An integrated device for dynamic training of large-scale UAV visual models with adaptive environment

Abstract

The present invention discloses an integrated, adaptive, large-scale visual model dynamic training device for unmanned aerial vehicles (UAVs). The device comprises a UAV, a connecting base disposed at the bottom of the UAV, and a mounting mechanism disposed at the bottom of the connecting base. The mounting mechanism comprises a mounting box, a flow guide box, a first wire trough, a first flow guide trough, a second flow guide trough, several air inlets, tempered glass, a position limiting mechanism, and a drag reduction mechanism. This solution effectively improves the device's heat dissipation efficiency, reduces wind resistance during flight, ensures endurance, and extends its service life.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and in particular to an integrated device for dynamic training of a large-scale UAV visual model capable of self-adapting to an environment. Background Art

[0002] The adaptive, large-scale, dynamic training and inference system for UAV vision is an intelligent visual processing system integrated into a drone. It uses multi-lens sensors to capture real-time environmental images and leverages a built-in large-scale computing unit to simultaneously perform real-time reasoning and dynamic training for visual tasks (such as target recognition, obstacle avoidance, and scene understanding). Its core capabilities include automatically optimizing model parameters for varying environments (such as bright light, haze, and nighttime conditions), learning new scene features while in flight, and adaptively evolving visual algorithms without relying on the cloud. This system is primarily used in scenarios requiring environmental adaptability and real-time decision-making, such as drone inspections, disaster monitoring, and intelligent logistics, enhancing drones' autonomous capabilities in complex environments.

[0003] Part of the existing adaptive environment UAV visual large model dynamic training integrated device uses an external plug-in to place the data processor at the bottom of the drone and protect it, which can make it more convenient to maintain and improve data transmission efficiency. However, in actual use, the drone with a large protective shell will increase energy consumption due to the large wind resistance during flight, which is not conducive to long-term work. In addition, the heat dissipation effect is average. After long-term use, heat accumulation will cause data processor failure.

[0004] For this, a solution is needed. Summary of the Invention

[0005] (1) Technical problems solved

[0006] In view of the shortcomings of the existing technology, the present invention provides an integrated device for dynamic training of a large-scale UAV visual model with adaptive environment to solve the problems raised in the above background technology.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] An integrated device for dynamic training of a large-scale visual model of an unmanned aerial vehicle (UAV) in an adaptive environment includes a UAV, a connecting seat, and a mounting mechanism. The connecting seat is arranged at the bottom of the UAV, and the mounting mechanism is arranged at the bottom of the connecting seat.

[0010] The installation mechanism includes an installation box, a guide box, a wire trough 1, a guide trough 1, a guide trough 2, a plurality of air inlet holes, tempered glass, a limiting mechanism and a drag reduction mechanism. The guide box is arranged at the left end of the installation box, the wire trough 1 is arranged at the top of the installation box, the guide trough 1 is arranged at the top of the guide box in a left-right structure, the guide trough 2 is arranged at the top of the installation box corresponding to each position arranged on the guide trough 1, a plurality of air inlet holes are evenly arranged through the bottom of the guide trough 1, the guide trough 2 and the rear end of the installation box, the tempered glass is arranged oppositely at the top and bottom of the guide box, the limiting mechanism is arranged inside and below the installation box, and the drag reduction mechanism is arranged inside and outside the installation box.

[0011] Preferably, the installation box is a rectangular structure, the guide box is an isosceles trapezoidal structure, the installation box and the guide box are integrally formed, the left and right surfaces of the guide box are inclined toward the middle, the thickness of the rear end of the guide groove one is greater than the thickness of the front end, and the guide groove two is an isosceles trapezoidal structure and the height of the bottom of the rear end is greater than the height of the bottom of the front end.

[0012] Preferably, the orientation of each of the air inlet holes is consistent with the orientation of the bottom of the corresponding guide groove 1 and guide groove 2 and the orientation of the rear end of the installation box.

[0013] Preferably, the limiting mechanism includes a base plate, several supporting feet, several cameras, a second wire trough, a reinforcement strip, several screw holes one, several screw holes two, a clamping seat, a data processor, a wire plug board, a warning light and an integrated system, wherein the base plate is arranged at the bottom of the installation box, several supporting feet are evenly arranged at the bottom of the base plate, several cameras are arranged at the four corners of the bottom of the base plate, the second wire trough is arranged on the base plate corresponding to the position of each camera, the reinforcement strip is arranged on the top of the base plate in a front, back, left and right structure and is connected to the inner wall of the installation box, several screw holes one are evenly arranged on each reinforcement strip, the second screw hole is arranged on the base plate corresponding to the position of each screw hole one, the clamping seats are arranged on the front and back halves of the top of the base plate in a left and right structure opposite to each other, the data processor is arranged on the top of the base plate and between the four clamping seats, the wire plug board is arranged at the left and right ends of the data processor, the warning light is arranged at the front end of the top of the data processor, and the integrated system is arranged inside the data processor.

[0014] Preferably, the base plate and the supporting legs are L-shaped, and the reinforcement strip and the installation box are integrally formed.

[0015] Preferably, the pressing seat has a Z-shaped structure, and a shock-absorbing pad is provided on the side of the pressing seat facing the data processor. A wire groove three is provided in the middle position of the pressing seat, and a connecting plate is provided at the end of the wire groove three away from the data processor. A rotary sleeve is provided on the top of the pressing seat, and a screw hole three is provided inside the rotary sleeve, and the screw hole three passes downward through the top of the pressing seat, and a threaded pin is provided on the rotary sleeve.

[0016] Preferably, the connecting plate is in an isosceles trapezoidal structure and is integrally formed with the pressing seat and the rotary sleeve, and the bottom of the threaded pin abuts against the data processor.

[0017] Preferably, the integrated system includes a perception module, a data processing and computing module, a model training module, a navigation and control module, a communication module, a storage and heat dissipation module, a power management module and a middleware and algorithm library module. The perception module and the data processing and computing module are electrically connected. The data processing and computing module is electrically connected to the model training module, the navigation and control module, the communication module and the storage and heat dissipation module respectively. The navigation and control module is electrically connected to the communication module. The model training module is electrically connected to the storage and heat dissipation module. The power management module is electrically connected to the perception module, the data processing and computing module, the model training module, the navigation and control module, the communication module, the storage and heat dissipation module, the middleware and algorithm library module. The middleware and algorithm library module is electrically connected to the data processing and computing module, the model training module and the navigation and control module.

[0018] Preferably, the drag reduction mechanism includes side wing plates, a plurality of guide grooves, a reinforcement plate, a base plate, a conduction plate and a plurality of heat dissipation fins. The side wing plates are arranged at the left and right ends of the installation box, and a plurality of the guide grooves are arranged laterally and equidistantly at the upper and lower ends of the side wing plates. The reinforcement plates are arranged at the upper and lower ends of one side of each side wing plate close to the installation box. The base plate is arranged at one end of each side wing plate facing the installation box and located inside the side wall of the installation box. The conduction plate is arranged at one end of each base plate away from the adjacent side wing plate, and a plurality of the heat dissipation fins are evenly arranged at one end of the conduction plate away from the adjacent base plate.

[0019] Preferably, the side wing plate is a right-angled triangle structure and the obtuse-angled part is an arrow-shaped structure, the acute angle of the side wing plate faces the front end, the upper and lower edges of the side wing plate converge at a point from back to front, the guide groove three is an arc-shaped structure and is parallel to the left and right side walls of the installation box, the reinforcement plate is an arc-shaped structure and the upper and lower edges converge at a point from back to front, the shape of the base plate is a triangular structure, the heat dissipation fins are a rectangular parallelepiped structure and are evenly distributed on the conduction plate from back to front and from top to bottom, there is a gap between each two adjacent rows of conduction plates, and the side wing plates, reinforcement plates, base plate, conduction plate and heat dissipation fins are formed as one piece.

[0020] (3) Beneficial effects

[0021] The present invention provides an integrated device for dynamic training of large-scale UAV visual models that is adaptive to the environment. It has the following beneficial effects:

[0022] 1. Wind resistance optimization - reduced energy consumption and extended flight time: The side wing panels are designed as a right-angled triangle structure with an arrow-shaped obtuse angle. Combined with the three arc-shaped guide grooves, they allow airflow to be smoothly diverted from the side of the drone. Compared with traditional large protective shells, they can effectively reduce wind resistance, directly reduce flight energy consumption, and solve the problem of "difficulty in long-term operation".

[0023] 2. Enhanced Heat Dissipation - Efficient Temperature Control, Reliable Hardware: The guide box guides airflow into the mounting box through the air inlet, rapidly dissipating heat through the conductive plate and heat dissipation fins. Air enters through the front air inlet and dissipates heat through the top and rear air inlets. The segmented and spaced layout of the heat dissipation fins increases the heat exchange area, eliminating issues such as poor heat dissipation and heat accumulation.

[0024] 3. Fluid mechanics innovation of aerodynamic layout

[0025] The drag reduction mechanism of the streamlined structure of the side wing panels: The side wing panels adopt a right-angled triangle + arrow-shaped obtuse angle design, and their upper and lower edges converge into a point from back to front, reducing the boundary layer thickness of the airflow on the surface of the device by 20%; wind tunnel tests show that when the wind speed is 15m / s, the turbulence intensity on the surface of the device drops from 0.18 to 0.15, and the corresponding drag coefficient drops from 0.85 to 0.76, which is equivalent to reducing the windward area of ​​the traditional rectangular protective shell by 10.6%.

[0026] The airflow diversion effect of the three-way guide groove: The curved three-way guide grooves on the upper and lower ends of the side panels (parallel to the side walls of the installation box) create an airflow acceleration gradient of 0.3 m / s, increasing the airflow separation angle on the side of the device from 65° to 82°, reducing energy loss caused by eddy currents. In the earthquake-stricken area with strong winds of 18 m / s, this design reduced the UAV's roll angle fluctuation from ±15° to ±5°, improving flight stability by 60%.

[0027] 4. Balance between weight and strength of the integrated drag reduction structure

[0028] Lightweight Design: Components such as the side panels and reinforcement plates are integrally molded from aviation aluminum alloy, increasing overall weight by only 280g (compared to approximately 500g for a traditional protective case) while reducing wind resistance by 10.6%, achieving a dual optimization of "weight reduction and drag reduction." In cold chain logistics scenarios, this design increases the drone's payload capacity by 12% (allowing it to carry an additional 500g of cargo).

[0029] Strengthened structural strength: The arc-shaped convergent structure of the reinforcing plate forms a triangular support with the base plate, which increases the wind pressure resistance of the device from 1200Pa of the traditional solution to 1800Pa. It can withstand a force 7 gale (wind speed 13.9-17.1m / s), making it suitable for strong wind scenarios such as offshore wind power inspections.

[0030] 5. Thermal management innovation with three-dimensional heat dissipation channels

[0031] The airflow dynamics of the air inlet angles are designed to align with the bottom of the guide trough and the rear end of the enclosure, creating a three-dimensional convection pattern of "diagonal air intake from the front and vertical exhaust from the top / rear." When the drone flies at 12 m / s, the airflow velocity inside the enclosure reaches 1.8 m / s (compared to 1.2 m / s for conventional solutions). This improves the heat exchange coefficient on the data processor surface by 25%, reducing the temperature from 75°C to 59°C within 30 minutes (compared to 68°C for conventional solutions).

[0032] Optimized heat sink topology: The heat sink fins adopt a segmented layout (3mm spacing at the front and 7mm spacing at the back) with a 0.5mm thick copper foil coating on the conductive plate, improving heat transfer efficiency by 18%. In high-temperature environments (40°C), this design keeps the processor junction temperature below 85°C (compared to 95°C with traditional solutions), preventing the chip from triggering frequency throttling due to overheating.

[0033] 6. Dust-proof and anti-condensation design

[0034] Positive pressure air intake and dust prevention: The trapezoidal structure of the guide box creates a positive pressure of 5Pa at the air inlet. In an environment with a dust concentration of 800mg / m³, the amount of dust accumulation inside the installation box is reduced by 70% compared with traditional solutions, and the probability of heat sink fin blockage is reduced from 35% to 10%.

[0035] Condensation protection: The nano-coating on the surface of the heat sink (contact angle 110°) prevents the accumulation of condensation in low-temperature environments. At 90% humidity and 5°C, no water droplets condense inside the device (traditional solutions have obvious water droplets), avoiding the risk of circuit short circuits.

[0036] 7. Anti-vibration and quick maintenance design of the limit mechanism

[0037] Multi-stage shock-absorbing structure: The Z-shaped structure of the compression seat + shock-absorbing pad (Shore hardness 50A) form a "rigid support - elastic buffer" dual shock absorption. Under the vibration conditions of the drone's 6000rpm motor, the resonant amplitude of the data processor is reduced from 0.25mm to 0.08mm, and the hard disk read and write error rate is reduced from 0.5% to 0.05%.

[0038] Tool-free assembly and disassembly: The quick-release design of the screw sleeve and threaded pin reduces the data processor replacement time from 15 minutes to 2 minutes. Combined with the cable storage function of the cable duct three, maintenance efficiency is increased by 6 times, making it suitable for on-site equipment replacement during emergency rescue.

[0039] 8. Edge computing capabilities of the integrated system

[0040] Model lightweight inference: The middleware and algorithm library modules integrate lightweight neural network compression technology (such as knowledge distillation), reducing visual model parameters by 40%. Under the condition of 10TOPS computing power, the target detection speed is increased from 25FPS to 35FPS, meeting the real-time obstacle avoidance requirements of drones.

[0041] Offline training storage optimization: The storage and cooling module uses an NVMe solid-state drive + heat sink combination with a storage capacity of 1TB. It can support 72 hours of continuous 4K video recording and model training data storage, and the read and write speed is maintained at 1500MB / s (traditional solutions are reduced to 800MB / s due to insufficient heat dissipation). BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of the disassembled structure of the drone and the mounting mechanism of the present invention;

[0043] Figure 2 This is a schematic diagram of the installation mechanism structure of the present invention;

[0044] Figure 3 This is a schematic diagram of the bottom structure of the mounting mechanism of the present invention;

[0045] Figure 4 This is a schematic structural diagram of the second guide channel of the present invention;

[0046] Figure 5 This is a schematic diagram of the internal structure of the installation box of the present invention;

[0047] Figure 6 This is a structural schematic diagram of the compression seat of the present invention;

[0048] Figure 7 This is a rear perspective structural diagram of the installation box of the present invention;

[0049] Figure 8 This is a schematic diagram of the side wing plate and its surrounding structure of the present invention;

[0050] Figure 9 This is a schematic diagram of the conductive plate and heat dissipation fin structure of the present invention;

[0051] Figure 10 It is a schematic diagram of the integrated system module of the present invention.

[0052] In the figure, 1-UAV; 2-Connection base; 3-Mounting mechanism; 31-Mounting box; 32-Direction box; 33-Wire trough 1; 34-Direction trough 1; 35-Direction trough 2; 36-Several air inlets; 37-Tempered glass; 38-Limiting mechanism; 381-Base plate; 382-Several support legs; 383-Several cameras; 384-Wire trough 2; 385-Reinforcement strip; 386-Several screw holes 1; 387-Several screw holes 2; 388-Compression seat ;3881- shock-absorbing pad; 3882- wire trough three; 3883-connecting plate; 3884-screw sleeve; 3885-screw hole three; 3886-threaded pin; 389-data processor; 3810-wire plug board; 3811-warning light; 3812-integrated system; 39-drag reduction mechanism; 391-side wing plate; 392-several guide grooves three; 393-reinforcement plate; 394-base plate; 395-conduction plate; 396-several cooling fins. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0054] Example 1:

[0055] See also Figures 1-10 The embodiment of the present invention provides a technical solution to achieve this: it includes a drone 1, a connecting seat 2 and a mounting mechanism 3, the connecting seat 2 is arranged at the bottom of the drone 1, and the mounting mechanism 3 is arranged at the bottom of the connecting seat 2.

[0056] The installation mechanism 3 includes an installation box 31, a guide box 32, a wire trough 1 33, a guide trough 1 34, a guide trough 2 35, a plurality of air inlet holes 36, a tempered glass 37, a limiting mechanism 38 and a drag reduction mechanism 39. The guide box 32 is arranged at the left end of the installation box 31, the wire trough 1 33 is arranged at the top of the installation box 31, the guide trough 1 34 is arranged at the top of the guide box 32 in a left-right structure, the guide trough 2 35 is arranged at the top of the installation box 31 corresponding to each position set on the guide trough 1 34, a plurality of air inlet holes 36 are evenly arranged through the bottom of the guide trough 1 34, the guide trough 2 35 and the rear end of the installation box 31, the tempered glass 37 is oppositely arranged at the top and bottom of the guide box 32, the limiting mechanism 38 is arranged inside and below the installation box 31, and the drag reduction mechanism 39 is arranged inside and outside the installation box 31.

[0057] The installation box 31 is a rectangular parallelepiped, while the air guide box 32 is an isosceles trapezoid. The installation box 31 and air guide box 32 are integrally formed, with the left and right sides of the air guide box 32 tilted toward the center. The thickness of the rear end of the air guide groove 1 34 is greater than that of the front end. The air guide groove 2 35 is an isosceles trapezoid, with the rear end bottom height greater than the front end bottom height. The orientation of each air inlet 36 aligns with the bottom of the corresponding air guide groove 1 34 and air guide groove 2 35, as well as the rear end of the installation box 31.

[0058] The limiting mechanism 38 includes a bottom plate 381, a plurality of supporting legs 382, ​​a plurality of cameras 383, a second wire trough 384, a reinforcement strip 385, a plurality of screw holes 1 386, a plurality of screw holes 2 387, a pressing seat 388, a data processor 389, a wire plug board 3810, a warning light 3811 and an integrated system 3812. The bottom plate 381 is set at the bottom of the installation box 31, the plurality of supporting legs 382 are evenly arranged at the bottom of the bottom plate 381, and the plurality of cameras 383 are set at the four corners of the bottom of the bottom plate 381. The second wire trough 384 is set on the bottom plate 381 corresponding to the position of each camera 383. The reinforcement strips 385 are respectively arranged in a front, back, left and right structure. At the top of the base plate 381, connected to the inner wall of the installation box 31, several screw holes 386 are evenly distributed on each reinforcement strip 385. Screw holes 387 are arranged throughout the base plate 381, corresponding to the positions of each screw hole 386. Compressive seats 388 are arranged in a left-right configuration, opposite each other, at the front and back halves of the top of the base plate 381. A data processor 389 is located at the top of the base plate 381, between the four compressive seats 388. Wire connectors 3810 are located at the left and right ends of the data processor 389. A warning light 3811 is located at the front end of the top of the data processor 389. An integrated system 3812 is located within the data processor 389. The base plate 381 and support legs 382 are L-shaped, and the reinforcement strips 385 and installation box 31 are integrally formed.

[0059] The compression seat 388 has a Z-shaped structure. A shock-absorbing pad 3881 is provided on the side of the compression seat 388 facing the data processor 389. A third wire slot 3882 is cut in the middle of the compression seat 388. A connecting plate 3883 is provided on the end of the third wire slot 3882 facing away from the data processor 389. A screw sleeve 3884 is provided at the top of the compression seat 388. A third screw hole 3885 is provided within the screw sleeve 3884, extending downward through the top of the compression seat 388. A threaded pin 3886 is provided on the screw sleeve 3884. The connecting plate 3883 has an isosceles trapezoidal structure and is integrally formed with the compression seat 388 and the screw sleeve 3884. The bottom of the threaded pin 3886 abuts against the data processor 389.

[0060] The integrated system 3812 includes a perception module, a data processing and computing module, a model training module, a navigation and control module, a communication module, a storage and heat dissipation module, a power management module and a middleware and algorithm library module. The perception module and the data processing and computing module are electrically connected. The data processing and computing module is electrically connected to the model training module, the navigation and control module, the communication module and the storage and heat dissipation module respectively. The navigation and control module is electrically connected to the communication module. The model training module is electrically connected to the storage and heat dissipation module. The power management module is electrically connected to the perception module, the data processing and computing module, the model training module, the navigation and control module, the communication module, the storage and heat dissipation module, and the middleware and algorithm library module. The middleware and algorithm library module is electrically connected to the data processing and computing module, the model training module and the navigation and control module.

[0061] 1. Perception Module

[0062] Functional positioning: As the "visual nerve" of the system, it is responsible for collecting external environment data in real time and providing original information for subsequent processing.

[0063] Composition and working principle:

[0064] Hardware support: Includes cameras at the four corners of the bottom plate, which can collect multi-directional image data and are suitable for tasks such as target recognition and scene understanding.

[0065] Data collection range: In different scenarios (such as high-voltage transmission line inspections and earthquake-stricken area reconnaissance), it can identify 1mm bolt cracks and human posture within 50m, covering multi-spectral bands such as visible light and infrared.

[0066] Collaborative relationship: Directly connected electrically to the data processing and computing module, the collected raw data (such as images and videos) are transmitted to the module in real time for analysis.

[0067] 2. Data Processing and Computing Module

[0068] Functional positioning: The "brain center" of the system, responsible for the core functions of data analysis, algorithm operation and task scheduling.

[0069] Core Competencies:

[0070] Data processing: Perform pre-processing such as noise reduction and feature extraction on the raw data input by the perception module (such as 4K video and multispectral images), for example, processing NDVI vegetation index data in agricultural plant protection.

[0071] Computing power support: Integrating lightweight neural network compression technologies (such as knowledge distillation) increases the target detection speed from 25FPS to 35FPS under 10TOPS computing power conditions, meeting real-time decision-making needs.

[0072] Task scheduling: Coordinate the workflows of modules such as model training and navigation control, for example, prioritizing life sign recognition tasks during earthquake rescue.

[0073] Connection relationship: Electrically connected to the model training module, navigation and control module, communication module, storage and heat dissipation module respectively to achieve two-way data interaction.

[0074] 3. Model Training Module

[0075] Functional positioning: The system's "learning engine" supports online reasoning and dynamic training of visual models, achieving adaptive environmental evolution.

[0076] Technological innovation:

[0077] Dynamic Training: Without relying on the cloud, the system can learn new scene features while flying. For example, during power line inspections, the recognition accuracy of a new anti-vibration hammer increased from 75% to 92% (with only 5 flight trainings).

[0078] Model optimization: Offline training is performed using locally stored samples (such as earthquake debris and fire scene images). For example, when communications are interrupted, the accuracy of life sign recognition is increased from 60% to 88%.

[0079] Collaborative relationship: Interacts with the data processing and computing module to annotate data and optimize models, and connects with the storage and heat dissipation module to read training data and store model parameters.

[0080] 4. Navigation and Control Module

[0081] Functional positioning: The "motor nerve" of the system, responsible for the flight control, path planning and attitude stabilization of the drone.

[0082] Key features:

[0083] Flight control: Adjusts drone flight parameters based on environmental data (such as wind speed and obstacle location). For example, in strong winds, the roll angle fluctuation can be reduced from ±15° to ±5° through the side wing guide structure.

[0084] Path Planning: Combined with the navigation algorithms in the middleware algorithm library, it enables obstacle avoidance and terrain-mimicking flight in complex terrain (such as terraced fields and urban buildings), increasing the obstacle avoidance success rate from 85% to 98%.

[0085] Connection relationship: Interacts with the data processing and computing module to control strategies, and connects with the communication module to receive remote commands or return flight status.

[0086] 5. Communication Module

[0087] Functional positioning: The system's "information bridge", supporting two-way data transmission between drones and ground stations and the cloud.

[0088] Technical features:

[0089] Interference-resistant transmission: In cross-border, strong electromagnetic interference environments (such as offshore wind power and urban security), the middleware algorithm reduces the packet loss rate from 12% to 2%, ensuring stable data transmission.

[0090] Multi-mode communication: supports 4G / 5G, satellite communication and other modes. For example, when communication is interrupted in an earthquake-stricken area, it can switch to local storage mode to temporarily store data.

[0091] Collaborative relationship: exchanges instructions and status data with the navigation and control module, and connects with the data processing and calculation module to transmit environmental understanding results.

[0092] 6. Storage and Cooling Module

[0093] Functional positioning: The system's "data warehouse" and "temperature control system", taking into account both data storage and hardware protection.

[0094] Dual Action:

[0095] Storage capacity: Using NVMe solid-state drives with a capacity of up to 1TB, it can continuously store 72 hours of 4K video and training data, with a read and write speed maintained at 1500MB / s (traditional solutions are reduced to 800MB / s due to insufficient heat dissipation).

[0096] Heat dissipation design: Through the coordination of the conduction plate, heat dissipation fins (segmented and spaced apart) and the air inlet holes of the guide box, a three-dimensional heat dissipation channel of "front air intake - top / rear exhaust" is formed, reducing the data processor temperature from 75°C to 59°C (traditional solutions reduce it to 68°C).

[0097] Technological innovation: Nano-coating on the surface of the cooling fins (contact angle 110°) prevents dust clogging and condensation accumulation. In an environment with a dust concentration of 800mg / m³, the probability of cooling fin clogging is reduced from 35% to 10%.

[0098] 7. Power Management Module

[0099] Functional positioning: The "energy center" of the system, responsible for power distribution, energy consumption optimization and power monitoring of each module.

[0100] Core role:

[0101] Intelligent power supply: Provides unified power supply to all modules, including the perception module, data processing and computing module, to ensure stable operation of each component.

[0102] Energy consumption optimization: In low-power scenarios (such as long-distance flights in cold chain logistics), by dynamically adjusting computing power allocation, energy consumption can be reduced by 10%, and the cruising range can be extended from 42km to 51km.

[0103] Battery monitoring: Real-time monitoring of battery status, automatically switching to low-power mode when the battery level drops below 20%, extending operating time to 7.5 hours (6 hours with traditional solutions).

[0104] 8. Middleware and algorithm library module

[0105] Functional positioning: The system's "algorithm treasure house" and "coordinator", providing general algorithm support and inter-module protocol conversion.

[0106] Key Values:

[0107] Algorithm support: Integrates basic algorithms such as target detection and semantic segmentation, as well as industry-specific algorithms (such as wind power bolt crack detection and forest fire spread prediction). For example, it improves the leaf area index inversion accuracy to 94% in agricultural plant protection.

[0108] Protocol conversion: Unify the data interfaces and communication protocols of each module to achieve seamless collaboration among perception, computing, and control modules. For example, it can reduce data processing latency from 300ms to 240ms.

[0109] Connection relationship: Electrically connected to the data processing and computing module, model training module, and navigation and control module, providing algorithm support and protocol analysis services.

[0110] 9. Module Collaborative Workflow

[0111] Data collection and processing: The perception module collects environmental data → the data processing and calculation module preprocesses and extracts features.

[0112] Model reasoning and optimization: The model training module performs real-time reasoning based on feature data and dynamically trains the model using historical data (e.g., improving recognition accuracy).

[0113] Decision-making and control: The data processing module transmits the inference results to the navigation and control module → generates flight control strategies (such as obstacle avoidance and hovering).

[0114] Data storage and transmission: Key data is stored in the storage module and transmitted back to the ground station through the communication module; the power management module monitors energy consumption throughout the process.

[0115] Algorithm support: The middleware and algorithm library modules provide algorithm support for the entire process (such as lightweight model compression and path planning algorithms), ensuring efficient collaboration among various modules.

[0116] Through the organic integration of the above eight modules, the integrated system realizes the full-process intelligence from environmental perception, data processing to decision-making and control, significantly improving the adaptive operation capability of drones in complex scenarios.

[0117] Analyzing the above content: The guide box 32 guides airflow into the mounting box 31 through the air inlet 36, rapidly dissipating heat through the conductive plate 395 and the heat dissipation fins 396. Air enters through the front air inlet 36 and dissipates heat through the top and rear air inlet 36. The segmented and spaced layout of the heat dissipation fins 396 increases the heat exchange area, solving the problem of poor heat dissipation and heat accumulation.

[0118] The perception module collects data → the computing module processes it → the training module optimizes it → the navigation module executes it. The entire process is coordinated through the middleware / algorithm library and cooperates with the communication module to achieve "ground-ground collaborative training", allowing UAV 1 to dynamically adapt to complex environments (such as automatically adjusting model parameters in strong winds and high temperatures).

[0119] The mounting mechanism 3 is mounted on the bottom of the drone 1 through the connecting seat 2. The data processor 389 is mounted on the top of the base plate 381 by pressing the threaded pin 3886. The base plate 381 is mounted on the bottom of the reinforcement bar by bolts and nuts. When working, the light of the warning light 3811 can be emitted outward through the upper and lower tempered glasses 37, so that the front end protection capability of the mounting box 31 is stronger without affecting the observation of the warning light 3811.

[0120] Example 2:

[0121] See also Figures 1-10, an embodiment of the present invention provides a technical solution to achieve this: the drag reduction mechanism 39 includes side wing plates 391, a plurality of guide grooves 392, a reinforcement plate 393, a base plate 394, a conduction plate 395 and a plurality of heat dissipation fins 396, the side wing plates 391 are arranged at the left and right ends of the installation box 31, a plurality of guide grooves 392 are arranged equidistantly at the upper and lower ends of the side wing plates 391, the reinforcement plates 393 are arranged at the upper and lower ends of each side wing plate 391 close to the installation box 31, the base plate 394 is arranged at one end of each side wing plate 391 facing the installation box 31 and is located inside the side wall of the installation box 31, the conduction plate 395 is arranged at one end of each base plate 394 away from the adjacent side wing plate 391, and a plurality of heat dissipation fins 396 are evenly arranged at one end of the conduction plate 395 away from the adjacent base plate 394.

[0122] The side wing plate 391 has a right-angled triangle structure and the obtuse angle part has an arrow-shaped structure. The acute angle of the side wing plate 391 faces the front end. The upper and lower edges of the side wing plate 391 converge at a point from back to front. The guide groove three 392 has an arc structure and is parallel to the side walls of the left and right ends of the installation box 31. The reinforcement plate 393 has an arc structure and the upper and lower edges converge at a point from back to front. The base plate 394 has a triangular structure. The heat dissipation fins 396 have a rectangular structure and are evenly distributed on the conduction plate 395 from back to front and from top to bottom. There is a gap between each adjacent column of conduction plates 395. The side wing plates 391, reinforcement plates 393, base plate 394, conduction plates 395 and heat dissipation fins 396 are formed as one piece.

[0123] Analysis of the above content: The side wing panel 391 is designed as a right-angled triangle structure with an arrow-shaped obtuse angle. Combined with the arc-shaped guide groove 392, it allows the airflow to be smoothly diverted from the side of the drone 1. Compared with traditional large protective shells, it can effectively reduce wind resistance, directly reduce flight energy consumption, and solve the problem of "difficulty in working for a long time".

[0124] The present invention includes: 1-unmanned aerial vehicle; 2-connecting seat; 3-mounting mechanism; 31-mounting box; 32-flow guide box; 33-wire trough 1; 34-flow guide trough 1; 35-flow guide trough 2; 36-several air inlets; 37-tempered glass; 38-limiting mechanism; 381-bottom plate; 382-several supporting feet; 383-several cameras; 384-wire trough 2; 385-reinforcement strip; 386-several screw holes 1; 387-several screw holes 2; 388-pressing seat; 3881-shock-absorbing pad; 3882-wire trough 3; 3883-connecting plate; 3884-rotating sleeve; 3885-screw hole 3; 3886-threaded pin; 389-data processor; 3810-wire plug board; 3811-warning light; 3812-integrated system; 39-drag reduction mechanism; 391-side wing plate; 392-if Dry guide groove three; 393-reinforcement plate; 394-base plate; 395-conduction plate; 396-several heat dissipation fins. These components are all universal standard parts or components known to those skilled in the art. Their structures and principles can be known to those skilled in the art through technical manuals or conventional experimental methods. The problem solved by the present invention is that a part of the existing adaptive environment UAV visual large model dynamic training integrated device adopts an external plug-in method to place the data processor at the bottom of the UAV and protect it, which can be more convenient to maintain and improve data transmission efficiency. However, in actual use, the UAV with a large protective shell will increase energy consumption due to large wind resistance during flight, which is not conducive to long-term work. In addition, the heat dissipation effect is general. After a long time of use, heat accumulation will cause data processor failure. The present invention effectively improves the heat dissipation efficiency of the device, effectively reduces the wind resistance of the device during flight, ensures endurance, and extends the use time.

[0125] Furthermore, the innovative features of this solution are described.

[0126] (1) Structural innovation and working principle of the wind resistance optimization system

[0127] Aerodynamic structural design of side wing panels

[0128] The side wing panels provided at the left and right ends of the installation box of the present invention are in a right-angled triangle structure, and the obtuse-angled portion thereof adopts an arrow-shaped streamlined design, with the acute angle facing the front end and the upper and lower edges converging into a point from the back to the front. This structure replaces the right-angled edge of the traditional external device with a gradually shrinking guide surface, so that the separation point when the airflow contacts the device moves backward. The arc-shaped guide grooves three, which are evenly distributed laterally at the upper and lower ends of the side wing panels and are parallel to the side walls of the installation box, can guide the airflow to flow smoothly along the grooves and reduce the generation of vortices. For example, when the drone flies at a speed of 15m / s, the arrow-shaped side wing panels can extend the attachment time of the airflow on the surface of the device by 12%, thereby reducing the pressure difference resistance caused by airflow separation.

[0129] Synergistic effect of integrated drag reduction structure

[0130] The reinforcement plates at the upper and lower ends of the side panels, located near the mounting box, also feature an arc-shaped converging structure, forming a unified airflow guide surface with the side panels. This design reduces the turbulent area on the side of the mounting box by approximately 30%. Combined with the integrated structure of the baseplate, conduction plate, and heat dissipation fins, the overall outer contour of the device meets aerodynamic streamlined requirements. Compared to traditional rectangular protective shells, this solution reduces the drag coefficient by splitting the frontal airflow into two streams through the arrow-shaped structure of the side panels. After being combed through the guide grooves three times, the airflow flows smoothly along the sides of the device, avoiding the airflow separation and vortices caused by the right-angled edges of traditional structures, thereby reducing wind resistance by approximately 10.6%.

[0131] (2) Multi-dimensional innovative design of the heat dissipation enhancement system

[0132] Airflow guidance mechanism of the guide box and air inlet

[0133] The deflector box is an isosceles trapezoidal structure, integrally formed with the mounting box. Its left and right sides slope toward the center, with deflector troughs 1 and 2 located at the top. Air inlets are evenly distributed at the bottom and rear of the mounting box. The orientation of the air inlets aligns with the bottom of the deflector trough and the rear of the mounting box, creating a convection path of "front intake - top / rear exhaust." During drone flight, oncoming airflow is accelerated by the trapezoidal surface of the deflector box before entering the mounting box through the front air inlet, increasing flow velocity by approximately 18% compared to traditional solutions. For example, at a wind speed of 10 m / s, the design of deflector trough 1, which is thicker at the rear than at the front, creates an acceleration gradient of 0.2 m / s within the trough, ensuring efficient intake of the air through the air inlets.

[0134] Segmented spacing layout of heat dissipation fins

[0135] The heat dissipation fins on the conductive plates are rectangular in structure, evenly distributed from back to front and from top to bottom, with a 5mm gap between adjacent rows of conductive plates. This layout increases the heat exchange area by 22% while avoiding the airflow obstruction caused by the traditional dense arrangement of heat dissipation fins. Experimental data shows that segmented heat dissipation fins can reduce the turbulence intensity of air flowing through it by 15% and improve heat dissipation efficiency by 12%. For example, when the data processor power is 50W, the heat dissipation fins of this solution can stabilize the surface temperature at 59°C within 30 minutes, while the traditional solution, due to the dense arrangement of heat dissipation fins and the poor airflow, can reach a temperature of 68°C.

[0136] (3) Innovative design of mechanical structure and system integration

[0137] Shock absorption and fixing design of the limit mechanism

[0138] The clamping mount features a Z-shaped structure, with a shock-absorbing pad on the side facing the data processor. A screw-on sleeve and threaded pin secure the data processor to the top. This design reduces the vibration amplitude of the drone during flight by 35%, eliminating the vibration-induced contact problems associated with traditional bolt-mounted mounting. For example, when maneuvering at an acceleration of 8 m / s², the shock-absorbing pad keeps the data processor's vibration amplitude within 0.1 mm, compared to 0.3 mm with traditional rigid mounting.

[0139] Modular collaboration of integrated systems

[0140] The integrated system comprises eight modules: a perception module, a data processing and computing module, and a model training module. Data interaction is achieved through middleware and an algorithm library. For example, environmental data collected by the perception module is analyzed by the data processing module and then transmitted in real time to the model training module to optimize the visual algorithm. Compared to the serial processing model of traditional plug-in systems, decision latency is reduced by approximately 20%, from 300ms to 240ms.

[0141] Specifically, data comparison of relevant tests and actual applications of this solution.

[0142] (1) Comparison of wind resistance and energy consumption

[0143] Test items Traditional solution Solution of the present invention Data Description Drag coefficient 0.85 0.76 Wind tunnel test conditions: wind speed 15m / s, ambient temperature 20°C, device frontal area 0.03m². This solution reduces the drag coefficient by 10.6% due to the side wing guide structure, meeting the practical range of aerodynamic optimization for small UAVs. Flight energy consumption 220W 202W In the endurance test, equipped with the same visual task load (4K cameras × 4, real-time inference computing power of 10TOPS), this solution reduced the power system power consumption by 8.2% due to the reduced wind resistance. The energy consumption data was collected in real time through a power meter. Battery life 28min 31min Under the condition of a battery capacity of 11.1V / 5200mAh, the reduced energy consumption extends the flight time by 10.7%. This data is recorded in the drone flight control system log with an error of ±0.5min.

[0144] (2) Comparison of heat dissipation efficiency

[0145] Test items Traditional solution Solution of the present invention Data Description Data processor surface temperature 68℃ (after 30 minutes of operation) 59℃ (after 30 minutes of operation) The ambient temperature is 25°C, and the data processor power consumption is 50W, monitored by an infrared thermometer (accuracy ±1°C). This solution reduces the temperature by 9°C due to the synergistic effect of the guide box and heat sink fins, preventing computing power throttling due to overheating. Heat dissipation rate 1.2℃ / min 1.5℃ / min In the shutdown cooling test, the heat dissipation rate of this solution increased by 25% in the process of dropping the temperature from 60°C to 30°C. In the actual working scenario, the overall heat dissipation efficiency increased by about 12%. The data was recorded by a temperature sensor (sampling frequency 1Hz). High temperature failure rate 15% (continuous work for 2 hours) 3% (continuous work for 2 hours) In a simulated summer sunlight environment (ambient temperature 35°C, sunlight intensity 1000W / m²), the traditional solution has a 15% probability of processor overheating and restarting due to insufficient heat dissipation. This solution reduces the failure rate by 80% through efficient heat dissipation.

[0146] (3) Comparison of system response and stability

[0147] Test items Traditional solution Solution of the present invention Data Description Target recognition delay 300ms 240ms Visual task test: Identifying a circular target with a diameter of 50 cm and a distance of 10 meters. This solution reduces latency by 20% due to the coordinated optimization of integrated system modules. The timestamps of the data processing link were measured using an oscilloscope. Vibration amplitude 0.3mm 0.1mm When the drone is in hovering state, the surface vibration of the data processor is monitored by an acceleration sensor (range ±10g, accuracy 0.01g). The shock absorption design of this solution reduces the amplitude by 66.7%, reducing hardware fatigue damage.

[0148] Data Validity Description

[0149] Test environment controllability

[0150] Wind resistance tests are conducted in a closed wind tunnel laboratory with a wind speed control accuracy of ±0.5m / s. Heat dissipation tests are completed in a constant temperature and humidity environmental chamber (temperature control accuracy of ±0.5°C, humidity 50%±5%) to eliminate the impact of environmental parameter fluctuations.

[0151] Test equipment accuracy

[0152] The drag coefficient is obtained using a six-component force balance (measurement accuracy 0.1% FS); temperature is monitored using an infrared thermal imager (resolution 640×512, accuracy ±2°C); and energy consumption data is recorded using a high-precision power analyzer (sampling rate 10kHz, accuracy 0.5%) to ensure data traceability.

[0153] Representativeness of comparison objects

[0154] The traditional solution uses the mainstream external UAV visual processing device on the market (such as a certain brand's VT-200 model). Its rectangular protective shell structure and densely arranged heat dissipation fins are representative of the industry and form an effective contrast with the solution of the present invention.

[0155] Data rationality verification

[0156] Data such as the reduction in wind resistance (10.6%) and the improvement in heat dissipation efficiency (12%) were averaged through three sets of repeated experiments, with a standard deviation of less than 5%. This is consistent with the engineering practice of optimizing the mechanical structure of small UAVs and does not contain exaggerated data that violates the principles of aerodynamics or thermodynamics.

[0157] Regarding application examples 1-3 of this solution.

[0158] Application Example 1: Application of Dynamic Training Device in High-Voltage Transmission Line Inspection

[0159] 1. Scenario Challenges and Technical Requirements

[0160] During inspections of high-voltage transmission lines in mountainous areas, drones need to traverse complex terrain (such as canyons above 1,500 meters above sea level) and face three core problems: traditional external visual devices have high wind resistance, which shortens the drone's flight time to less than 25 minutes in level 5 winds (wind speed 8-10m / s); when the temperature in the equipment area exceeds 40°C in the summer afternoon, overheating of the data processor can easily cause distortion of the infrared temperature measurement data; when flying in areas with dense poles and towers, the vibration amplitude of the fuselage exceeds 0.3mm, causing the insulator crack recognition rate to drop below 70%.

[0161] 2. Scenario Adaptation of Technical Solutions

[0162] Field application of aerodynamic drag reduction design

[0163] During a 220kV power line inspection in Liangshan, Sichuan, the right-angled triangular side panels (arrow-shaped structures with blunt-angled ends) on the device's left and right sides reduced the drone's drag coefficient from 0.82 to 0.71 when traversing a 10m-wide gap between towers. Measured data showed that at a flight speed of 12m / s, power system power consumption dropped from 210W to 190W, flight time increased from 26 minutes to 30 minutes, and the number of towers covered in a single inspection increased from 18 to 22, improving inspection efficiency by approximately 22%. The isosceles trapezoidal structure of the guide box guides airflow along the first and second guide troughs. This reduced airflow separation by 30% when the drone rounded a mountain corner, preventing flight fluctuations caused by vortices.

[0164] Optimization of cooling system for high temperature environments

[0165] The air inlet holes on the top of the guide box and the mounting box form a "front intake - rear exhaust" channel. During inspections at an ambient temperature of 38°C, the surface temperature of the data processor remained stable at 59°C, 9°C lower than the 68°C of a traditional solution. The segmented spacing of the heat sink fins (5mm spacing between adjacent fins) increases the heat exchange area by 22%. When the processor consumes 50W, the heat dissipation rate reaches 1.5°C / min, 0.3°C / min faster than traditional densely packed heat sinks. In field tests at a substation, the device operated continuously for three hours without experiencing overheating or frequency throttling. However, after 1.5 hours, the traditional device experienced a surge in infrared image noise due to excessive temperatures, making it impossible to identify the 0.5mm hot zone on the cable clamp.

[0166] Intelligent inspection empowerment of the integrated system

[0167] The perception module's four-corner cameras work in conjunction with the data processing module, enabling it to identify a 1mm crack in a bolt at a distance of 10m, reducing target recognition latency from 300ms to 240ms. The model training module, through on-the-fly learning during inspections, has increased the recognition accuracy of new anti-vibration hammers from an initial 75% to 92% (with only five training flights). The navigation and control module, combined with middleware algorithms, limits the vibration amplitude of the drone to less than 0.1mm when traversing strong winds, a 67% reduction compared to traditional solutions. This improves the clarity of captured insulator images and increases the crack recognition rate from 70% to 95%.

[0168] 3. Actual application effect

[0169] In a power transmission line inspection project in the mountainous area of ​​Zhaotong, Yunnan, drones equipped with this device reduced the time it takes to complete a 30km line inspection from 4.5 hours to 3.8 hours, increasing daily operational efficiency by 15%. During the hot season (June-August), the inspection failure rate dropped from 12% to 3%, eliminating four repeat inspections per month due to equipment overheating and saving approximately 80,000 yuan in annual operation and maintenance costs. The device's dynamic training capabilities increased the system's recognition speed for hazards unique to mountainous areas, such as ice cover and tree obstructions, by 40%. The discovery time for a hazard on a section of line prone to ice accumulation was shortened from two days to five hours compared to traditional solutions, saving critical time for de-icing operations.

[0170] Application Example 2: Application of dynamic training device in emergency reconnaissance in earthquake-stricken areas

[0171] 1. Scenario Challenges and Technical Requirements

[0172] Reconnaissance in earthquake-stricken areas faces three difficulties: strong winds (15-20m / s) and turbulent airflow created by building debris cause excessive wind resistance in traditional devices, resulting in a probability of drone loss of control exceeding 20%; the cooling fins are easily clogged in dusty environments, and temperature fluctuations of the data processor exceeding 15°C can cause the image recognition algorithm to fail; when communications are interrupted, traditional fixed models cannot adapt to changes in the ruins scene, and the accuracy rate of identifying life signs is less than 60%.

[0173] 2. Emergency Response of Technical Solutions

[0174] Enhanced wind resistance of drag-reducing structures

[0175] The arrow-shaped streamlined design of the side panels reduced the drone's drag coefficient from 0.85 to 0.76 in 18m / s winds during rescue operations in an earthquake-stricken area in Gansu, improving flight stability by 55%. When the drone traversed narrow passages (8-10m wide) between collapsed buildings, the three guide grooves smoothly channeled airflow, preventing the oscillation of the fuselage caused by airflow separation in traditional devices (swing amplitude was reduced from ±1.5m to ±0.5m). Field tests showed that under comparable wind conditions, the drone equipped with this device could hover stably 5m above the ruins, while traditional devices were unable to maintain hover due to excessive wind resistance. The qualified rate of reconnaissance images increased from 50% to 90%.

[0176] Dust environment adaptation of cooling system

[0177] The air inlet orientation of the guide box (aligned with the airflow direction) creates positive air pressure, reducing dust accumulation. The segmented, spaced-apart layout of the cooling fins (5mm apart) prevents dust clogging. In an environment with a dust concentration of 1000mg / m³, heat dissipation efficiency only decreased by 8%, compared to a 35% decrease for traditional systems with densely packed fins. In field tests at a disaster area, after four hours of continuous operation, the data processor temperature remained stable at 62°C. However, in a traditional system, the temperature rose to 75°C due to clogged cooling fins, causing the accuracy of the person recognition algorithm to drop from 85% to 55%.

[0178] Emergency intelligent upgrade of offline training

[0179] The integrated system's model training module dynamically optimizes its model based on 200 locally stored debris samples during communication interruptions. At one rescue site, the system's initial confusion rate for identifying concrete blocks and human bodies was 40%. After 30 minutes of offline training, this confusion rate dropped to 10%, and the accuracy of identifying vital signs increased from 60% to 88%. The power management module and storage cooling module work together to automatically switch to low-power mode when the battery charge falls below 20%. This extends the device's continuous operating time from 6 hours to 7.5 hours on a full charge, providing more sustained on-site data support for rescue operations.

[0180] 3. Actual application effect

[0181] During a 2024 earthquake rescue operation, a swarm of drones equipped with this device completed a comprehensive reconnaissance of a 15km² disaster area, reducing the time required by traditional devices by 40% and increasing the efficiency of finding trapped people by 60%. The device's high stability improved the clarity of images of the rubble captured by rescuers, shortening the life detection time for a collapsed building from 2 hours to 40 minutes, directly assisting in the rescue of 23 trapped people. The system's dynamic training capabilities increased the system's recognition speed for new types of rubble structures (such as prefabricated slab bracing) by three times, eliminating missed detections often associated with traditional solutions due to model lag and increasing the rescue success rate by approximately 30%.

[0182] Application Example 3: Application of Dynamic Training Device for Cross-Border Cold Chain Logistics UAVs

[0183] 1. Scenario Challenges and Technical Requirements

[0184] Cross-border cold chain logistics faces four major challenges: long-distance flights (over 50km one way) require low-energy consumption designs, and traditional devices have high wind resistance, resulting in a range of less than 40km; the cooling system in low-temperature environments (-15℃ to -10℃) is prone to cause processor low-temperature failures; diverse cargo types (such as vaccines and fresh produce) require real-time identification of different packaging, and the adaptation time of traditional fixed models exceeds 2 hours; when cross-border communications fluctuate, insufficient data transmission stability can lead to a delivery address error rate of over 5%.

[0185] 2. Logistics scenario customization of technical solutions

[0186] Low wind resistance design for long-range optimization

[0187] The right-angled triangle structure of the side wing panels reduces the drag coefficient from 0.8 to 0.72 during cross-border flight (at 15m / s), lowering energy consumption by 10%. With an 11.1V / 6000mAh battery, the range is extended from 42km to 51km, meeting the requirements of a cross-border delivery route (48km one-way) along the China-Vietnam border. The curved layout of the guide trough reduces energy consumption fluctuations caused by airflow disturbances when the drone traverses valleys. Tests show that this device consumes 0.15kWh less power than traditional solutions at the same range, equivalent to completing one or two more cross-border deliveries per day.

[0188] Wide temperature environment adaptation of the cooling system

[0189] The air inlet holes in the guide box can adjust the air flow rate. In a -15°C environment, the data processor temperature is maintained at 25-30°C by controlling the amount of cold air intake. The segmented design of the heat sink fins prevents airflow blockage at low temperatures. When the ambient temperature is -12°C, the processor temperature fluctuation is controlled to ±2°C, while traditional devices have a 15% probability of crashing due to excessive heat dissipation. In a cross-border vaccine delivery test, the device operated continuously for 5 hours without any low-temperature failures, ensuring the real-time return of GPS positioning data. However, the traditional device lost its delivery position three times due to processor crashes after 3 hours.

[0190] Dynamic training for cargo identification optimization

[0191] The perception module, integrated with the model training module, has increased the recognition accuracy of insulated containers of varying sizes from an initial 78% to 96% in cross-border fresh produce delivery (after only three training runs). When the delivery shifted from vaccines to seafood, the system's adaptive adjustment of the recognition algorithm reduced the time required to adjust from two hours to 12 minutes. The communication module, combined with the middleware algorithm, reduced the packet loss rate from 12% to 2% in weak signal areas across borders (such as mountainous borders), ensuring real-time transmission of delivery addresses and temperature data, reducing the address error rate for cross-border deliveries from 5% to 0.5%.

[0192] 3. Actual application effect

[0193] In a cross-border cold chain logistics pilot along the China-Laos Railway, drones equipped with this device reduced one-way delivery time from 3.5 hours to 2.8 hours, increasing delivery efficiency by 25%. Due to reduced energy consumption, the daily delivery cost of a single drone dropped from 90 yuan to 78 yuan. Based on a fleet of 20 drones, this translates to an average annual cost savings of 86,400 yuan. Dynamic training capabilities have reduced the system's time to identify and adapt to cargo packaging standards in different countries (such as Laotian bamboo insulated boxes and Chinese foam boxes) from one day to half a day, significantly improving cross-border logistics response speed. The delivery time for a batch of emergency vaccines was reduced from 48 hours to 20 hours, ensuring the effectiveness of biological products.

[0194] Application Example 4: Application of dynamic training device in agricultural plant protection and crop monitoring

[0195] 1. Scenario Challenges and Technical Requirements

[0196] In modern agricultural operations, drone plant protection and crop monitoring face multiple technical bottlenecks: complex farmland terrain (such as hilly terraces and contiguous orchards) causes turbulent airflow during drone flight, and traditional external visual devices have large wind resistance. Under level 4 wind conditions (wind speed 5.5-7.9m / s), the flight time is shortened to 22 minutes, and the coverage area of ​​a single operation is less than 30 mu; when the field temperature exceeds 35°C at noon in summer, overheating of the data processor can easily lead to distortion of multispectral images, and the chlorophyll content inversion error exceeds 15%; the morphology of crops in different growth stages (such as rice tillering stage and corn jointing stage) is significantly different, and the pest and disease identification accuracy of traditional fixed models is only 65%-70%, which cannot meet the needs of precise plant protection.

[0197] 2. Scenario Adaptation of Technical Solutions

[0198] (1) Aerodynamic drag reduction and endurance optimization

[0199] The device's right-angled triangular side wings, used in an apple orchard in Shandong, reduced the drone's drag coefficient from 0.81 to 0.70 when flying between fruit trees (3-4m spacing). Test data showed that at a flight speed of 10m / s, power system power consumption dropped from 180W to 165W, flight time increased from 23 minutes to 28 minutes, and the area covered by a single operation increased from 28 mu to 35 mu, improving operational efficiency by approximately 25%. The curved layout of the diversion trough three guides airflow smoothly through the fruit tree canopy, reducing turbulence caused by foliage obstruction. This reduces the drone's roll angle fluctuation from ±12° to ±4° when flying over terraced fields, ensuring stable multispectral camera recording.

[0200] (2) Adaptation of the cooling system to high-temperature farmland

[0201] The air intake holes on the top of the guide box and the mounting box form a "front oblique air intake - rear vertical exhaust" channel. In a farmland setting with an ambient temperature of 37°C, the surface temperature of the data processor remained stable at 58°C, 9°C lower than the 67°C of a traditional solution. The segmented spacing of the heat sink fins (3mm spacing at the front and 7mm spacing at the rear) increases the heat exchange area by 20%. When the processor consumes 45W, the heat dissipation rate reaches 1.4°C / min, 0.2°C / min faster than traditional densely packed heat sinks. In field tests at a rice cultivation base, the device operated continuously for four hours without experiencing overheating or throttling. However, after two hours, a traditional device experienced increased noise in its multispectral data due to excessive temperatures, and the error in the NDVI (Normalized Difference Vegetation Index) calculation increased from 8% to 18%.

[0202] (3) Crop Recognition Optimization with Dynamic Training

[0203] The perception module's quad-spectrum camera (red, green, blue, and near-infrared) works in conjunction with the data processing module to identify crop disease spots as small as 0.5 cm² at an operating altitude of 15 meters, reducing target recognition latency from 280ms to 220ms. The model training module, through on-the-fly learning, has increased the accuracy of corn leaf blight recognition from an initial 72% to 89% (with only three training flights). The navigation and control module, combined with middleware algorithms, keeps the drone's vibration amplitude to within 0.12mm during simulated flight along terraced fields, a 60% reduction compared to traditional solutions. This improves the clarity of crop canopy images and increases the accuracy of leaf area index (LAI) retrieval from 82% to 94%.

[0204] 3. Actual application effect

[0205] In a 10,000-acre farmland demonstration area in Henan, a swarm of drones equipped with this device reduced the time it takes to monitor a single crop's full growth period from 15 days to 10, improving monitoring efficiency by 33%. During the hot season (July-August), the operational failure rate dropped from 10% to 2%, eliminating three repetitive operations per month due to equipment overheating, saving approximately 50,000 yuan in annual plant protection costs. The device's dynamic training capabilities increased the system's identification of the growth periods of different crop varieties (such as hybrid rice and conventional wheat) by 35%. This reduced water and fertilizer management decision-making time for a corn field from the traditional 24 hours to 8 hours, improving nitrogen fertilizer utilization efficiency by 12% and increasing yields by approximately 80 kg per mu.

[0206] Application Example 5: Application of dynamic training device in offshore wind power inspection

[0207] 1. Scenario Challenges and Technical Requirements

[0208] The offshore wind turbine inspection environment is extremely harsh. Drones must operate in strong winds (level 6-7, wind speed 10.8-17.1m / s), high salt fog (chloride ion concentration 0.5mg / m³) and complex sea conditions. They face three major technical challenges: traditional devices have high wind resistance and a flight time of less than 20 minutes in level 7 winds, making it impossible to complete a full-scale inspection of a single wind turbine (over 100m in height); the cooling fins are prone to corrosion and clogging in salt fog environments, and temperature fluctuations of the data processor exceeding 10°C can result in a bolt crack detection miss rate of over 15%; the defects of wind turbine towers, blades, nacelles and other components have diverse forms (such as blade cracks, loose bolts, and overheated bearings), and the comprehensive recognition accuracy of traditional fixed models is only 68%.

[0209] 2. Scenario Adaptation of Technical Solutions

[0210] (1) Strengthening wind resistance and drag reduction structures

[0211] During an offshore wind farm inspection, the arrow-shaped streamlined design of the side panels reduced the drone's drag coefficient from 0.85 to 0.76 in 15m / s winds, improving flight stability by 60%. When the drone orbited the blades (flying in a 5m radius circle), the three guide grooves created an acceleration gradient of 0.4m / s, increasing the airflow separation angle on the device's sides from 65° to 82° and reducing vortex-induced oscillation (from ±0.8m to ±0.3m). Field measurements showed that under comparable wind conditions, the drone with this device could hover stably 2m from the wind turbine tower, while conventional devices were unable to maintain a fixed position due to excessive wind resistance. The inspection image qualification rate increased from 45% to 88%.

[0212] (2) Optimization of heat dissipation and anti-corrosion systems

[0213] The positive pressure design (5Pa) of the air inlet of the guide box creates an airflow barrier in salt spray environments, reducing the amount of salt spray particles entering the installation box by 70%. The nano-coating (contact angle of 110°) on the surface of the heat sink fins prevents salt spray condensation droplets from adhering. In an environment with a chloride ion concentration of 0.8mg / m³, the probability of heat sink fin blockage was reduced from 35% to 8%. In field measurements at an offshore wind power project, after 5 hours of continuous operation, the data processor temperature remained stable at 60°C. In contrast, the temperature of a conventional device rose to 72°C due to corrosion and blockage of the heat sink fins, resulting in a missed detection rate for bolt cracks from 8% to 22%. The 0.5mm copper foil coating on the conductive plate improves heat transfer efficiency by 18%. Under high sunlight conditions (surface temperature of 45°C), the processor junction temperature is kept below 80°C, preventing overheating and triggering frequency reduction.

[0214] (III) Wind power defect identification in integrated systems

[0215] The perception module's 4K camera, coupled with an infrared thermal imager, can identify surface cracks as small as 0.2mm on blades at a distance of 30m, with an infrared temperature measurement accuracy of ±1°C. The model training module, through offline learning (using 500 wind turbine defect samples stored locally), has increased the accuracy of bearing overheating detection from an initial 75% to 91% (training time is only 40 minutes). The communication and navigation modules work together to reduce the data packet loss rate from 15% to 3% in strong electromagnetic interference environments at sea, ensuring real-time transmission of defect location coordinates (with an accuracy of ±0.5m). The wind turbine-specific detection algorithm in the middleware algorithm library has tripled the speed of tower weld defect identification, reducing the inspection time for a single wind turbine from 15 minutes to 8 minutes.

[0216] 3. Actual application effect

[0217] During the operation and maintenance of an offshore wind farm in Guangdong, drones equipped with this device reduced the time it takes to complete a single wind turbine inspection from 20 minutes to 10 minutes, increasing daily operational efficiency by 50%. Equipment failure rates in salt spray environments dropped from 18% to 4%, eliminating two equipment replacements per year due to salt spray corrosion, and saving approximately 120,000 yuan in annual operation and maintenance costs. Dynamic training capabilities have reduced the system's defect identification and adaptation time for new wind turbine equipment (such as flexible DC wind turbines) from one week to one day. This has enabled one wind farm to detect potential hazards in real time, instead of seven days after regular inspections. This has created a critical window for equipment maintenance and avoided downtime losses of approximately 300,000 yuan per incident due to bearing overheating.

[0218] Application Example 6: Application of Dynamic Training Device in Urban Security and Emergency Command

[0219] 1. Scenario Challenges and Technical Requirements

[0220] Urban security scenarios are highly complex. Drones need to operate in environments with high-rise canyon winds (wind speeds of 8-12m / s), strong electromagnetic interference, and dynamic changes in multiple targets. They face core technical challenges: traditional devices have high wind resistance and a flight time of less than 25 minutes when flying between buildings, which cannot meet the emergency duty requirement of 30 minutes / flight; densely populated urban areas have poor heat dissipation conditions, and data processor temperatures exceeding 70°C will cause the face recognition frame rate to drop from 30FPS to 15FPS, missing key targets; the scenes of urban emergencies (such as terrorist attacks and mass incidents) change rapidly, and the accuracy of abnormal behavior recognition of traditional fixed models is only 60%-65%.

[0221] 2. Scenario Adaptation of Technical Solutions

[0222] 1. Urban environment optimization of aerodynamic layout

[0223] The streamlined structure of the device's side wings reduces the drag coefficient from 0.83 to 0.73 and energy consumption by 12% when flying within urban buildings. With an 11.1V / 5200mAh battery, flight time is extended from 26 minutes to 30 minutes, meeting the 30-minute emergency response requirement for a single flight. The curved design of the third guide channel, parallel to the side walls of the installation box, reduces turbulence caused by airflow separation when the drone traverses the narrow passage (15-20m wide) between two tall buildings, reducing the roll angle fluctuation from ±10° to ±3° and ensuring stable tracking of the optoelectronic pod. Field measurements show that at a flight speed of 12m / s, the drone with this device consumes 20W less power in urban canyons than traditional solutions, enabling one to two additional patrol missions per day.

[0224] (2) Urban Heat Island Adaptation of Cooling Systems

[0225] The guide box directs airflow, creating a three-dimensional "front intake - top exhaust" convection pattern. Under the urban heat island effect (ambient temperature 38°C), the surface temperature of the data processor remains stable at 61°C, 9°C lower than the 70°C of a traditional solution. The "dense front, sparse back" layout of the heat sink fins (3mm spacing at the front, 7mm spacing at the back) increases the heat exchange area by 25%. When the processor runs a deep learning model (computing power of 8TOPS), the heat dissipation rate reaches 1.6°C / min, 0.4°C / min faster than a traditional solution. In a city security test, the device operated continuously for three hours without experiencing overheating or throttling, while a traditional device experienced a drop in facial recognition accuracy from 92% to 78% after 1.5 hours due to overheating.

[0226] (3) Intelligent security upgrade with dynamic training

[0227] The perception module's starlight-class camera, coupled with a lidar radar, can identify human gestures within 50 meters at night in 0.1 lux lighting conditions, reducing target detection latency from 250ms to 200ms. The model training module, through on-the-fly learning, has increased the accuracy of identifying crowd gatherings from an initial 70% to 88% (with just five training patrols). The navigation and control module, integrated with the city's digital twin map, has tripled the speed of drone path planning for obstacle avoidance, increasing the obstacle avoidance success rate from 85% to 98%. The crowd density estimation algorithm in the middleware algorithm library has reduced the warning time for abnormal crowd flow during large-scale event security from 30 seconds to 10 seconds.

[0228] 3. Actual application effect

[0229] During security operations in a provincial capital, a swarm of drones equipped with this device completed routine patrols of a 30-square-kilometer urban area, reducing time by 30% compared to traditional systems and shortening emergency response time from 8 minutes to 5. The equipment failure rate during the hot season (June-August) dropped from 12% to 3%, eliminating mission interruptions due to overheating by 5 per month and saving approximately 60,000 yuan in emergency support costs annually. Dynamic training capabilities have reduced the system's adaptation time for identifying new types of suspicious behavior (such as illegal drone flights and unusual vehicle parking) from 24 hours to 2 hours compared to traditional solutions. This has increased the effectiveness of early warning of suspicious individuals in a commercial center by 40%, directly helping police prevent 17 cases.

[0230] Application Example 7: Application of dynamic training device in forest fire fighting and fire monitoring

[0231] 1. Scenario Challenges and Technical Requirements

[0232] Forest firefighting scenarios require extremely high environmental adaptability of drones and face four major technical challenges: the gusty winds (wind speed 12-18m / s) generated by the complex terrain of the forest area (such as mountains and canyons) make the wind resistance of traditional devices too large, and the flight time is less than 20 minutes, which is unable to complete 50 square kilometers of fire reconnaissance; the high temperature at the fire scene (the ambient temperature exceeds 50°C at 500m from the fire source) causes the data processor to overheat and crash with a probability of more than 20%; the fire source identification accuracy of traditional visual models in smoky environments (visibility <50m) is less than 50%; the spread of forest fires changes dynamically and the traditional fixed model cannot update the fire line prediction in real time, with an error of more than 1 kilometer.

[0233] 2. Scenario Adaptation of Technical Solutions

[0234] (1) Wind resistance, drag reduction and long-endurance design

[0235] The arrow-shaped structure of the side panels reduces the drag coefficient from 0.86 to 0.75 when flying in forested areas. This extends the flight time from 18 minutes to 23 minutes in force 7 wind conditions, increasing the reconnaissance area from 35 square kilometers to 48 square kilometers, and improving reconnaissance efficiency by 37%. The airflow diversion effect of the third guide channel reduces turbulence-induced fluctuations in the drone's flight attitude when crossing valleys, reducing the pitch angle fluctuation from ±15° to ±5°, ensuring stable imaging for the infrared thermal imager. Tests show that at a speed of 15m / s, a drone equipped with this device consumes 30W less power in gusty winds than a conventional solution, equivalent to covering 13 square kilometers of additional forest area per reconnaissance.

[0236] (2) High temperature protection and heat dissipation enhancement

[0237] The positive pressure design (8Pa) of the air inlet in the guide box creates an airflow barrier in smoky environments, reducing smoke particle entry into the installation box by 60%. The nano-coating on the heat sink fins (contact angle 115°) prevents condensation from high-temperature smoke. At 300 meters from the fire source (ambient temperature 55°C), the data processor temperature remained stable at 65°C, while traditional devices, due to insufficient heat dissipation, rose to 80°C, causing distortion in the infrared thermal imager data. The segmented spacing of the heat sink fins, combined with the copper foil coating on the conductive plate, improves heat transfer efficiency by 20%. When the processor is running at full load (10TOPS of computing power), the temperature drops from 70°C to 62°C in 30 minutes, while traditional solutions can only reach 68°C.

[0238] (3) Intelligent fire identification based on dynamic training

[0239] The perception module's dual-light camera (infrared + visible light) works in conjunction with a lidar to identify high-temperature fires (temperature > 300°C) within a 1 km radius in smoky environments, reducing target positioning error from 50m to 15m. The model training module, through offline learning from fire samples (300 fire point images stored locally), has increased the accuracy of understory fire identification from an initial 55% to 85% (training time: 30 minutes). The navigation and control module, incorporating wind speed and direction data from the fire scene, updates the flight path in real time, increasing obstacle avoidance success rate from 80% to 95%. The forest fire spread prediction model in the middleware algorithm library has reduced the fire front advance prediction error from 1.2 km to 0.4 km, buying critical time for firefighting deployment.

[0240] 3. Actual application effect

[0241] During a forest fire rescue operation, a swarm of drones equipped with this device completed a full-coverage reconnaissance of a 200-square-kilometer fire area, reducing the time by 45% compared to traditional equipment and increasing fire location efficiency by 60%. The equipment failure rate in high-temperature, smoky environments dropped from 25% to 5%, eliminating three reconnaissance interruptions per rescue operation due to equipment failures and directly assisting in extinguishing 12 fires. Dynamic training capabilities have reduced the system's recognition and adaptation time for new fire behaviors (such as crown fires and flying fires) from one day to three hours compared to traditional solutions. The fire front prediction time for a forest fire was shortened from 30 minutes to 10 minutes, saving two hours of valuable time for establishing firebreaks and reducing the burned area by approximately 500 hectares.

[0242] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from all perspectives, the embodiments should be regarded as illustrative and non-restrictive. The scope of the present invention is defined by the appended claims, not the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be included within the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.

[0243] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. An integrated device for dynamic training of large-scale UAV visual models that is adaptive to the environment, characterized by: It comprises a drone (1), a connecting seat (2) and a mounting mechanism (3), wherein the connecting seat (2) is arranged at the bottom of the drone (1), and the mounting mechanism (3) is arranged at the bottom of the connecting seat (2); The installation mechanism (3) includes an installation box (31), a guide box (32), a line trough 1 (33), a guide trough 1 (34), a guide trough 2 (35), a plurality of air inlet holes (36), a tempered glass (37), a limit mechanism (38) and a drag reduction mechanism (39), wherein the guide box (32) is arranged at the left end of the installation box (31), the line trough 1 (33) is arranged at the top of the installation box (31), the guide trough 1 (34) is arranged at the top of the guide box (32) in a left-right structure, and the guide trough 2 (35) is arranged at the left end of the guide box (31). Each position corresponding to the guide groove 1 (34) is arranged at the top of the installation box (31), a plurality of the air inlet holes (36) are evenly arranged through the guide groove 1 (34), the bottom of the guide groove 2 (35) and the rear end of the installation box (31), the tempered glass (37) is oppositely arranged at the top and bottom of the guide box (32), the limiting mechanism (38) is arranged inside and below the installation box (31), and the drag reduction mechanism (39) is arranged inside and outside the installation box (31); The system further comprises: an integrated system (3812), wherein the integrated system (3812) comprises a perception module, a data processing and computing module, a model training module, a navigation and control module, a communication module, a storage and heat dissipation module, a power management module, and a middleware and algorithm library module, wherein the perception module and the data processing and computing module are electrically connected, the data processing and computing module is electrically connected to the model training module, the navigation and control module, the communication module, and the storage and heat dissipation module respectively, the navigation and control module is electrically connected to the communication module, the model training module is electrically connected to the storage and heat dissipation module, the power management module is electrically connected to the perception module, the data processing and computing module, the model training module, the navigation and control module, the communication module, the storage and heat dissipation module, the middleware and algorithm library module, and the middleware and algorithm library module is electrically connected to the data processing and computing module, the model training module, and the navigation and control module; The drag reduction mechanism (39) includes a side wing plate (391), a plurality of guide grooves (392), a reinforcing plate (393), a base plate (394), a conduction plate (395) and a plurality of heat dissipation fins (396), wherein the side wing plate (391) is arranged at the left and right ends of the installation box (31), a plurality of the guide grooves (392) are arranged laterally and equidistantly at the upper and lower ends of the side wing plate (391), the reinforcing plate (393) is arranged at the upper and lower ends of a side of each side wing plate (391) close to the installation box (31), the base plate (394) is arranged at one end of each side wing plate (391) facing the installation box (31) and located inside the side wall of the installation box (31), the conduction plate (395) is arranged at one end of each base plate (394) away from the adjacent side wing plate (391), and a plurality of the heat dissipation fins (396) are evenly arranged at one end of the conduction plate (395) away from the adjacent base plate (394).

2. The adaptive environment UAV visual large model dynamic training integrated device according to claim 1 is characterized by: The installation box (31) is a rectangular parallelepiped structure, and the guide box (32) is an isosceles trapezoidal structure. The installation box (31) and the guide box (32) are integrally formed. The left and right sides of the guide box (32) are inclined toward the middle. The thickness of the rear end of the guide groove (34) is greater than the thickness of the front end. The guide groove (35) is an isosceles trapezoidal structure, and the height of the bottom of the rear end is greater than the height of the bottom of the front end.

3. The adaptive environment UAV visual large model dynamic training integrated device according to claim 2 is characterized by: The orientation of each air inlet hole (36) is consistent with the orientation of the bottom of the corresponding guide groove 1 (34), guide groove 2 (35) and the orientation of the rear end of the installation box (31).

4. The adaptive environment UAV visual large model dynamic training integrated device according to claim 3 is characterized by: The limiting mechanism (38) includes a bottom plate (381), a plurality of supporting legs (382), a plurality of cameras (383), a second wire trough (384), a reinforcement strip (385), a plurality of first screw holes (386), a plurality of second screw holes (387), a pressing seat (388), a data processor (389), a wire plug board (3810), a warning light (3811) and an integrated system (3812). The bottom plate (381) is arranged at the bottom of the installation box (31), the plurality of supporting legs (382) are evenly arranged at the bottom of the bottom plate (381), the plurality of cameras (383) are arranged at the four corners of the bottom of the bottom plate (381), the second wire trough (384) is arranged on the bottom plate (381) corresponding to the position of each camera (383), and the reinforcement strip (385) is arranged in a front-back, left-right structure at the bottom of the installation box (31). The top of the base plate (381) is connected to the inner wall of the installation box (31), a plurality of screw holes (386) are evenly arranged on each reinforcement strip (385), and the screw holes (387) are arranged on the base plate (381) corresponding to the position of each screw hole (386). The pressing seats (388) are arranged in a left-right structure opposite to each other at the front and rear halves of the top of the base plate (381). The data processor (389) is arranged on the top of the base plate (381) and is located between the four pressing seats (388). The wire plug board (3810) is arranged at the left and right ends of the data processor (389), the warning light (3811) is arranged at the front end of the top of the data processor (389), and the integrated system (3812) is arranged inside the data processor (389).

5. The adaptive environment UAV visual large model dynamic training integrated device according to claim 4 is characterized by: The bottom plate (381) and the supporting legs (382) are L-shaped structures, and the reinforcing strip (385) and the installation box (31) are integrally formed.

6. The adaptive environment UAV visual large model dynamic training integrated device according to claim 5 is characterized by: The pressing seat (388) has a Z-shaped structure. A shock-absorbing pad (3881) is provided on the side of the pressing seat (388) facing the data processor (389). A wire groove three (3882) is provided in the middle position of the pressing seat (388). A connecting plate (3883) is provided at the end of the wire groove three (3882) away from the data processor (389). A rotary sleeve (3884) is provided on the top of the pressing seat (388). A screw hole three (3885) is provided inside the rotary sleeve (3884), and the screw hole three (3885) passes downward through the top of the pressing seat (388). A threaded pin (3886) is provided on the rotary sleeve (3884).

7. The adaptive environment UAV visual large model dynamic training integrated device according to claim 6 is characterized by: The connecting plate (3883) has an isosceles trapezoidal structure and is integrally formed with the pressing seat (388) and the rotary sleeve (3884), and the bottom of the threaded pin (3886) abuts against the data processor (389).

8. The adaptive environment UAV large-scale visual model dynamic training integrated device according to claim 1 is characterized by: The side wing plate (391) is in a right-angled triangle structure and the obtuse angle portion is in an arrow-shaped structure. The acute angle of the side wing plate (391) faces the front end. The upper and lower edges of the side wing plate (391) converge to form a point from the back to the front. The guide groove three (392) is in an arc-shaped structure and is parallel to the left and right side walls of the installation box (31). The reinforcing plate (393) is in an arc-shaped structure and the upper and lower edges converge to form a point from the back to the front. The base plate (394) is in a triangular structure. The heat dissipation fins (396) are in a rectangular parallelepiped structure and are evenly distributed on the conductive plate (395) from the back to the front and from the top to the bottom. There is a gap between each two adjacent rows of the conductive plates (395). The side wing plate (391), the reinforcing plate (393), the base plate (394), the conductive plate (395) and the heat dissipation fins (396) are integrally formed.

Images