Method and system for returning flight data of unmanned aerial vehicle

Abstract

The invention discloses an unmanned aerial vehicle flight data return method and system, and relates to the technical field of wireless communication, and the method comprises the steps: obtaining real-time data collected by an unmanned aerial vehicle; screening key data; carrying out lossless coding on the protocol data, the file data and the data in the key area, and carrying out lossy coding on the data outside the key area in the video data; a VPN tunnel is constructed based on a QUIC protocol, data obtained after protocol data, file data and video data are coded are divided into multiple fragments, key data fragments form a high-priority queue, and non-key data fragments form a low-priority queue; the bandwidth higher than that of the low-priority queue is allocated to the high-priority queue, and the bandwidth is transmitted back to the monitoring center through the VPN tunnel; and decrypting the received data, and decompressing the decrypted data. According to the method, the VPN tunnel, the QUIC and the adaptive coding technology are combined, and the time delay of data return is reduced as much as possible on the premise of ensuring lossless transmission of key data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, in particular to a method and system for returning flight data of a UAV. BACKGROUND

[0002] The UAV is a device that is widely used in many fields such as industry and agriculture. For example, in the process of power line inspection, the UAV can replace manual work to perform periodic and reliable data collection and judgment, greatly improving the efficiency and quality of power line inspection.

[0003] During the operation of the UAV, various data including the state data of the UAV itself and target data need to be returned to the ground control device in real time, so that the control device can execute accurate and effective strategies, such as adjusting the height and speed of the UAV. There are two common data return methods, one is dependent on the ground network, and the other is based on satellite communication. The ground network relies on the construction of basic equipment, and in some areas where the base stations are less arranged, the data return effect will be poor, while the satellite communication is not limited by location, but the satellite communication has high cost. In order to solve the cost problem of satellite communication, CN120434056A provides a system and method for returning flight data of a UAV, which uses VPN tunnel technology to reduce the cost while ensuring the return effect.

[0004] However, the VPN tunnel technology also has its technical limitations. The VPN tunnel needs to go through link negotiation, identity authentication, data encapsulation and decapsulation, etc. in the data transmission process. Each link has a fixed protocol overhead, usually 10-20ms. This inherent transmission delay cannot meet the millisecond-level return requirement of some data, mainly protocol data, and the actual use effect is not ideal. SUMMARY

[0005] The embodiments of the present application provide a method and system for returning flight data of a UAV to solve the problem of large inherent delay when using VPN tunnel to transmit data in the prior art.

[0006] In one aspect, the embodiments of the present application provide a method for returning flight data of a UAV, comprising: obtaining real-time data collected by the UAV, the real-time data including protocol data, file data and video data; filtering key fields in the protocol data through a feature weight model, filtering key modules in the file data using an information entropy method, and identifying key regions in the video data using a YOLOv8 target detection algorithm; The LZ4 lossless compression algorithm is used to perform lossless encoding of the protocol data, the Deflate lossless compression algorithm is used to perform lossless encoding of the file data, the H.265 encoding standard is used to perform lossless encoding of the data in the key area, and the data outside the key area of ​​the video data is performed lossy encoding. The video data after lossless encoding and lossy encoding is then spliced ​​together. A VPN tunnel is built based on the QUIC protocol. The encoded data of protocol data, file data and video data is divided into multiple fragments. The fragments that store the encoded data corresponding to the data of key fields, key modules and key areas are designated as key data fragments. The other fragments are designated as non-key data fragments. Key data fragments form a high-priority queue and non-key data fragments form a low-priority queue. The high-priority queue is allocated a higher bandwidth than the low-priority queue, and the data in the high-priority queue and the low-priority queue are transmitted back to the monitoring center through the VPN tunnel according to the allocated bandwidth. The received data is first decrypted, then reassembled into complete data according to the fragment sequence number. The encoded packets of protocol data in the complete data are decompressed using the LZ4 algorithm, the encoded packets of file data in the complete data are decompressed using the Deflate algorithm, and the encoded packets of video data in the complete data are decoded using H.265.

[0007] On the other hand, embodiments of this application also provide a system for transmitting UAV flight data back to the user, including: The data acquisition module is used to acquire real-time data collected by the drone, including protocol data, file data, and video data. The key data filtering module is used to filter key fields in protocol data through a feature weight model, filter key modules in file data using the information entropy value method, and identify key regions in video data using the YOLOv8 object detection algorithm. The data encoding module is used to perform lossless encoding of protocol data using the LZ4 lossless compression algorithm, lossless encoding of file data using the Deflate lossless compression algorithm, lossless encoding of data in key areas using the H.265 encoding standard, lossy encoding of data outside key areas in video data, and splicing the bitstreams after lossless and lossy encoding of video data. The data segmentation module is used to build a VPN tunnel based on the QUIC protocol. It divides the encoded data of protocol data, file data and video data into multiple segments. The segments that store the encoded data corresponding to the data of key fields, key modules and key areas are designated as key data segments, and the other segments are designated as non-key data segments. Key data segments form a high-priority queue, and non-key data segments form a low-priority queue. The data transmission module is used to allocate higher bandwidth to high-priority queues than to low-priority queues, and transmit the data in the high-priority queues and low-priority queues back to the monitoring center through a VPN tunnel according to the allocated bandwidth. The data decompression module is used to first decrypt the received data, then reassemble it into complete data according to the fragment sequence number. The LZ4 algorithm is used to decompress the encoded packets of protocol data in the complete data, the Deflate algorithm is used to decompress the encoded packets of file data in the complete data, and H.265 decoding is performed on the encoded packets of video data in the complete data.

[0008] The method and system for transmitting UAV flight data disclosed in this application have the following advantages: By combining VPN tunneling, QUIC, and adaptive coding techniques, this method minimizes data return latency while ensuring lossless transmission of critical data. Experiments show that, regardless of whether it is protocol data, file data, or video data, the technology described in this application achieves the required levels of latency and compression ratio for UAV data return. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a method for transmitting UAV flight data in an embodiment of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] Figure 1 A flowchart illustrating a method for transmitting UAV flight data according to an embodiment of this application. This application provides a method for transmitting UAV flight data, including the following steps: S100 acquires real-time data collected by the drone, including protocol data. D pro File data D file and video dataD vid .

[0013] For example, the protocol data includes position parameters, attitude parameters, and control parameters; the file data includes sensor log files and task report files; and the video data includes real-time video and historical video.

[0014] Specifically, the UAV's flight status parameters are collected in real time via a TCP connection port 10001 between the intranet unit and the command and control station DM host. Location parameters include latitude and longitude. Lon , Lat ), relative height H rel Absolute height H abs Attitude parameters include pitch angle. θ Roll angle Heading angle ψ Control parameters include navigation mode. M nav ,airspeed V air , ground speed V ground Sampling frequency f pro =50Hz, ensuring real-time response to control commands.

[0015] Structured files are collected from the command and control station archive server according to the naming rule of "task ID + timestamp". These files include: sensor log files, which record the raw data sequences of sensors such as lidar and infrared thermography; and task report files, which contain task objectives, flight path planning, and abnormal event records. Collection frequency... f file =1Hz, while extracting file metadata Meta file This data includes file size. S file Creation time T create , check value H file .

[0016] Connect the onboard camera via the HDMI interface of the video data acquisition unit to capture real-time onboard video; connect the first network video recorder (NVR) via the RJ45 Ethernet port to capture historical field video. Basic video parameters are as follows.

[0017] Real-time video: frame rate f vid =25fps, raw bitrate R raw =8Mbps.

[0018] Historical Videos: Frame Rate f vid =15fps, raw bitrate R raw =4Mbps.

[0019] Video frame timestamps are recorded synchronously during acquisition. T frame This ensures the continuity of the frame sequence.

[0020] Furthermore, after acquiring real-time data, the protocol data is formatted uniformly, the file data is verified for integrity, and the video data is synchronized frame by frame.

[0021] Specifically, standardizing the protocol data format involves converting the collected UDP multicast data and TCP composite data into a unified JSON format, standardizing field names to avoid ambiguity in subsequent identification. File data integrity verification involves calculating the SHA-256 hash value of the collected files. H curr , and archived metadata H file If compared, H curr = H file If the data acquisition is complete, then confirm that the acquisition is complete; otherwise, re-acquire the data. Video data frame synchronization is based on timestamps. T frame Duplicate frames are removed; frames with an interval of less than 40ms are considered duplicates. Missing frames are then filled in to ensure the continuity of the video stream.

[0022] S110 uses a feature weighting model to filter key fields in protocol data, employs the information entropy method to filter key modules in file data, and uses the YOLOv8 object detection algorithm to identify key regions in video data.

[0023] For example, key fields in protocol data are filtered using a feature weight model, including: Assign weights to each protocol field in the protocol data; Calculate the corresponding importance score based on the weight of each protocol field; Protocol fields with importance scores higher than the protocol threshold are designated as key fields.

[0024] Specifically, different fields in the protocol data have different weights influencing flight control, and key fields need to be selected through a feature weight model.

[0025] According to drone control safety standards, such as GB / T38948-2020, weights are assigned to each protocol field. w i A higher weight indicates that the field is more critical.w Lon = w Lat =0.95、 w θ = = w ψ =0.9、 w Mnav =0.85、 w Vair = w ground =0.8, other auxiliary fields w i =0.3, where w Lon and w Lat These are the weighting coefficients for the longitude and latitude fields, respectively. w θ , and w ψ These are the weighting coefficients for the pitch angle, roll angle, and yaw angle fields, respectively. w Mnav This is the weighting coefficient for the navigation mode field. w Vair and w ground These are the weighting coefficients for the airspeed and ground speed fields, respectively.

[0026] Calculate the importance score for each protocol field. S j The score is higher than the protocol threshold. T pro Fields with a value of 0.75 are identified as key fields. K pro :

[0027] Where, Δ D j For the current frame and the previous frame fields j The larger the difference, the more significant the change in the field, and the greater its impact on control decisions. For fields j The maximum range of values ​​used to normalize the score. S j ∈(0,1).

[0028] Furthermore, the information entropy method is used to filter key modules in the file data, including: Divide the file data into multiple modules according to a fixed size; Calculate the information entropy of each module; Calculate the mean of all information entropy, and designate modules whose information entropy is 20% higher than the mean as key modules.

[0029] Specifically, the file data D file According to fixed size Block size =1024 bytes divided into N Each module Block 1, Block 2,..., Block N .

[0030] Information entropy H k Reflection Module Block k The data uncertainty is such that a higher entropy value indicates that the module contains more effective information.

[0031] in, m For module Block k The number of different byte values ​​in the middle. m ≤256; p k,i Byte value i exist Block k The probability of occurrence in p k,i = count i / Block size , count i Byte value i The number of times it appears.

[0032] Calculate the mean of the information entropy of all modules. bar H Set threshold T file =1.2· bar H ,but K file ={ Block k | H k ≥ T file}

[0033] The key regions of the video data are those related to the task objective. The key regions are identified using the YOLOv8 object detection algorithm, as follows.

[0034] Training dataset construction: Collect video frames of typical drone scenarios, annotate the bounding boxes of target regions, and construct the training set. Dataset train .

[0035] Model training and inference: Input the training set into the YOLOv8 model and train it until the loss function is reached. Loss <0.05, this loss function includes classification loss and bounding box regression loss; for real-time acquired video frames Frame t The model outputs the confidence level of the target region. Conf ,0< Conf ≤1.

[0036] Key region identification: setting confidence threshold T vid =0.8, if Conf ≥ T vid Then the target area is the critical area. K vid Record its bounding box coordinates; simultaneously calculate inter-frame coordinates. ROI The Intersection over Union (IOU) ratio of the region of interest ensures regional continuity.

[0037] in, K vid,t For the first t The key region of the frame, Area() is the area calculation function for the region, if IOU If the value is ≥0.6, it is determined to be the same target region, and there is no need to reinitialize the encoding parameters; if IOU If the value is less than 0.6, it is determined to be a new target region, and the encoding strategy is updated.

[0038] The S120 uses the LZ4 lossless compression algorithm to perform lossless encoding on protocol data, the Deflate lossless compression algorithm to perform lossless encoding on file data, and the H.265 encoding standard to perform lossless encoding on data in key areas. It also performs lossy encoding on data outside key areas in video data and then splices the lossless and lossy encoded bitstreams of the video data together.

[0039] For example, the LZ4 lossless compression algorithm has a fast compression speed and low decompression overhead, which is suitable for the high real-time requirements of protocol data. The encoding process is as follows.

[0040] Sliding window initialization: Set the size of the sliding window W lz4 =64KB.

[0041] Repeating sequence matching: Traverse the protocol data byte by byte. If the current sequence matches a historical sequence within the window, i.e., the matching length is... L match If the length is ≥4 bytes, the original sequence is replaced with the offset plus the matching length; if there is no match, the original bytes are output directly.

[0042] Compression ratio calculation: Size of compressed data S pro,comp Original size S pro,raw Compression ratio CR pro :

[0043] Require CR pro ≤50%, meaning the compressed file size is reduced by more than 50%, while ensuring that the decompressed data is completely consistent with the original data, i.e., passing SHA-256 verification. H pro,comp = H pro,raw .

[0044] The Deflate lossless compression algorithm is used, which can meet the high reliability requirements of file data. The compression process is as follows.

[0045] Dictionary compression phase: Build size D deflate A 32KB dictionary stores the recurring data sequences that appear in the file, replacing the original sequences with dictionary indices.

[0046] Huffman coding stage: For the output after dictionary compression, a Huffman tree is constructed based on the frequency of character occurrence, and short codes are assigned to high-frequency characters to further reduce the amount of data.

[0047] Critical module protection: Protection of critical modules K file Add a critical flag bit (1) to the header of the compressed file, where 1 indicates critical and 0 indicates non-critical, to ensure that the receiving end prioritizes decompressing critical modules; simultaneously calculates... K file Independent hash value H K,file This is used for integrity verification after decompression. The performance metric is the compression ratio of the file data using the Deflate algorithm. CR file ≈30%, decompression speed >200MB / s, and after decompression H file,decomp = H file,raw To achieve 100% lossless results.

[0048] Using the H.265 encoding standard, key areas are... K vid The lossless mode is used, while the lossy mode is used for non-critical areas to balance image quality and bandwidth. The encoding process is as follows.

[0049] For key areas K vid For lossless encoding, the encoding mode is set to intra_period=1, meaning each frame is an I-frame, avoiding the lossless effect caused by prediction errors in P / B frames. The quantization parameter is set to... QP K =0; Bitrate control adopts CBR constant bitrate mode, bitrate R K =20Mbps, ensuring the bandwidth required for lossless encoding.

[0050] For non-critical areas Non - K vid The lossy encoding is used, with the encoding mode set to intra_period=30, meaning one I-frame every 30 frames, and the rest are P-frames; quantization parameters... QP Non-K Based on real-time bandwidth B real Dynamic adjustment, the adjustment formula is:

[0051] when B real =8Mbps, QP Non-K =30, at which point the image quality loss in non-critical areas such as the sky background is acceptable to the human eye, that is, the peak signal-to-noise ratio PSNR≥35dB.

[0052] Rate control uses VBR (Variable Rate Bracket) mode, with a maximum bitrate of [missing value]. R Non-K,max =5Mbps, to avoid occupying bandwidth in critical areas.

[0053] Will K vid lossless bitstream and Non - K vid Lossy bitstream by frame timestamp T frame splicing into a complete video stream Stream vid At the same time, ROI coordinate information is added to the header of the bitstream so that the receiving end can identify key areas during decoding.

[0054] S130 constructs a VPN tunnel based on the QUIC protocol, dividing the encoded protocol data, file data, and video data into multiple fragments. Among these fragments, the fragments that store the encoded data corresponding to key fields, key modules, and key areas are designated as key data fragments, while the other fragments are designated as non-key data fragments. Key data fragments form a high-priority queue, while non-key data fragments form a low-priority queue.

[0055] For example, building a VPN tunnel based on the QUIC protocol includes: The 0-RTT handshake method stores the TLS session ticket during the initial handshake connection, and uses the TLS session ticket directly for authentication during subsequent reconnection. The AAA server is split into edge nodes and central nodes. The edge nodes pre-store a list of authorized devices and perform primary authentication during the first handshake connection. The central node completes secondary authentication through device fingerprints. After successful authentication, a VPN tunnel key is generated. After receiving the data at the monitoring center, the received data is decrypted using the VPN tunnel key.

[0056] Specifically, traditional VPNs, such as L2TP, require a 3-way TCP handshake plus TLS authentication, resulting in latency >50ms. In contrast, the QUIC protocol, based on UDP, supports 0-RTT handshakes. This means that the client router stores the server's TLS session ticket during the initial connection, and uses this ticket directly for authentication during subsequent reconnections, significantly reducing handshake latency. T handshake <10ms.

[0057] To avoid single points of failure, the AAA server is split into edge nodes and a central node. The edge nodes are used to pre-store the list of authorized devices, such as the MAC address and device ID of the external network unit, and complete the initial authentication upon first connection. The central node, on the other hand, completes secondary authentication through device fingerprints, such as the LSB steganography information of the internal network unit. After successful authentication, a VPN tunnel key is generated. Key vpn The key is encrypted using AES-256.

[0058] The tunnel parameters are set to MTU=1472 bytes, adapting to the Internet standard MTU=1500 bytes, reserving a 28-byte QUIC header, and adjusting the sliding window size. W quic =2MB, to avoid congestion.

[0059] Based on the MTU size, the encoded data is divided into fragments. Frag 1, Frag 2,..., Frag MPiece size S frag ≤MTU-QUIC header size = 1472 - 28 = 1444 bytes; For critical data fragments, add a critical flag bit of 1.

[0060] High-priority queue Q high Storage K pro Protocol key fields K file Key modules of the document K vid Video key region segmentation, priority weight α high =0.8.

[0061] low priority queue Q low Non-critical data is stored in shards with priority and weight. α low =0.2.

[0062] S140 allocates higher bandwidth to high-priority queues than to low-priority queues, and transmits data from both high-priority and low-priority queues back to the monitoring center via VPN tunnel according to the allocated bandwidth.

[0063] For example, the scheduler allocates bandwidth according to priority weights to ensure Q high Obtain more transmission resources, scheduling time T sched The calculation formula is:

[0064]

[0065] in, T interval =20ms is the scheduling interval, which is consistent with the protocol data acquisition frequency, so the scheduling time of the high-priority queue is... T sched,high =16ms, scheduling time for low-priority queues T sched,low =4ms, meaning that within every 20ms, 16ms are used to transmit high-priority data and 4ms are used to transmit low-priority data.

[0066] Furthermore, during the data backhaul process through the VPN tunnel, real-time bandwidth is monitored. When real-time bandwidth decreases, the size of the fragments is reduced to avoid fragment timeout retransmission.

[0067] Specifically, the real-time bandwidth is calculated using the RTT sampling method. Breal :

[0068] in, S trans This represents the total amount of data transmitted over the past 10 scheduling intervals. RTT avg The average round-trip time is the average of the past 10 RTTs.

[0069] like B real If the speed drops, for example from 10 Mbps to 5 Mbps, then reduce the fragment size. S frag To avoid fragment timeout retransmissions, the fragment size has been adjusted. S frag,new for:

[0070] when B real =5Mbps, T interval =20ms, B real × T interval =5×10 6 ×20×10 -3 =100,000 bytes, much larger than MTU, therefore S frag,new =1444 bytes; if B real =1Mbps, then B real × T interval =20000 bytes, still larger than MTU S frag,new Keep 1444 bytes; if B real =0.5Mbps, then B real × T interval =10000Byte, S frag,new =10000 bytes, then it needs to be split into multiple fragments.

[0071] right Q high For fragmented data, a selective retransmission (SR) strategy is used, retransmitting only the lost fragments; for Q low The data is fragmented and uses a cumulative retransmission strategy to reduce retransmission overhead. If the link packet loss rate...Loss rate If the percentage is greater than 5%, a temporary increase will be implemented. Q high weight to α high =0.9, ensuring the reliability of critical data transmission.

[0072] S150 first decrypts the received data, then reassembles it into complete data according to the fragment sequence number. The encoded packets of protocol data in the complete data are decompressed using the LZ4 algorithm, the encoded packets of file data in the complete data are decompressed using the Deflate algorithm, and the encoded packets of video data in the complete data are decoded using H.265.

[0073] For example, tunnel decryption uses Key vpn Decrypt the QUIC-VPN data packets to extract the raw fragments. Based on the fragment sequence number, which is contained in the header of each fragment, reassemble the fragments into complete encoded data, including protocol data encoded packets, file data encoded packets, and video stream Streamvid.

[0074] Priority verification: Checks the integrity of received high-priority data, such as... K pro Whether all fragments are received, and if any are missing, a retransmission request is triggered only for high-priority data.

[0075] Furthermore, after decompressing the encoded packets of protocol data in the complete data using the LZ4 algorithm, the hash value of the key fields in the decompressed data is calculated. This hash value is compared with the hash value pre-transmitted by the UAV. If the two hash values ​​are the same, it is determined that the data of the key fields is intact; otherwise, the data of the key fields is requested to be retransmitted.

[0076] Specifically, LZ4 decompression is performed on the protocol data encoded packets, and the restored protocol data is output. D pro,decomp .

[0077] Key field validation: calculation D pro,decomp middle K pro hash value H K,pro,recv , and the pre-transmitted by the sending end H K,pro,send If compared, H K,pro,recv = H K,pro,send Then determine K pro Lossless; otherwise, request a retransmission.

[0078] The restored protocol data is transmitted to the monitoring computer for drone command control, such as according to... K pro The attitude parameters in the navigation command adjust the response delay of the control command. T resp = T lat + T decomp ,Require T resp <30ms, including transmission delay T lat <25ms, data decompression latency T decomp <5ms.

[0079] Furthermore, when using the Deflate algorithm to decompress the encoded packets of file data in the complete data, the data corresponding to the key modules in the encoded packets is decompressed first. The SHA-256 hash value of the decompressed data is calculated, and this SHA-256 hash value is compared with the SHA-256 hash value pre-transmitted by the UAV corresponding to the file data. If the two SHA-256 hash values ​​are the same, it is confirmed that the file data as a whole is intact. At the same time, the SHA-256 hash value of the key modules in the decompressed data is also calculated, and this SHA-256 hash value is compared with the SHA-256 hash value pre-transmitted by the UAV corresponding to the key modules. If the two SHA-256 hash values ​​are the same, it is confirmed that the data of the key modules is intact.

[0080] Specifically, the file data encoding package is decompressed using Deflate, prioritizing decompression based on key identifier bits. K file Output the complete file D file,decomp .

[0081] Integrity verification: calculation D file,decomp SHA-256 hash value H file,recv , and the hash value pre-stored at the sending end H file,send Comparison confirmed that the document was intact; calculation K file hash value H K,file,recv , and the hash value pre-stored at the sending end H K,file,send The comparison confirmed that the key modules were undamaged.

[0082] Data storage: D file,decomp Transmitted to the enterprise intranet via a one-way optical shutter, while simultaneously recording the storage time. Tstore File size S file,decomp This facilitates subsequent retrieval.

[0083] During the video data restoration process, the video bitstream is parsed. Stream vid ROI coordinate information in the image, for identification K vid and Non - K vid Region; To K vid Lossless decoding is used. QP K =0, restore the original pixel; for Non - K vid Using lossy decoding, according to QP Non-K Inverse quantization.

[0084] Decoded K vid and Non - K vid Regions are stitched together according to coordinates to generate a complete video frame. Frame decomp Then Frame decomp Transmitted to display devices, such as 4K displays in a monitoring center, to show latency Δ T disp = T recv - T frame <150ms T recv For video reception time, the complete video is finally transmitted to the enterprise intranet NVR via a one-way optical shutter, stored in MP4 format, with a bitrate of [missing information]. R store = R K + R Non-K , R Non-K The bitrate for non-critical areas of the video data is set to ensure that the stored image quality is consistent with the transmitted image quality.

[0085] This application embodiment also provides a system for transmitting UAV flight data back to the user, the system comprising: The data acquisition module is used to acquire real-time data collected by the drone, including protocol data, file data, and video data. The key data filtering module is used to filter key fields in protocol data through a feature weight model, filter key modules in file data using the information entropy value method, and identify key regions in video data using the YOLOv8 object detection algorithm. The data encoding module is used to perform lossless encoding of protocol data using the LZ4 lossless compression algorithm, lossless encoding of file data using the Deflate lossless compression algorithm, lossless encoding of data in key areas using the H.265 encoding standard, lossy encoding of data outside key areas in video data, and splicing the bitstreams after lossless and lossy encoding of video data. The data segmentation module is used to build a VPN tunnel based on the QUIC protocol. It divides the encoded data of protocol data, file data and video data into multiple segments. The segments that store the encoded data corresponding to the data of key fields, key modules and key areas are designated as key data segments, and the other segments are designated as non-key data segments. Key data segments form a high-priority queue, and non-key data segments form a low-priority queue. The data transmission module is used to allocate higher bandwidth to high-priority queues than to low-priority queues, and transmit the data in the high-priority queues and low-priority queues back to the monitoring center through a VPN tunnel according to the allocated bandwidth. The data decompression module is used to first decrypt the received data, then reassemble it into complete data according to the fragment sequence number. The LZ4 algorithm is used to decompress the encoded packets of protocol data in the complete data, the Deflate algorithm is used to decompress the encoded packets of file data in the complete data, and H.265 decoding is performed on the encoded packets of video data in the complete data.

[0086] Experimental instructions 1. Test Scenario and Parameters Scenario: Power line inspection, drone wingspan 3.5m, maximum flight altitude 500m, flight area coverage urban 5G coverage + weak 5G in mountainous areas, some areas rely on Mesh private network.

[0087] Test data volume: Protocol data acquired at 50Hz for 60 minutes, approximately 1.8 × 10⁻⁶. 6 Each frame and file is 10MB, for a total of 10 frames; the video data is 1080P / 25fps, lasting 60 minutes, approximately 9 x 10. 4 frame.

[0088] Link conditions: 5G bandwidth fluctuation range 2-15Mbps, Mesh private network bandwidth fluctuation range 1-5Mbps, packet loss rate 0.1%-5%.

[0089] 2. Core performance indicators

[0090] Key data, including K pro , K file , K vid Lossless encoding and hash verification achieved 100% data restoration, with no data loss or errors, meeting the requirements for flight control and post-event review. Protocol data transmission latency remained stable at 22-28ms, below the design threshold of 30ms, ensuring no lag in monitoring center commands and control. Video stuttering rate was 0.5%, below the design threshold of 1%, and... K vid In critical areas, such as insulators, image quality remains lossless (PSNR = 45dB). In non-critical areas, such as the sky, image quality loss is acceptable to the human eye (PSNR = 36dB). Data compression is better than design requirements, significantly reducing bandwidth consumption. Even in mountainous mesh private network scenarios with bandwidth of 1-5Mbps, stable data transmission with no link interruptions is still guaranteed.

[0091] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0092] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for backhauling flight data of a UAV, the method comprising: The application comprises the following steps: acquiring real-time data collected by a UAV, wherein the real-time data comprises protocol data, file data and video data; screening key fields in the protocol data by a feature weight model, screening key modules in the file data by an information entropy method, and identifying key areas in the video data by a YOLOv8 target detection algorithm; performing lossless encoding on the protocol data by an LZ4 lossless compression algorithm, performing lossless encoding on the file data by a Deflate lossless compression algorithm, performing lossless encoding on data in the key areas by an H.265 encoding standard, and performing lossy encoding on data outside the key areas in the video data, and splicing code streams obtained by performing lossless encoding and lossy encoding on the video data; constructing a VPN tunnel based on a QUIC protocol, dividing data obtained by encoding the protocol data, the file data and the video data into multiple segments, storing encoded data corresponding to data of the key fields, the key modules and the key areas in segments of the multiple segments as key data segments, storing other segments as non-key data segments, grouping the key data segments into a high-priority queue, and grouping the non-key data segments into a low-priority queue; allocating higher bandwidth to the high-priority queue than to the low-priority queue, and returning data in the high-priority queue and the low-priority queue to a monitoring center through the VPN tunnel according to the allocated bandwidth; decrypting received data, reorganizing the received data into complete data according to sequence numbers of the segments, decompressing encoded packets of the protocol data in the complete data by an LZ4 algorithm, decompressing encoded packets of the file data in the complete data by a Deflate algorithm, and performing H.265 decoding on encoded packets of the video data in the complete data. 2.The method of claim 1, wherein, The protocol data comprises position parameters, attitude parameters and control parameters, the file data comprises sensor log files and task report files, and the video data comprises real-time videos and historical videos. 3.The method of claim 1, wherein, After acquiring the real-time data, performing format unification on the protocol data, integrity checking on the file data, and frame synchronization on the video data.

4. The method of claim 1, wherein, Screening key fields in the protocol data by a feature weight model comprises the following steps: allocating weights to each protocol field in the protocol data; calculating importance scores of each protocol field based on the weights of the protocol field; regarding protocol fields with importance scores higher than a protocol threshold as the key fields.

5. The method of claim 1, wherein, Screening key modules in the file data by an information entropy method comprises the following steps: dividing the file data into multiple modules according to a fixed size; calculating information entropy of each module; calculating a mean value of all the information entropy, and regarding modules with information entropy higher than the mean value by 20% as the key modules.

6. The method of claim 1, wherein, Constructing a VPN tunnel based on a QUIC protocol comprises the following steps: The way of using 0-RTT handshake stores TLS session ticket when first handshake connection, and directly uses TLS session ticket for authentication when subsequent reconnection; The AAA server is split into edge node and center node, the edge node pre-stores authorized device list, and performs primary authentication when first handshake connection, the center node completes secondary authentication through device fingerprint, and generates VPN tunnel key after authentication; After receiving data in the monitoring center, the received data is decrypted using the VPN tunnel key.

7. The method of claim 1, wherein, During data transmission through the VPN tunnel, the real-time bandwidth is monitored, and when the real-time bandwidth decreases, the size of the fragment is reduced to avoid timeout retransmission of the fragment. 8.The method of claim 1, wherein, After decompressing the encoded packet of the protocol data in the complete data by using the LZ4 algorithm, the hash value of the key field in the decompressed data is calculated, and the hash value is compared with the hash value pre-transmitted by the unmanned aerial vehicle. If the two hash values are the same, it is determined that the data of the key field is lossless, otherwise, the data of the key field is requested to be retransmitted.

9. The method of claim 1, wherein, When the encoded packet of the file data in the complete data is decompressed by using the Deflate algorithm, the data corresponding to the key module in the encoded packet is preferentially decompressed, the SHA-256 hash value of the decompressed data is calculated, and the SHA-256 hash value is compared with the SHA-256 hash value corresponding to the file data pre-transmitted by the unmanned aerial vehicle. If the two SHA-256 hash values are the same, it is confirmed that the file data is lossless as a whole, and the SHA-256 hash value of the key module in the decompressed data is also calculated, and the SHA-256 hash value is compared with the SHA-256 hash value corresponding to the key module pre-transmitted by the unmanned aerial vehicle. If the two SHA-256 hash values are the same, it is confirmed that the data of the key module is lossless.

10. The system for the backhaul of flight data of the UAV according to any one of claims 1-9, characterized in that, The system comprises: A data acquisition module for acquiring real-time data collected by an unmanned aerial vehicle, the real-time data comprising protocol data, file data and video data; A key data screening module for screening key fields in the protocol data by a feature weight model, screening key modules in the file data by an information entropy method, and identifying key regions in the video data by a YOLOv8 target detection algorithm; A data encoding module for losslessly encoding the protocol data by using an LZ4 lossless compression algorithm, losslessly encoding the file data by using a Deflate lossless compression algorithm, losslessly encoding data in the key regions by using an H.265 encoding standard, and lossily encoding data outside the key regions in the video data, and splicing code streams obtained by lossless encoding and lossy encoding of the video data; The data segmentation module is configured to construct a VPN tunnel based on a QUIC protocol, encode the protocol data, the file data, and the video data, and segment the encoded data into multiple segments, wherein a segment storing encoded data corresponding to the key fields, the key modules, and the key regions is a key data segment, and other segments are non-key data segments; the key data segments form a high-priority queue, and the non-key data segments form a low-priority queue; The data transmission module is configured to allocate a higher bandwidth to the high-priority queue than to the low-priority queue, and transmit data in the high-priority queue and the low-priority queue to a monitoring center through the VPN tunnel according to the allocated bandwidth; The data decompression module is configured to decrypt received data, reorganize the decrypted data into complete data according to the sequence numbers of the segments, decompress encoded packets of the protocol data in the complete data by using an LZ4 algorithm, decompress encoded packets of the file data in the complete data by using a Deflate algorithm, and perform H.265 decoding on encoded packets of the video data in the complete data.

Images