Unmanned aerial vehicle parking and charging optimization method and system based on roadside lamp pole multiplexing

Abstract

The application discloses an unmanned aerial vehicle parking and charging optimization method and system based on roadside lamp pole multiplexing, relates to the technical field of unmanned aerial vehicle parking and charging, and comprises the following steps: transforming an existing roadside lamp pole to build a lamp pole parking point; planning a patrol route; performing vehicle target detection by using an on-board model, and selecting a local image according to a detection result and sending the local image to a nearest edge node; performing orthographic correction processing on the collected image; storing evidence data after performing an encryption signature operation; relying on positioning and visual guidance to land and perform wireless charging on the unmanned aerial vehicle; monitoring the running state of the unmanned aerial vehicle in real time on a cloud side, and transferring an unfinished patrol task when the unmanned aerial vehicle fails.The application solves the technical problems of the prior art, such as difficulty in laying out a special parking and charging station, difficulty in redistributing a patrol task after a failure, and insufficient precision of evidence pictures for archiving and storing, and achieves the technical effects of realizing lamp resource multiplexing and intelligent scheduling of a patrol task, and improving the efficiency of illegal parking patrol and the reliability of evidence.

Description

Technical Field

[0001] This invention relates to the field of drone docking and charging technology, specifically to an optimized method and system for drone docking and charging based on the reuse of roadside light poles. Background Technology

[0002] Currently, in the operation of drone inspections for illegal parking on roads, dedicated drone parking and charging platforms require separate land acquisition and construction, resulting in large investment in site construction and limited site selection, making it difficult to deploy on a large scale based on existing urban road infrastructure. Conventional inspection routes often adopt an equal arrangement, failing to differentiate inspection resources based on the frequency of illegal parking at each parking space, leading to wasted inspection resources and frequent missed inspections of key parking spaces. Drone-collected images are uncorrected and exhibit significant distortion, making license plate recognition and illegal parking determination prone to errors, and evidence materials lack standardization. Evidence data lacks reliable means of preservation and is easily tampered with. Furthermore, when a drone suddenly goes offline due to a malfunction, the original inspection tasks lack a standardized allocation mechanism, making it difficult to connect the remaining parking space inspection work, significantly reducing the overall continuity of inspection operations.

[0003] Existing technologies face technical challenges such as difficulty in deploying dedicated offline charging stations, difficulty in redistributing inspection tasks after a failure, and insufficient precision in evidence-gathering images, making archiving and preservation difficult. Summary of the Invention

[0004] This application provides an optimized method and system for drone parking and charging based on the reuse of roadside light poles, which addresses the technical problems of difficulty in deploying dedicated parking and charging stations, difficulty in reallocating inspection tasks after a failure, and difficulty in archiving and storing evidence due to insufficient image accuracy.

[0005] In view of the above problems, this application provides an optimized method and system for drone parking and charging based on the reuse of roadside light poles.

[0006] The first aspect of this application provides an optimized method for drone parking and charging based on roadside light pole reuse, the method comprising:

[0007] The existing roadside lampposts are modified to create lamppost parking points; the cloud determines parking space weight parameters based on basic parking space data, and plans inspection routes based on real-time parking space data and drone battery status; the drone follows the planned inspection route, and its onboard model performs vehicle target detection, selecting local images based on the detection results and sending them to the nearest edge node; the edge node receives the images, completes license plate recognition and parking space occupancy determination, and performs orthorectification processing on the collected images based on positioning data; the edge node packages the evidence data, performs encryption and signature operations, and stores it on the blockchain; when the drone's battery meets preset conditions, it flies to the nearest lamppost parking point, lands using positioning and vision guidance, and performs wireless charging; the cloud monitors the drone's operating status in real time, and in case of drone malfunction, the unfinished inspection task is transferred based on the parking space hash allocation rule.

[0008] A second aspect of this application provides an optimized drone parking and charging system based on the reuse of roadside light poles, the system comprising:

[0009] The system includes the following modules: a parking spot construction module for modifying existing roadside lampposts to create parking spots; a parking space weight parameter determination module for determining parking space weight parameters in the cloud based on basic parking space data, and planning inspection routes by combining real-time parking space data and UAV battery status; a target detection module for UAVs to perform vehicle target detection along the planned inspection routes using an onboard model, and selecting optimal local images based on the detection results to send to the nearest edge node; an orthorectification processing module for edge nodes to perform license plate recognition and parking space occupancy determination after receiving images, and to perform orthorectification processing on the acquired images based on positioning data; an on-chain storage module for edge nodes to package evidence data, perform encryption and signature operations, and then store it on the blockchain; a wireless charging module for UAVs to fly to the nearest lamppost parking spot when their battery level meets preset conditions, land using positioning and vision guidance, and perform wireless charging; and a UAV monitoring module for real-time monitoring of UAV operation status in the cloud, and transferring unfinished inspection tasks based on parking space hash allocation rules when UAV malfunctions.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The system modifies existing roadside lampposts to create lamppost parking points; the cloud determines parking space weight parameters based on basic parking space data, and plans inspection routes by combining real-time parking space data and drone battery status; the drone follows the planned inspection route, and its onboard model performs vehicle target detection, selecting optimal local images based on the detection results and sending them to the nearest edge node; the edge node receives the images, performs license plate recognition and parking space occupancy determination, and performs orthorectification processing on the collected images based on positioning data; the edge node packages the evidence data, performs encryption and signature operations, and stores it on the blockchain; when the drone's battery meets preset conditions, it flies to the nearest lamppost parking point, lands using positioning and vision guidance, and wirelessly charges; the cloud monitors the drone's operating status in real time, and in case of drone malfunction, transfers unfinished inspection tasks based on parking space hash allocation rules. This achieves the technical effect of realizing street light resource reuse and intelligent scheduling of inspection tasks, improving the efficiency of illegal parking inspection and the reliability of evidence collection. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic diagram of the optimized method for drone docking and charging based on roadside light pole reuse provided in this application embodiment;

[0014] Figure 2 A schematic diagram of the structure of the drone docking and charging optimization system based on roadside light pole reuse provided in this application embodiment.

[0015] Figure labeling: 10, berth location construction module, 20, 30, target detection module, 40, on-chain storage module, 50, wireless charging module, 60, drone monitoring module. Detailed Implementation

[0016] This application provides an optimized method and system for drone parking and charging based on the reuse of roadside light poles, which addresses the technical problems of difficulty in deploying dedicated parking and charging stations, difficulty in reallocating inspection tasks after a fault, and difficulty in archiving and storing evidence due to insufficient image quality.

[0017] 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 a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides an optimized method for drone parking and charging based on the reuse of roadside light poles, the method comprising:

[0019] Step S100: Modify existing roadside light poles to create light pole parking points.

[0020] Specifically, existing Q235 material municipal roadside lighting poles were selected as the renovation carrier. First, the safety verification of the poles was completed through ultrasonic thickness measurement, foundation flaw detection, and verticality testing. L50×5 angle steel and E4303 welding rods were used to reinforce the upper part of the poles with three-sided welding. The aluminum alloy support components were fixed with 8.8 grade M16 stainless steel clamps. A carbon fiber docking platform was installed on the support structure. YJV-4×4mm² power supply cables were laid along the poles and equipped with circuit breakers, leakage protection devices, and power / signal surge protectors (SPDs). At the platform location, ublox ZED-F9P model RTK reference station, Qi EPP standard wireless charging transmitter components with an operating frequency of 87kHz, 940nm infrared and white light dual-mode LED guide light strips, and Apriltag positioning markers were integrated. After passing insulation and grounding tests, standardized pole docking points were built that combine the functions of precise drone landing, wireless fast charging, differential reference output, and visual docking guidance.

[0021] Step S200: The cloud determines the berth weight parameters based on the berth basic data, and plans the inspection route by combining real-time berth data and UAV battery status.

[0022] Specifically, the cloud first extracts three basic indicators for each berth: historical illegal parking probability, berth turnover rate, and complaint frequency, and then performs weighted calculations to generate a berth value coefficient, which serves as the berth weight parameter for route planning. The cloud then uses a lightweight CNN model, inputting time, weather, holidays, and surrounding POI data, to predict the regional parking heat map distribution for the next hour, obtaining real-time berth congestion dynamics. Simultaneously, it aggregates operational parameters for each drone, including real-time coordinates, remaining battery power, rated battery capacity, 500W power consumption during cruising, 800W power consumption during hovering, and a cruising speed of 8m / s. The maximum flight range supported by the remaining battery power is used as a hard constraint. Constraints: The route planning algorithm consists of two levels of scheduling. First, the Hungarian algorithm is used to initially assign parking spaces across multiple UAVs, achieving coarse allocation of tasks over a wide range. Then, a heuristic greedy algorithm is used for parking spaces under the jurisdiction of a single UAV, prioritizing the selection of the next inspection parking space based on the ratio of parking space value to unit energy consumption. The initial inspection path is generated iteratively point by point. Subsequently, a 2-opt local search algorithm is introduced to iteratively optimize the path nodes by exchanging them pairwise, eliminating empty flight segments, and repeatedly verifying that the total flight and hovering energy consumption of the path does not exceed the remaining available power of the UAV. After iterative convergence, the final dynamic inspection route is output.

[0023] Step S300: The UAV follows the planned inspection route, and the onboard model performs vehicle target detection. Based on the detection results, it selects the best local images and sends them to the nearest edge node.

[0024] Specifically, the drone flies at a cruising speed of 8m / s according to the optimized inspection route issued by the cloud. It is equipped with a YOLO-Nano lightweight detection model quantized by INT8 and performs real-time inference to identify vehicles from the pre-processed 416×416 airborne images. When the vehicle detection confidence is greater than 0.5 and less than 0.8 and is in the pending range that needs to be reviewed, the ROI is locally cropped by expanding the edge of the target detection box by 20%. The cropped image is compressed into a small JPEG file and transmitted via the MQTT communication protocol through the Wi-Fi 6 link to the nearest JetsonAGX Orin edge node deployed within the 1km service radius of the drone. If the confidence is higher than 0.8, it is directly determined to be compliant and no image needs to be sent back. This simplifies the transmission of data and reduces the transmission time of the link.

[0025] Step S400: After receiving the image, the edge node completes license plate recognition and parking space occupancy determination, and performs orthorectification processing on the acquired image based on the positioning data.

[0026] Specifically, the roadside edge node, configured with an NVIDIA Jetson AGX Orin 64GB, receives JPEG compressed ROI images from the drone via a Wi-Fi 6 communication link. In the first stage of license plate recognition: a lightweight detection algorithm is used to locate the coordinates of the four corners of the license plate. Four-point perspective transformation is used to correct the tilted license plate into a standard rectangular 96×32 grayscale image, which is then fed into the LPRNet lightweight license plate recognition network, accelerated by the INT8 neural network quantization method and the TensorRT tensor real-time inference acceleration engine, to perform character inference. The CTC connection-time classification loss decoding algorithm is used to eliminate character alignment errors, and the province and number characters and their corresponding confidence scores are output. In the second stage of berth occupancy determination: the centimeter-level drone latitude and longitude coordinates and the local plane coordinates of the four corners of the berth are retrieved from the RTK real-time dynamic differential positioning. The length and width dimensions of the berth boundary are calculated, and the coordinate projection mapping of the four vertices of the vehicle landing in the aerial coordinate system is performed. The vehicle is then compared with the coordinates of the two vertices. The offset distance between the boundary and the berth edge line is used to calculate the span ratio by the ratio of the length of the vehicle exceeding the berth edge line to the total width of the berth, and to quantitatively determine the type of berth occupation violation such as vehicle spanning or pressing the line. In the third step of orthorectification, the fx, fy, cx, cy intrinsic parameter matrices obtained from the calibration of the airborne camera are retrieved. Combined with the UAV RTK elevation data and the Roll, Pitch, Yaw three-axis attitude angles output by the IMU inertial measurement unit, the spatial rotation matrix R and the camera projection matrix P are calculated. The DLT direct linear transformation method is used to solve the image homography matrix H. The warpPerspective perspective transformation algorithm is called to complete the image resampling pixel by pixel, and a uniform orthorectified image with a fixed ground sampling distance GSD is generated. This completely eliminates the perspective distortion caused by aerial pitch and yaw, and provides standard scale evidence images for subsequent judicial evidence preservation.

[0027] Step S500: The edge node packages the evidence collection data and performs encryption and signature operations before storing it on the blockchain.

[0028] Specifically, the edge node encapsulates orthorectified evidence images, license plate recognition results, RTK real-time dynamic differential positioning spatiotemporal location information, and parking space determination data into a unified JSON format evidence data package. First, a unique hash digest is generated for the data package using the SHA-256 hash algorithm. Then, a digital signature is completed based on the private key using the RSA-2048 asymmetric encryption algorithm, and the signature content is Base64 encoded. Subsequently, the Hyperledger Fabric v2.4 blockchain software development kit (SDK) is called to run the EvidenceContract smart contract, submitting the complete evidence package with the digital signature to the blockchain. After the platform completes block packaging and consensus confirmation, the data is written into the blockchain to achieve on-chain solidification and evidence storage. The hash verification and asymmetric signature ensure that the evidence data is tamper-proof and its source is verifiable, meeting the requirements for evidence validity.

[0029] Step S600: When the drone's battery level meets the preset conditions, it flies to the nearest lamppost docking point, lands using positioning and visual guidance, and wirelessly charges.

[0030] Specifically, the drone relies on its onboard control system to monitor the remaining battery power of the TB60 in real time. When the remaining power reaches a preset safety threshold, it automatically queries the coordinates of the nearest lamppost docking point and generates the nearest landing route. During the landing process, it integrates RTK real-time dynamic differential positioning centimeter-level differential positioning with visual dual guidance composed of Apriltag marking and a 940nm infrared + white light ring LED guide light strip to complete the docking. After the drone lands, the onboard wireless charging receiving coil is precisely coupled with the Qi EPP standard transmitting coil at the docking point. It adopts a PID closed-loop control strategy to carry out wireless charging. In the early stage, it uses a high-power fast charging of up to 500W. As the battery temperature rises, the charging power is automatically reduced to 500W for temperature control protection when the battery reaches 80%. The recharging operation is completed in 18 minutes. The overall charging efficiency of the entire wireless charging system can reach 67.5%.

[0031] Step S700: The cloud monitors the drone's operating status in real time. When a drone malfunctions, the incomplete inspection task is transferred based on the berth hash allocation rule.

[0032] Specifically, the cloud-based scheduling layer continuously collects the location, remaining battery power, communication link, and onboard equipment operation data of each drone according to a 5-second heartbeat cycle, and monitors abnormal equipment conditions in real time. If two consecutive heartbeat messages are lost, it is determined that the drone has a communication interruption fault. Then, the MurmurHash3 hash algorithm is used to split the uninspected berth tasks under the jurisdiction of the faulty drone according to the partitioning and allocation rule of modulo the total number of drones by the berth number. The split berths are then redistributed to the other normal drones. The time taken for a single task re-hash allocation is about 15ms. After adding the time for message push and drone response, the overall task migration is controlled within 2.5 seconds. Dynamic task transfer ensures the continuous operation of berth inspections along the entire route and improves the operational availability of the entire inspection system.

[0033] In one possible implementation, step S100 further includes:

[0034] Step S110: Conduct structural inspection on the existing light pole and use reinforcement components to fix the load-bearing platform.

[0035] Step S120: Assemble the charging transmitter, reference positioning device, light guide and positioning mark on the carrier platform.

[0036] Step S130: Install power supply lines and protective devices, and complete electrical performance testing.

[0037] Specifically, an ultrasonic thickness gauge was used to test the wall thickness of the original roadside light pole made of Q235 material, ground-penetrating radar was used to test the compactness of the concrete foundation, and a total station was used to verify the vertical deviation of the light pole. After the structural inspection was qualified, L50×5 Q235 angle steel was used as a reinforcement. E4303 welding rods were used to weld and fix the light pole on three sides with a welding current of 120A and a voltage of 22V. The 6061-T6 aluminum alloy bracket was locked with 304 stainless steel double clamps with M16 high-strength bolts of grade 8.8. Finally, a carbon fiber bearing platform with a diameter of 600mm was fixed on top of the bracket, and the overall installation level of the platform was controlled to be 1.5mm / m, thus completing the reinforcement and assembly of the docking platform.

[0038] The components were assembled sequentially on the carbon fiber support platform. A wireless charging transmitter with an operating frequency of 87kHz and conforming to the Qi EPP standard was installed. The ublox ZED-F9P RTK real-time dynamic differential positioning reference positioning device and TW5400 multi-frequency GNSS antenna were installed. A ring-shaped LED light guide with 940nm infrared and white light dual-mode was deployed around the outer edge of the platform. An Apriltag positioning mark with a size of 300mm×300mm and ID number 001 was fixed on the platform surface. At the same time, YJV-4×4mm² power cables were laid along the light pole, and a C16A circuit breaker, a 30mA leakage current protector, and power and signal SPD surge protectors were connected in series to complete the electrical wiring of the entire set of components.

[0039] A φ32 PVC protective conduit was installed along the modified light pole, and a YJV-4×4mm² power cable was laid inside. The cable was then connected to a C16A / 2P circuit breaker, a 30mA residual current device, a 40kA power surge protection device, and a 10kA signal surge protection device, among other electrical protection devices. The wiring terminals were secured with OT-4 cold-pressed terminals. After the wiring was completed, a 500V megohmmeter was used to conduct cable insulation testing. The measured insulation resistance reached 12MΩ, and the grounding resistance was 3.2Ω. With both indicators meeting the standards, the electrical acceptance was completed.

[0040] In one possible implementation, step S200 further includes:

[0041] Step S210: Calculate berth weights based on various statistical indicators of berths.

[0042] Step S220: Initially select inspection berths according to the relationship between berth weight and flight energy consumption.

[0043] Step S230: Determine the available inspection range based on the remaining power, and adjust the inspection route using a path optimization algorithm.

[0044] Specifically, the cloud-based system extracts three raw statistical indicators from the berth database: historical illegal parking probability, berth turnover rate, and complaint frequency. These three indicators are then normalized to the 0-1 range before a weighted calculation rule is configured: historical illegal parking probability accounts for 40% of the weight, berth turnover rate accounts for 35%, and complaint frequency accounts for 25%. The berth weight is calculated as: Normalized Illegal Parking Rate × 0.4 + Normalized Turnover Rate × 0.35 + Normalized Complaint Frequency × 0.25. The resulting dimensionless value is the berth value coefficient, or berth weight. For four typical application scenarios—commercial streets, office buildings, residential areas, and areas surrounding schools—differentiated weights are generated by substituting the corresponding indicator data for each scenario into this fixed weighting formula. A higher weight indicates a higher priority for berth inspection, providing a quantitative basis for subsequent route berth selection.

[0045] Using three typical parking spaces in a city's urban commercial district, old residential areas, and main urban roads as the actual measurement scenarios, parking space operation data was collected to calculate weight parameters. Specific statistical data are shown in Table 1.

[0046] Table 1. Actual Measured Data on Parking Space Indicators and Weights for Three Types of Road Sections

[0047] B1 Business District Roads 0.92 0.88 0.95 0.9195 B2 Old residential areas 0.53 0.45 0.38 0.4675 B3 Urban main roads 0.21 0.25 0.18 0.2175

[0048] The cloud uses the calculated berth weight to represent the inspection revenue. Combined with the drone's cruising speed of 8m / s, cruising power consumption of 500W, and hovering power consumption of 800W, the flight energy consumption to the corresponding berth is calculated based on the distance between the berth and the drone's real-time position. The revenue-energy consumption ratio is obtained by dividing the berth weight by the total energy consumption of a single berth. The berths are sorted from high to low according to this ratio, and the berths with the higher ratio are selected as the initial inspection points. This realizes the initial selection logic of prioritizing the inspection of high-value berths.

[0049] The cloud platform calculates the drone's maximum range based on its current remaining battery power, rated battery capacity, 500W cruise power, 800W hovering power, and 8m / s cruise speed. This range is used to define the geographical boundaries for the current inspection area, eliminating initially selected parking spaces outside the designated area. A greedy heuristic algorithm is first used to connect points in order of benefit-energy consumption ratio to generate an initial inspection path. Then, a 2-opt local optimization algorithm is used to iteratively select two points in the path and swap their positions. The total cruise energy consumption and hovering energy consumption of the entire route are recalculated. The total energy consumption values ​​before and after optimization are continuously compared. Under the constraint that the total energy consumption is always lower than the remaining available battery power, redundant backtracking routes are repeatedly eliminated. After iterative convergence, the final inspection route with energy consumption compliance and optimal path is obtained.

[0050] In one possible implementation, step S300 further includes:

[0051] The airborne system uses a lightweight neural network to detect vehicles and performs three processing actions based on confidence level: skipping, local evidence storage, and selecting target areas and transmitting images.

[0052] Specifically, the UAV is equipped with a lightweight target detection neural network that has been accelerated and optimized with INT8 integer quantization and TensorRT tensor inference. The onboard main controller scales the real-time aerial footage to a standard size, performs image preprocessing, and then sends it to the network for inference to output vehicle target detection boxes and corresponding confidence scores. The system has two preset confidence thresholds to divide three types of processing logic: when the detection confidence is below the lower threshold, it is determined that there are no valid illegal vehicles in the image, and subsequent processing is skipped without saving or uploading the image; when the detection confidence is above the upper threshold, it is determined that the vehicle target detection result is conclusive, and only the complete original image is stored locally onboard to preserve evidence, without transmitting the image to the edge nodes; when the detection confidence falls in the middle range between the upper and lower thresholds, the region of interest (ROI) is cropped outward with a fixed pixel margin based on the vehicle detection bounding box, compressed into a small-volume JPEG image in the joint image expert group format, and sent to the nearest roadside edge node via the onboard wireless communication link for secondary refined recognition and analysis.

[0053] In one possible implementation, step S400 further includes:

[0054] Step S410: The light pole reference station provides differential positioning data to the UAV.

[0055] Step S420: Construct a transformation matrix by combining camera parameters and UAV pose parameters.

[0056] Step S430: Based on the transformation matrix, perform perspective transformation on the original image to generate an orthophoto image, and overlay a berth marking layer.

[0057] Specifically, the RTK real-time dynamic differential positioning reference positioning equipment installed on the support platform of each modified light pole continuously receives raw observation data from multi-band satellite navigation. The equipment calculates differential correction information based on the accurate on-site coordinates of the calibrated light poles and broadcasts differential messages through the platform's wireless radio frequency transmission module. The airborne positioning receiver of the UAV conducting aerial inspections searches for differential data sent by nearby reference stations in real time, and completes differential calculation by combining it with the satellite observation values ​​of the UAV itself. This corrects the position deviation caused by the airborne single-point positioning and brings the UAV's positioning accuracy to the centimeter level, providing a continuous and reliable differential positioning data source for UAV flight path cruising, fixed-point hovering, landing alignment, and aerial photography evidence collection.

[0058] The intrinsic parameters of the airborne camera obtained from the previous calibration are retrieved, including the lateral focal length fx, longitudinal focal length fy, the horizontal coordinate of the image principal point cx, and the vertical coordinate of the image principal point cy, which form the camera intrinsic parameter matrix. Then, the spatial three-dimensional coordinates acquired by the UAV's real-time dynamic differential positioning and the three sets of attitude data (roll angle, pitch angle, and yaw angle) output by the IMU inertial measurement unit are read. The camera spatial rotation matrix is ​​generated based on the three-axis Euler angle conversion. The translation vector is constructed by combining the equipment's ground elevation. The camera extrinsic parameter projection matrix is ​​obtained by combining the intrinsic parameter matrix, rotation matrix, and translation parameters. Then, the homography transformation matrix is ​​solved by fusing the image image points and the ground coordinates using the DLT direct linear transformation algorithm. The mapping transformation matrix from the aerial image pixel coordinate system to the ground geographic coordinate system is fully constructed, providing the matrix operation foundation for subsequent aerial image orthorectification.

[0059] The edge nodes call the solved homography transformation matrix and use the perspective transformation algorithm to resample and map the original ROI region of interest image returned by the UAV pixel by pixel, eliminating the geometric distortion caused by the tilted aerial photography, and generating a standard orthophoto with a unified GSD ground sampling distance and a single pixel corresponding to the actual size specifications. The system retrieves the pre-stored berth vector geographic data in the background and accurately overlays the berth edge line and berth number layer onto the corresponding position of the orthophoto image based on the geographic coordinate registration relationship, intuitively presenting the relative positional relationship between the vehicle and the berth boundary, and providing a visual standard evidence image for berth occupancy determination and illegal evidence collection.

[0060] In one possible implementation, step S500 further includes:

[0061] Step S510: Summarize the images, license plates, and spatiotemporal location information to generate an evidence data package.

[0062] Step S520: Perform a hash digest operation on the evidence data packet, generate a digital signature using asymmetric encryption, and write the signed evidence data into the blockchain.

[0063] Specifically, the edge node filters standardized orthophoto images after homography transformation correction, complete license plate number data output by the LPRNet license plate recognition network, and the latitude, longitude, elevation, and precise capture time of the captured images synchronously collected by RTK real-time dynamic differential positioning, combined with the judgment conclusion of parking space over-line or lane occupation, are integrated and classified according to the agreed encapsulation format. After removing redundant and invalid fields, they are uniformly encapsulated into a structured evidence data package, which completely collects all elements of evidence collection information and provides a standardized data source for subsequent hash encryption and on-chain evidence storage.

[0064] Edge nodes first use the SHA-256 hash algorithm to perform digest operations on the complete evidence data packet byte by byte. This algorithm iteratively compresses the original data in 512-bit blocks, and outputs a unique hash digest of fixed length 256 bits after multiple rounds of logical operations and constant iterations. Any change to any field in the original data packet will directly cause a change in the digest value. Then, the RSA-2048 asymmetric encryption algorithm is used to encrypt the aforementioned hash digest using the 2048-bit private key stored locally on the edge node to generate a digital signature. The RSA algorithm generates a public-private key pair based on the large prime number factorization problem. The private key is only used for signing, and the public key is used for subsequent signature verification. After the signature is completed, the original evidence packet, hash digest, and digital signature are uploaded to the Hyperledger Fabric consortium blockchain. A customized evidence storage smart contract is called, and the legality of the data is verified through consensus among nodes. After the verification is passed, a new block is generated and written to the consortium blockchain distributed ledger. The evidence data is solidified and stored by relying on the uniqueness of hash and the tamper-proof characteristics of asymmetric signature.

[0065] In one possible implementation, step S700 further includes:

[0066] Step S710: Identify the inspection drones belonging to each berth using a hash modulo method.

[0067] Step S720: After identifying the faulty drone, rehashed and reassigned the uninspected berths to normal drones to continue operations.

[0068] Specifically, the cloud-based scheduling system extracts the unique code for each berth as the hash input source and uses the MurmurHash3 hash algorithm to perform data processing. This algorithm relies on multiple rounds of shift, XOR, and constant perturbation operations to perform discrete hash mapping on the berth code, outputting irregular integer hash values. Then, using the total number of currently online and normally inspecting drones as the modulo divisor, the hash result is used to perform a modulo operation on the number of drones. Based on the modulo remainder, the corresponding numbered drone is bound one by one. The remainder value range matches the range of drone numbers in use. Relying on the discrete characteristics of hashing, the berth tasks in the entire area are evenly divided into segments, avoiding the problem of task clustering and establishing a fixed berth-drone mapping relationship for routine inspections.

[0069] The cloud relies on periodic heartbeat messages to continuously verify the online status of drones. After determining that the target drone is offline due to a fault, it extracts the codes of all uninspected berths within the jurisdiction of the faulty drone. It then uses the MurmurHash3 hash algorithm to perform shift, XOR, and constant perturbation hash operations on the berth codes to obtain discrete integer hash values. The modulo divisor is changed to the total number of normal drones remaining after removing the faulty drone. The hash operation result is used to re-modulate the new divisor, and the normal drones with corresponding numbers are matched according to the modulo remainder. This completes the secondary redistribution of the faulty berth task. The drone receiving the task updates its own inspection list and continues to complete the inspection of the remaining berths. The unified hash rule ensures that the task redistribution logic is consistent and the redistribution results are evenly distributed.

[0070] Example 2, based on the same inventive concept as the drone docking and charging optimization method based on roadside light pole reuse in the previous examples, such as... Figure 2 As shown, this application provides an optimized drone parking and charging system based on the reuse of roadside light poles. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0071] The parking spot construction module 10 is used to modify existing roadside light poles to create parking spots.

[0072] The berth weight parameter determination module 20 is used to determine the berth weight parameters in the cloud based on the basic berth data, and to plan the inspection route by combining real-time berth data and the power status of the drone.

[0073] The target detection module 30 is used for the UAV to perform vehicle target detection along the planned inspection route, and selects the best local image based on the detection results and sends it to the nearest edge node.

[0074] The orthorectification processing module 40 is used to complete license plate recognition and parking space occupancy determination after the edge node receives the image, and to perform orthorectification processing on the acquired image based on the positioning data.

[0075] The on-chain storage module 50 is used to package the evidence collection data at the edge node, perform encryption and signature operations, and then store it on the blockchain.

[0076] The wireless charging module 60 is used to fly to the nearest lamppost docking point when the drone's battery level meets preset conditions, and lands wirelessly guided by positioning and vision.

[0077] The drone monitoring module 70 is used to monitor the drone's operating status in real time on the cloud. When a drone malfunctions, it transfers the unfinished inspection tasks based on the berth hash allocation rule.

[0078] Furthermore, the system is also used to implement the following functions:

[0079] The existing light poles were structurally inspected, and the support platform was fixed with reinforcement components. A charging transmitter, a reference positioning device, a light guide, and positioning markers were installed on the support platform. Power supply lines and protective devices were laid out, and electrical performance testing was completed.

[0080] Furthermore, the system is also used to implement the following functions:

[0081] The berth weights are calculated based on various statistical indicators of the berths; the berths are initially selected for inspection according to the relationship between the berth weights and flight energy consumption; the available inspection range is determined in combination with the remaining power, and the inspection route is adjusted through a path optimization algorithm.

[0082] Furthermore, the system is also used to implement the following functions:

[0083] The airborne system uses a lightweight neural network to detect vehicles and performs three processing actions based on confidence level: skipping, local evidence storage, and selecting target areas and transmitting images.

[0084] Furthermore, the system is also used to implement the following functions:

[0085] Differential positioning data is provided to the UAV by the light pole reference station; a transformation matrix is ​​constructed by combining camera parameters and UAV pose parameters; an orthophoto image is generated by perspective transformation of the original image based on the transformation matrix, and a berth marking layer is superimposed.

[0086] Furthermore, the system is also used to implement the following functions:

[0087] The evidence data package is generated by summarizing images, license plates, and spatiotemporal location information; a hash digest operation is performed on the evidence data package, and a digital signature is generated using asymmetric encryption; the signed evidence data is then written into the blockchain.

[0088] Furthermore, the system is also used to implement the following functions:

[0089] The inspection drones belonging to each berth are assigned by hash modulo method; after identifying the faulty drone, the uninspected berths are re-hash-assigned to the working drones to continue the operation.

[0090] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Specific embodiments of this specification have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0091] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0092] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An optimized method for drone parking and charging based on the reuse of roadside light poles, characterized in that, The method includes: Modify existing roadside light poles to create light pole docking points; The cloud platform determines berth weight parameters based on berth basic data and plans inspection routes by combining real-time berth data and drone battery status. The UAV conducts target detection of the airborne model vehicle along the planned inspection route, and selects the best local images based on the detection results to send to the nearest edge node; After receiving the image, the edge node completes license plate recognition and parking space occupancy determination, and performs orthorectification processing on the acquired image based on the positioning data. The edge node packages the evidence collection data and performs encryption and signing operations before storing it on the blockchain; When the drone's battery level meets the preset conditions, it flies to the nearest lamppost docking point, lands using positioning and visual guidance, and wirelessly charges. The cloud platform monitors the drone's operational status in real time, and in the event of a drone malfunction, it transfers the unfinished inspection tasks based on the berth hash allocation rule.

2. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, Modify existing roadside light poles to create light pole docking points, including: The original light poles underwent structural testing, and the load-bearing platform was fixed with reinforcement components. The carrying platform is equipped with a charging transmitter, a reference positioning device, a light guide, and positioning markers. Install power supply lines and protective devices, and complete electrical performance testing.

3. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, Planning inspection routes by combining real-time berth data and drone battery status, including: Calculate berth weights based on various statistical indicators of berths; Initially select inspection berths based on the relationship between berth weight and flight energy consumption; The available inspection range is determined based on the remaining power, and the inspection route is adjusted using a path optimization algorithm.

4. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, Based on the detection results, select the best local images and send them to the nearest edge nodes, including: The airborne system uses a lightweight neural network to detect vehicles and performs three processing actions based on confidence level: skipping, local evidence storage, and selecting target areas and transmitting images.

5. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, Orthorectification processing is performed on the acquired images based on positioning data, including: Differential positioning data is provided to the UAV from the light pole reference station; A transformation matrix is ​​constructed by combining camera parameters and UAV pose parameters; An orthophoto image is generated by performing perspective transformation on the original image based on the transformation matrix, and a berth marking layer is superimposed on it.

6. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, After packaging the forensic data and performing cryptographic signing operations, it is stored on the blockchain, including: Generate evidence data packages by summarizing imagery, license plate information, and spatiotemporal location information; The evidence data packet is hashed and digested, and a digital signature is generated using asymmetric encryption. The signed evidence data is then written into the blockchain.

7. The optimized method for drone parking and charging based on roadside light pole reuse as described in claim 1, characterized in that, Transferring incomplete inspection tasks based on berth hash allocation rules, including: The inspection drones belonging to each berth are identified using a hash modulo method; After identifying the faulty drone, the uninspected berths are re-hashed and assigned to working drones to continue operations.

8. An optimized drone parking and charging system based on the reuse of roadside light poles, characterized in that, The system is used to implement the drone docking and charging optimization method based on roadside light pole reuse as described in any one of claims 1-7, the system comprising: The parking spot construction module is used to modify existing roadside light poles to create parking spots. The berth weight parameter determination module is used to determine the berth weight parameters in the cloud based on the basic berth data, and to plan the inspection route by combining real-time berth data and the drone's battery status. The target detection module is used to detect airborne model vehicles along the planned inspection route of the UAV, and select the best local images based on the detection results and send them to the nearest edge node. The orthorectification processing module is used to complete license plate recognition and parking space occupancy determination after the edge node receives the image, and to perform orthorectification processing on the acquired image based on the positioning data. The on-chain storage module is used to package the evidence collection data at the edge node, perform encryption and signing operations, and then store it on the blockchain. The wireless charging module is used to fly to the nearest lamppost docking point when the drone's battery meets the preset conditions, and lands and charges wirelessly by relying on positioning and vision guidance. The drone monitoring module is used to monitor the drone's operating status in real time in the cloud. When a drone malfunctions, it transfers the unfinished inspection tasks based on the berth hash allocation rule.

Images