In the initial topology graph, nodes represent verified drones, and edges represent connection paths that meet communication quality requirements. An adjacency matrix element value of 1 indicates a valid connection, and a value of 0 indicates no connection.

[0072] S1.5. Based on the list of legal drones, use the cryptogen tool to generate a node certificate set, configure the node certificate set, and obtain the genesis block.

[0073] Furthermore, based on the device ID and physical unclonable function feature vector in the legal drone list, the cryptogen tool generates a node certificate set containing an ECDSA key pair and an X.509 certificate. The node certificate set is written into the crypto-config.yaml file according to the configuration specifications of the Hyperledger Fabric framework, specifying the organizational relationship and certificate path of the Orderer node and the Peer node. The configuration is read by the configtxgen tool to generate a genesis block containing the MSP identity information and channel configuration.

[0074] S2. Collect video stream data, the inertial measurement unit records attitude data, and generates compressed data packets.

[0075] S2.1. The drone aligns the sensor clock through the GPS PPS signal to obtain a timestamp sequence. The frame rate is dynamically adjusted according to the timestamp sequence and IMU velocity data to obtain video stream data.

[0076] Furthermore, the drone receives the GPS PPS signal as a global clock reference, synchronizes the sampling clocks of the inertial measurement unit and camera through the PTP protocol, generates a timestamp sequence with microsecond accuracy, and dynamically calculates the target frame rate based on the interval distribution of the timestamp sequence and the three-dimensional velocity values ​​measured in real time by the inertial measurement unit. When the speed value is less than 5m / s, 30fps is selected, 60fps is selected for 5-15m / s, and 120fps is selected for speed values ​​greater than 15m / s. The camera collects video stream data according to the frame rate.

[0077] S2.2. The inertial measurement unit uses the timestamp sequence IMU velocity data to trigger differential encoding to obtain attitude data.

[0078] Furthermore, the inertial measurement unit triggers differential encoding based on the sampling time and inertial measurement unit velocity data recorded in the timestamp sequence. When the velocity change rate is less than 0.1g, the Euler angle data of adjacent timestamps are incrementally encoded and compressed attitude data is output. The timestamp sequence provides a synchronization reference with an accuracy of 10μs. The inertial measurement unit velocity data is obtained by fusion calculation of the three-axis accelerometer and gyroscope. The differential encoding process uses a first-order difference algorithm to quantize the difference between the attitude quaternion of the current moment and the previous moment into a 16-bit integer for storage. The attitude data includes a compressed representation of the roll angle, pitch angle and yaw angle. When the velocity change rate exceeds 0.1g, it switches to full-precision mode to record the original data. The compression algorithm achieves a 4:1 compression ratio by discarding high-order decimal places, and the continuous attitude is restored through linear interpolation during decompression.

[0079] S2.3. Concatenate the video stream data and the posture data to obtain a compressed data packet.

[0080] Furthermore, the H.265 encoded video stream data and the differentially compressed attitude data are aligned and spliced ​​according to the timestamps, and the attitude data at the corresponding moment is inserted into the header of each frame of the video stream data to form a composite data block; the composite data block is secondary compressed using the Zstandard algorithm, and the compression level is set to 9 to generate the final compressed data packet. The frame header of the video stream data reserves a 20-byte metadata area to store the timestamp index of the attitude data, and the attitude data uses little endian to store the 16-bit integer values ​​of the roll angle, pitch angle and yaw angle; Zstandard compression uses a dictionary (containing 100 groups of typical flight data patterns) to improve the compression ratio. The header of the compressed data packet contains a 4-byte synchronization identifier and a 2-byte length field, and a 4-byte CRC-32 check code is attached to the tail. The timestamp alignment uses the hardware clock synchronized by the PTP protocol, and the error is less than 1ms; the splicing of composite data blocks is achieved through DMA transmission to achieve zero-copy operation.

[0081] S3. The drone scans the metasurface pattern of the fuselage through the terahertz transmitter to obtain the scattering feature matrix, generates a composite hash value and appends it to the compressed data packet to form the data packet to be transmitted.

[0082] S3.1. The drone uses a 0.3THz transmitting module to scan the fuselage metasurface along a spiral trajectory to obtain the original scattered signal data. The short-time Fourier transform is performed on the original scattered signal data to obtain the time-frequency matrix.

[0083] Furthermore, the drone activates the 0.3THz transmission module, scans the fuselage metasurface structure according to a preset spiral path, emits linear frequency-modulated continuous waves and receives reflected signals, collects original scattered signal data containing amplitude and phase information, and performs short-time Fourier transform on the original scattered signal data with a 128ns window to generate a time-frequency matrix.

[0084] S3.2. Train the ResNet-18 model based on the sample data set in the time-frequency matrix, input the time-frequency matrix into the trained ResNet-18 model, and obtain the unique scattering fingerprint feature vector.

[0085] Furthermore, a dataset containing 10,000 sets of time-frequency matrix samples was used to perform transfer learning training on the ResNet-18 model. The input layer accepted a 128×128 complex time-frequency matrix, extracted features through a residual connection structure, and output a 128-dimensional unique scattering fingerprint feature vector. The Adam optimizer was used in the training process, and the loss function was a triplet loss. Training was stopped when the distance between the anchor sample and the positive sample was less than 0.2 and the distance between the anchor sample and the negative sample was greater than 0.8. In the inference stage, the time-frequency matrix collected in real time was input into the trained ResNet-18 model, and the fully connected layer output the normalized feature vector. The value range of each dimension of the unique scattering fingerprint feature vector was [0, 1]. The feature matching degree was calculated by cosine similarity, and the threshold was set to 0.95 to determine the validity.

[0086] S3.3. Read the current grid code, fuselage temperature, and unique scattering fingerprint feature vector of the drone to obtain anti-interference data.

[0087] Furthermore, the 10cm precision grid code converted from the drone's current GPS coordinates and the actual measured value of the fuselage temperature sensor are read and spliced ​​with the 128-dimensional unique scattering fingerprint feature vector output by the ResNet-18 model; the grid code is converted into a 20-bit binary coordinate identifier, and the fuselage temperature is quantized into an 8-bit unsigned integer (resolution 0.5°C), and input into the SHA3-256 hash operation together with the unique scattering fingerprint feature vector to generate anti-interference data.

[0088] S3.4. Perform a SHA3-256 hash operation on the anti-interference data to obtain a composite hash value.

[0089] Specifically, the expression is,

[0090] PUF hash =SHA3 256 (v feature ⊕v GPS ⊕T amb );

[0091] Among them, PUF hash is the composite hash value, v feature is the unique scattering fingerprint feature vector, v GPS is the current longitude and latitude, T amb为 Drone temperature.

[0092] S3.5. Concatenate the composite hash value with the compressed data packet to obtain the data packet to be transmitted.

[0093] Further, the SHA3-256 algorithm is used to splice the composite hash value with the Zstandard compressed data packet, the composite hash value is written into the authentication field of the packet header, the compressed data packet body is stored continuously from the 0x20 address, the packet header is attached with a 4-byte synchronization identifier and a 2-byte length field, the tail is appended with a 4-byte CRC-32 check code, and a complete data packet to be transmitted is formed.

[0094] S4. Measuring the refractive index structure constant of the current area to obtain a turbulence intensity level, adding the turbulence intensity level to the data packet to be transmitted, splitting the data packet to be transmitted into data segments according to the initial network topology, and selecting a relay node for routing and forwarding through the Kademlia algorithm.

[0095] S4.1, real-time collection and measurement of temperature, humidity and air pressure, and atmospheric data is summarized, and the refractive index structure constant is calculated according to the atmospheric data.

[0096] Specifically, the expression is,

[0097]

[0098] wherein, is the structure constant of the refractive index m, P is the atmospheric pressure, T is the temperature, is the temperature differential, is the infinitesimal change in height, is the specific humidity differential, and m is the refractive index.

[0099] S4.2, according to the current height of the unmanned aerial vehicle and the refractive index structure constant, the turbulence intensity level is obtained.

[0100] Further, based on the current altitude measured by the barometric altimeter and the refractive index structure constant, the turbulence intensity level is determined by table lookup method, when the altitude is lower than 100 meters and the refractive index structure constant is less than 5*10^-16 m^(-2 / 3), it is determined as L1 level (weak turbulence), the altitude is 100-500 meters and the refractive index structure constant is between 5*10^-16 and 2*10^-15 m^(-2 / 3) is L2 level (moderate turbulence), the altitude is higher than 500 meters or the refractive index structure constant is greater than 2*10^-15 m^(-2 / 3) is L3 level (strong turbulence).

[0101] S4.3, adding the turbulence intensity level to the packet header flag of the data packet to be transmitted to obtain a turbulence marked data packet, calculating the fragmentation size according to the initial network topology and the turbulence marked data packet, and splitting the turbulence marked data packet into a data segment sequence.

[0102] Specifically, the expression is,

[0103]

[0104] Among them, S frag is the fragment size, MTU is the maximum length of a single frame, Turb level is the turbulence level.

[0105] Furthermore, the fragment size is dynamically determined based on the intensity level identifier (L1 / L2 / L3) in the turbulence mark data packet header: L1 level uses 1024-byte fragments, L2 level uses 768-byte fragments, and L3 level uses 512-byte fragments; the data packet body is divided into continuous data blocks according to the fragment size, and a 6-byte header is attached to each data block (including a 2-byte fragment sequence number, a 2-byte total number of fragments, and a 2-byte turbulence mark) to generate an ordered data fragment sequence.

[0106] S4.4. Select relay nodes from the initial network topology using the Kademlia algorithm to obtain a node list, and send the data slice sequence to the target node according to the node list.

[0107] Specifically, the expression is,

[0108]

[0109] Among them, D ij is the normalized comprehensive distance between the sending node i and the target node j, ID i is the unique identity of node i, ID j is the unique identity of the target node j, SNR ij is the channel quality of the signal sent from the sending node i to the target node j, XOR is the bitwise exclusive OR, α is the logical distance weight, β is the turbulence intensity weight, and γ is the signal-to-noise ratio weight.

[0110] S5. Redundantly encode the data packets to be transmitted according to the turbulence intensity level to generate transmission units with error correction codes. The endorsing nodes of the private blockchain network verify the transmission units with error correction codes through the Byzantine fault-tolerant consensus mechanism.

[0111] S5.1. Select the Reed-Solomon coding scheme according to the turbulence intensity level to perform redundant encoding on the data packet to be transmitted to obtain the configuration instruction, divide the compressed data packet into byte blocks to obtain the data block sequence data fragment group to be encoded, encode the data block sequence data fragment group to obtain the original coding unit, add a packet header to the original coding unit, and generate the final transmission unit with the error correction code.

[0112] Furthermore, the corresponding Reed-Solomon coding parameters are selected according to the turbulence intensity level. L1 level adopts Reed-Solomon coding, L2 level adopts Reed-Solomon coding, and L3 level adopts Reed-Solomon coding parameters. The data packet to be transmitted is divided into 1024-byte blocks to generate a sequence of data blocks to be encoded. A 4-byte header (including a 2-byte fragment ID and a 2-byte length identifier) ​​is added to each data block. Redundant encoding is performed on the sequence of data blocks to be encoded through the Reed-Solomon encoder to generate an original coding unit containing check bytes. The number of check bytes is determined by the coding scheme. An 8-byte header (including a 4-byte synchronization code and a 4-byte timestamp) is appended to the head of the original coding unit, and a 4-byte CRC-32 check code is added to the tail to generate a final transmission unit with an error correction code.

[0113] S5.2. Perform a cyclic redundancy check on the transmission unit with the error correction code according to the endorsement node of the private blockchain network to obtain an initial verification state. Perform Reed-Solomon decoding on the initial verification state to obtain a decoding state. Check the difference between the original data shard timestamp and the current block height to obtain a time-space verification state. Verify the digital signature of the transmission unit with the error correction code within the Intel SGX enclave to obtain a security authentication state.

[0114] Furthermore, the endorsement node of the private blockchain network performs a CRC-32 check on the received transmission unit with the error correction code. The check range covers all data from the packet header to the check byte. When the checksum matches, an initial check pass status is generated. The Reed-Solomon code block is extracted from the transmission unit that passes the initial check, and Galois field decoding is performed according to the encoding parameters recorded in the packet header. A decoding pass status is generated when the original data block is successfully restored. The timestamp is extracted from the decoded data shard header and compared with the time window corresponding to the current blockchain height. A time-space check pass status is generated when the timestamp is within the valid range. The preset ECDSA public key is used in the Intel SGX enclave to verify the digital signature at the end of the transmission unit. A security authentication pass status is generated when the signature is valid and the certificate has not been revoked.

[0115] S5.3. Generate a local verification result based on the initial verification state, the decoding state, the time-space verification state, and the security authentication state.

[0116] Furthermore, a four-state AND logic is used to generate local verification results: when the cyclic redundancy check status is passed, the Reed-Solomon decoding status is data complete recovery, the timestamp deviation of the time-space check status is within ±3 seconds, and the digital signature verification of the security authentication status is passed, the local verification result is marked as valid; if any state fails, it is marked as invalid. The cyclic redundancy check status is determined by the CRC-32 checksum matching, the Reed-Solomon decoding status depends on whether the original data block can be successfully recovered, the time-space check status is generated by comparing the data shard timestamp with the highly corresponding time window of the blockchain, and the security authentication status is determined by the ECDSA signature verification result in the Intel SGX enclave.

[0117] S5.4. Verify the local verification result through the Byzantine fault-tolerant consensus mechanism to obtain the verified transmission unit with error correction code.

[0118] Furthermore, the endorsing nodes of the private blockchain network collect at least 2f+1 valid local verification results (f is the maximum number of fault-tolerant nodes) and conduct multiple rounds of voting verification through the Byzantine fault-tolerant consensus mechanism: in the first round, the master node broadcasts a proposal containing the local verification results, and the replica nodes compare the cyclic redundancy check status, Reed-Solomon decoding status, time-space verification status and security authentication status in the proposal to see if they are consistent with their own verification. After receiving f+1 matching preliminary responses, the submission phase begins, and finally 2f+1 consistently confirmed transmission units with error correction codes are marked as verified.

[0119] S6. Sort the service nodes in the private blockchain network to generate a new block, and transmit the verified transmission units with error correction codes to the new block.

[0120] S6.1. Generate a priority queue based on the CPU load, network latency, and storage IOPS of the service nodes in the private blockchain network. Select verified transmission units with error correction codes from the memory pool, sort them by reception timestamp, generate a block structure template, perform SHA3-256 hashing on the block structure template to obtain a block hash value, and use the private key to perform ECDSA signing on the block hash value to obtain a block signature.

[0121] Furthermore, the service nodes of the private blockchain network monitor CPU utilization, end-to-end network latency, and storage IOPS indicators in real time, calculate node priority and generate a sorted queue through a weighted scoring algorithm (CPU weight 0.4, latency weight 0.3, IOPS weight 0.3), extract transmission units with error correction codes that have been verified by the practical Byzantine fault tolerance consensus from the memory pool, and arrange them in ascending order according to the receiving timestamp synchronized with the IEEE 1588 precision time protocol. Generate a block structure block template containing a block header (parent hash, timestamp, Merkle root) and a transaction list, perform a SHA3-256 hash operation on the block structure block template to generate a 256-bit block hash value, and use the secp256k1 private key in the node certificate set to ECDSA sign the block hash value, outputting a 64-byte block signature.

[0122] S6.2 combines the block structure block template, block hash value and block signature to obtain a new block.

[0123] Furthermore, the block structure block template, block hash value and block signature are combined in a predefined format: the block structure block template is used as the main content (including the block header and transaction list), the block hash value is written into the hash field of the block header, and the block signature is attached to the end of the block; the combined data structure is broadcast to all network nodes through the gRPC protocol to form a new block.

[0124] S6.3 transmits the verified transmission unit with error correction code to the new block through the gRPC protocol.

[0125] Furthermore, the transmission units with error correction codes that have been verified by the practical Byzantine fault-tolerant consensus are sent to the generated new block through the streaming transmission channel of the gRPC protocol; the transmission process adopts HTTP / 2 multiplexing technology, and each transmission unit is encapsulated as an independent gRPC message frame (including a 4-byte length prefix and payload data), and the message header is attached with a block height identifier and a target node ID. The transmission units with error correction codes are sorted by the receiving timestamp and transmitted in batches. Each batch contains a maximum of 50 transmission units, and the maximum length of a single frame does not exceed 16KB; the receiver confirms the integrity of the data by checking the CRC-32 checksum in the message frame header, and writes the transaction list of the new block after successful reception. The gRPC connection maintains a long connection state, and the flow control adopts the token bucket algorithm (rate limit 10Mbps). The error retransmission mechanism is implemented based on sequence number confirmation, and the binding relationship between the transmission unit and the new block is verified by the Merkle root hash in the block header.

[0126] S7. The smart contract triggers the edge computing node to perform Reed-Solomon decoding on the transmission unit with the error correction code to restore it to the data packet to be transmitted. The TensorFlow Lite inference engine is used to perform target detection analysis and generate an analysis report.

[0127] S7.1. Use smart contracts to collect statistics on video stream data and posture data in the private blockchain network and obtain dynamic thresholds.

[0128] Specifically, the expression is,

[0129] Thresh old=max(100MB,0.2×AvgLast10Blocks);

[0130] Among them, Thresh old is the dynamic threshold, and vgLast10Blocks is the average value of 10 blocks in the private blockchain network.

[0131] S7.2. The smart contract monitors and verifies that the number of transmission units with error correction codes reaches the dynamic threshold, sends a trigger instruction to the edge node, and obtains a list of transmission unit IDs.

[0132] Furthermore, the smart contract counts the number of transmission units with error correction codes that have passed on-chain verification in real time. When the cumulative value reaches the dynamic threshold, a trigger instruction containing a list of transmission unit IDs is sent to the edge node. The transmission unit ID list is generated by the smart contract by traversing the memory pool. Each ID is a 32-byte SHA3-256 hash value. The list is sorted in ascending order by the timestamp when the transmission unit is on the chain. The dynamic threshold is calculated based on the block data volume recorded in the Hyperledger Fabric state database, and the statistical window is the average of the most recent 10 blocks; the trigger instruction is sent to the gRPC stream channel subscribed by the edge node through the event push mechanism. The instruction payload is encoded in Protocol Buffers. After receiving the instruction, the edge node retrieves the corresponding transmission unit data with error correction code from the IPFS distributed storage according to the transmission unit ID list.

[0133] S7.3. Obtain the transmission unit with the Reed-Solomon error correction code from IPFS according to the transmission unit ID list and perform decoding to obtain the data packet to be transmitted.

[0134] Further, the edge node retrieves the corresponding transmission unit with Reed-Solomon error correction code through the IPFS distributed storage network according to the 32-byte SHA3-256 hash value in the transmission unit ID list; extracts the Reed-Solomon encoding block parameters from the retrieved transmission unit, performs decoding operation on the Galois field, restores the original data fragments, reorganizes the decoded data fragments according to the fragment sequence number, removes the turbulence level mark and the check field in the packet header, generates the to-be-transmitted data packet, the IPFS retrieval uses the content identifier CIDv1 (based on the SHA2-256 hash algorithm), the Reed-Solomon decoder is configured with the same generating polynomial and Galois field parameters as the sending end; the integrity of the to-be-transmitted data packet is confirmed by the consistency of the reorganized CRC-32 value and the stored value in the packet header, and the check range covers all bytes of the data fragment load.

[0135] S7.4, decoding the to-be-transmitted data packet and inputting into the TensorFlow Lite model to generate an analysis report.

[0136] Further, the to-be-transmitted data packet is subjected to H.265 video decoding and attitude data decompression, and the decoded video frames are aligned with the inertial measurement unit attitude data according to the timestamps; the aligned data is input into the TensorFlow Lite model to perform target detection analysis to generate a structured result analysis result and a JSON format analysis report combined with the timestamp and location information; the H.265 decoding is realized using the FFmpeg library, and the attitude data decompression restores the original quaternion through differential encoding reverse operation; the TensorFlow Lite model adopts the quantized YOLOv5s architecture (INT8 precision), and the target detection threshold is set to 0.5; the JSON report field includes timestamp (Unix millisecond timestamp), location (WGS84 coordinates), and objects (target array); the alignment of the video frames and the attitude data is based on the PTP synchronous clock, the TensorFlow Lite inference is accelerated through the NEON instruction set, and the analysis report is generated.

[0137] The embodiment also provides a computer device suitable for the case of the construction method of the real-time data analysis model of the unmanned aerial vehicle based on a blockchain, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the construction method of the real-time data analysis model of the unmanned aerial vehicle based on a blockchain proposed in the above embodiment.

[0138] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0139] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for constructing a real-time data analysis model for a drone based on blockchain as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0140] In summary, the present invention achieves hardware-level identity authentication by: acquiring physical fingerprint features through terahertz scanning of the fuselage metasurface, generating interference-resistant hash values ​​based on spatiotemporal parameters, fundamentally eliminating the risk of drone identity cloning, dynamically adjusting the RS coding scheme and routing strategy based on real-time turbulence intensity, and accurately quantifying the channel state through the refractive index structure constant to improve coding efficiency and ensure the transmission success rate in a strong interference environment. Physical fingerprint authentication ensures the credibility of the data source, and dynamic coding ensures the reliability of the transmission process. Together with the blockchain consensus mechanism, a full-stack security system from the physical layer to the application layer is constructed.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a real-time data analysis model for drones based on blockchain, characterized by: include, The drones establish an ad-hoc mesh network using the 5G NR-U frequency band. The ground control station verifies the legitimacy of the network, generates an initial network topology, and builds a private blockchain network based on the Hyperledger Fabric framework to generate a genesis block. Collect video stream data, the inertial measurement unit records attitude data, and generates compressed data packets; The drone uses a terahertz transmitter to scan the metasurface pattern on its fuselage to obtain a scattering feature matrix, and generates a composite hash value that is appended to the compressed data packet to form a data packet to be transmitted; Measure the refractive index structure constant of the current area to obtain the turbulence intensity level, add the turbulence intensity level to the data packet to be transmitted, split the data packet into data segments according to the initial network topology, and select relay nodes for routing and forwarding using the Kademlia algorithm; Redundant encoding is performed on the data packets to be transmitted based on the turbulence intensity level to generate transmission units with error correction codes. The endorsing nodes of the private blockchain network verify the transmission units with error correction codes through the Byzantine fault-tolerant consensus mechanism, sort the service nodes in the private blockchain network to generate a new block, and transmit the verified transmission units with error correction codes to the new block. The smart contract triggers the edge computing node to perform Reed-Solomon decoding on the transmission unit with the error correction code to restore it to the data packet to be transmitted, use the TensorFlow Lite inference engine to perform target detection analysis, and generate a format analysis report.

2. The method for constructing a blockchain-based drone real-time data analysis model according to claim 1, characterized in that: The drone establishes an ad-hoc mesh network through the 5G NR-U frequency band. The ground control station verifies the legitimacy and generates an initial network topology. A private blockchain network is built based on the Hyperledger Fabric framework to generate a genesis block. The following steps are included: The baseband chip scans the 5925-7125MHz frequency band, measures the frequency signal-to-noise ratio, and generates a spectrum analysis report. The ground control station receives the spectrum analysis report, executes the frequency algorithm to obtain the center frequency, and then derives the TDMA time slot allocation plan based on the center frequency to obtain the time slot configuration table. The drone sends a beacon frame according to the time slot configuration table, establishes a connection node and generates a neighbor list. The ground control station sends a PUF challenge code to each drone to obtain a PUF feature vector set. The control station sends a maneuver command and generates a behavior authentication result through IMU data verification. Combine the behavior authentication results with the PUF feature vector set, perform threshold judgment, and output a list of legal drones; Based on the neighbor list and the legal drone list, the Dijkstra algorithm is used to generate the initial topology map; According to the list of legal drones, use the cryptogen tool to generate a node certificate set, configure the node certificate set, and obtain the genesis block.

3. The method for constructing a blockchain-based drone real-time data analysis model according to claim 2, characterized in that: Collect video stream data, the inertial measurement unit records attitude data, and generates compressed data packets. The following steps are included: The drone aligns the sensor clock through the GPS PPS signal to obtain a timestamp sequence, and dynamically adjusts the frame rate based on the timestamp sequence and IMU speed data to obtain video stream data; The inertial measurement unit uses the timestamp sequence IMU velocity data to trigger differential encoding to obtain attitude data; The video stream data and posture data are spliced ​​together to obtain a compressed data packet.

4. The method for constructing a blockchain-based drone real-time data analysis model according to claim 3, characterized in that: The drone scans the metasurface pattern of the fuselage through the terahertz transmitter to obtain the scattering feature matrix, generates a composite hash value and appends it to the compressed data packet to form the data packet to be transmitted. The following steps are included: The drone uses a 0.3THz transmitting module to scan the fuselage metasurface in a spiral trajectory to obtain raw scattered signal data. It then performs a short-time Fourier transform on the raw scattered signal data to obtain a time-frequency matrix. The ResNet-18 model is trained based on the sample data set in the time-frequency matrix. The time-frequency matrix is ​​input into the trained ResNet-18 model to obtain the unique scattering fingerprint feature vector. Read the drone's current grid code, fuselage temperature, and unique scattering fingerprint feature vector to obtain anti-interference data; Perform SHA3-256 hash operation on the interference data to obtain a composite hash value; The composite hash value is concatenated with the compressed data packet to obtain the data packet to be transmitted.

5. The method for constructing a blockchain-based drone real-time data analysis model according to claim 4, characterized in that: Measure the refractive index structure constant of the current area to obtain the turbulence intensity level, add the turbulence intensity level to the data packet to be transmitted, split the data packet to be transmitted into data segments according to the initial network topology map, and select the relay node for routing forwarding through the Kademlia algorithm, including the following steps: By collecting and measuring temperature, humidity and air pressure in real time, the atmospheric data are summarized and the refractive index structure constant is calculated based on the atmospheric data; The turbulence intensity level is obtained based on the current altitude of the UAV and the refractive index structure constant; The turbulence intensity level is added to the header flag of the data packet to be transmitted to obtain a turbulence-marked data packet. The fragment size is calculated based on the initial network topology map and the turbulence-marked data packet, and the turbulence-marked data packet is split into a sequence of data fragments. The Kademlia algorithm is used to select relay nodes from the initial network topology to obtain a node list, and the data slice sequence is sent to the target node according to the node list.

6. The method for constructing a blockchain-based drone real-time data analysis model according to claim 5, characterized in that: Redundant encoding is performed on the data packets to be transmitted according to the turbulence intensity level to generate transmission units with error correction codes. The endorsing nodes of the private blockchain network verify the transmission units with error correction codes through the Byzantine fault-tolerant consensus mechanism, which includes the following steps: Selecting a Reed-Solomon coding scheme according to the turbulence intensity level to perform redundant coding on the data packet to be transmitted, obtaining a configuration instruction, performing byte block division on the compressed data packet to obtain a data fragment group of a sequence of data blocks to be encoded, encoding the data fragment group of the sequence of data blocks to be encoded to obtain an original coding unit, adding a packet header to the original coding unit, and generating a final transmission unit with an error correction code; Perform a cyclic redundancy check on the transmission unit with the error correction code according to the endorsing node of the private blockchain network to obtain the initial verification state. Perform Reed-Solomon decoding on the initial verification state to obtain the decoding state. Check the difference between the original data shard timestamp and the current block height to obtain the time-space verification state. Verify the digital signature of the transmission unit with the error correction code in the Intel SGX enclave to obtain the security authentication state. Generate local verification results based on the initial verification state, decoding state, time and space verification state and security authentication state; The local verification results are verified through the Byzantine fault-tolerant consensus mechanism to obtain the verified transmission unit with error correction code.

7. The method for constructing a blockchain-based drone real-time data analysis model according to claim 6, characterized in that: Sorting the service nodes in the private blockchain network to generate a new block, and delivering the verified transmission units with error correction codes to the new block, includes the following steps: Based on the CPU load, network latency, and storage IOPS of the service nodes in the private blockchain network, a priority queue is generated. Verified transmission units with error correction codes are selected from the memory pool, sorted by reception timestamp, and a block structure template is generated. The block structure template is hashed with SHA3-256 to obtain a block hash value. The block hash value is then ECDSA-signed with the private key to obtain a block signature. Combine the block structure template, block hash value and block signature to obtain a new block; Through the gRPC protocol, the verified transmission unit with error correction code is delivered to the new block.

8. The method for constructing a blockchain-based drone real-time data analysis model according to claim 7, characterized in that: The smart contract triggers the edge computing node to perform Reed-Solomon decoding on the transmission unit with error correction code to restore it to the data packet to be transmitted, and uses the TensorFlow Lite inference engine to perform target detection analysis and generate an analysis report, including the following steps: The video stream data and posture data in the private blockchain network are counted through smart contracts to obtain dynamic thresholds; The smart contract monitors and verifies that the transmission units with error correction codes have reached the dynamic threshold, sends a trigger instruction to the edge node, and obtains the transmission unit ID list; According to the transmission unit ID list, the transmission unit with Reed-Solomon error correction code is obtained from IPFS and decoded to obtain the data packet to be transmitted; Decode the data packets to be transmitted, input them into the TensorFlow Lite model, and generate an analysis report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for constructing a blockchain-based drone real-time data analysis model are implemented as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a blockchain-based drone real-time data analysis model are implemented.

Images