Dynamic switching and self-adaptive compression system and method for multi-mode communication of low-altitude Internet of Things

Abstract

The invention discloses a dynamic switching and adaptive compression system and method for multi-mode communication of a low-altitude Internet of Things, and belongs to the technical field of intelligent ad hoc networks of unmanned aerial vehicles. The airborne end comprises an NVIDIA Jetson Xavier NX embedded computing platform, an airborne Raspberry Pi CM4-5G expansion board, an airborne ad hoc network radio station, a flight control computer and an optical pod camera; the ground end comprises a ground Raspberry Pi CM4-5G expansion board and a ground ad hoc network radio station; according to the low-altitude intelligent networking multi-mode communication dynamic switching and adaptive compression system and method, the advantages that an ad hoc network radio station is farther in transmission distance, independent of ground infrastructure, flexible in deployment and the like are utilized, the ad hoc network radio station serves as a supplementary communication means of a 5G public network, and therefore real-time online of an unmanned aerial vehicle is guaranteed; information sharing and collaborative operation between the unmanned aerial vehicles can be effectively realized, and the overall efficiency and collaborative capability of the system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle intelligent self-organizing networks, and in particular to a system and method for dynamic switching and adaptive compression of multi-modal communications in a low-altitude intelligent network. Background Art

[0002] In today's society, the "low-altitude economy" based on low-altitude airspace and dominated by general aviation and drone industries is being vigorously developed; the low-altitude economy has a broad application prospect in vertical industries such as emergency rescue, industry inspection, agricultural plant protection, public safety, geological exploration, environmental monitoring, film and television entertainment, and is expected to have a trillion-level market size. The development demand of the low-altitude economy has led to the emergence of the low-altitude intelligent network. The low-altitude intelligent network is an important infrastructure for realizing the intelligent interconnection of "man-machine-object" in the low-altitude airspace (3km / 1km), and can help realize the physical network space of networking, digitalization and intelligent operation of low-altitude business. By the end of 2023, there were 1.267 million drones registered with real names in my country, a year-on-year increase of 32.2%; in 2023, the cumulative flight hours of civil drones were 23.11 million, a year-on-year increase of 11.8%. Civil drones have been popularized in the fields of agriculture, forestry, animal husbandry and fishery, and entertainment aerial photography. As a key low-altitude new infrastructure, the low-altitude intelligent network is an essential basic information network facility to support the development of the low-altitude economy. Due to the current lack of low-altitude basic network facilities, a feasible approach is to slightly modify ground communication facilities (such as upgrading the existing base station software) or build a small number of new low-altitude communication base stations to achieve low-altitude three-dimensional networking coverage. However, in general, relying solely on traditional ground cellular network facilities is far from enough. How to achieve real-time online networking ("uninterrupted") of low-altitude aircraft such as drones and efficient return of corresponding data is a key issue that urgently needs to be solved in the low-altitude intelligent network, which is crucial to whether the low-altitude economy can take off.

[0003] To this end, the present invention designs a system and method for dynamic switching and adaptive compression of multi-modal communications in a low-altitude intelligent network, which can bring more technological innovation and application expansion possibilities to the communication coverage and computing power supply of the low-altitude intelligent network. Summary of the invention

[0004] The purpose of the present invention is to provide a low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system and method to solve the problems of the prior art that the traditional drone inspection system relies on manual operation, lacks real-time transmission and has limited onboard computing power.

[0005] To achieve the above-mentioned objectives, the present invention provides a low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system, including an airborne end and a ground end; the airborne end includes an NVIDIA Jetson Xavier NX embedded computing platform, an airborne Raspberry Pi CM4-5G expansion board, an airborne ad hoc network radio, a flight control computer and an optical pod camera; the ground end includes a ground Raspberry Pi CM4-5G expansion board and a ground ad hoc network radio.

[0006] Preferably, the NVIDIA Jetson Xavier NX embedded computing platform is equipped with a target detection model, a video compression module, a network status perception module, a data transmission module, and a route intelligent planning module.

[0007] Preferably, the target detection model adopts the YOLOv5 network structure, which is trained by a number of drone aerial image samples. The YOLOv5 network detects targets by dividing the input image into grids of different sizes and predicting the bounding box and category of the target in each grid. During the training process, according to the feedback of the loss function, the YOLOv5 network adjusts its weights to optimize the performance and accuracy of the model. During the detection process, the detection video is output for subsequent processing.

[0008] Preferably, the video compression module performs video compression by calling the ffmpeg tool in the onboard Raspberry Pi CM4-5G expansion board, and adjusts the compression rate by adjusting the CRF value when transmitting the video. The CRF parameter is used to control the video quality of H.264 encoding. CRF is the abbreviation of Constant Rate Factor, which is a quantization parameter used to control the compression rate and quality of video encoding. The value range of CRF is between 0 and 51, where 0 represents lossless compression (i.e., original video quality) and 51 represents the lowest quality and highest compression rate. The value of CRF is inversely proportional to the video quality and file size. Specifically, under normal circumstances, the smaller the CRF value, the higher the video quality, but the larger the file size; the larger the CRF value, the lower the video quality and the smaller the file size.

[0009] Preferably, the ground Raspberry Pi CM4-5G expansion board is used to establish an ad hoc network and 5G connection between the ground and the drone, sense the connection status of the current link in real time, switch the data transmission link according to the status, and transmit the current transmission quality to the NVIDIA Jetson Xavier NX embedded computer, so that it can perform video compression according to the value; and the ground Raspberry Pi CM4-5G expansion board uses the integrated 5G module to complete the connection of the 5G link, and at the same time connects to the ground ad hoc network radio through the network port, so that it completes the connection of the ad hoc network link with the airborne ad hoc network radio, and uses the ground-end Raspberry Pi CM4-5G expansion board to complete the reading and decompression of the transmitted video data for real-time playback.

[0010] Preferably, the airborne Raspberry Pi CM4-5G expansion board is used to sense the connection quality of the two links, and then transmit the data to the NVIDIA Jetson Xavier NX embedded computer. The NVIDIA Jetson Xavier NX embedded computer compresses the video, and after the compression is completed, it is transmitted to the ground end through the airborne Raspberry Pi CM4-5G expansion board. The two links are the ad hoc network link and the 5G link.

[0011] Preferably, the ground ad hoc network radio station and the airborne ad hoc network radio station are both used to establish an ad hoc network link at the airborne end and the ground end.

[0012] The present invention also provides a method for dynamic switching and adaptive compression of multi-modal communication in a low-altitude intelligent network, comprising the following steps: Step 1: The airborne end and the ground end establish a communication connection, and the airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board perceive the quality information of the two communication links in real time; Step 2: Specify the starting and ending points of the drone flight. The NVIDIA Jetson Xavier NX embedded computer will autonomously plan the flight path based on the environment and transmit the path information to the flight control computer. Step 3: During the drone cruise, the optical pod camera on board captures video data in real time, and then transmits the video data to the NVIDIA Jetson Xavier NX embedded computer; Step 4: The NVIDIA Jetson Xavier NX embedded computer obtains the link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board, determines the compression rate of the transmitted video based on the data, compresses the video data captured in step 3, and transmits the video data to the onboard Raspberry Pi CM4-5G expansion board after the compression is completed; Step 5. The airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board both select a video transmission link according to the connection status of the two current links, and transmit the video data from the selected link to the ground Raspberry Pi CM4-5G expansion board; Step 6: After the ground end receives the compressed video data, the ground Raspberry Pi CM4-5G expansion board decompresses the compressed video and plays it in real time.

[0013] Preferably, the specific process of step 4 is as follows: S41, the optical pod camera transmits the collected video data to the NVIDIA Jetson Xavier NX embedded computer; The S42, NVIDIA Jetson Xavier NX embedded computer obtains link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board and uses this data to determine the compression rate of the video transmission; S43, the video compression rate is divided into four gears. When the link signal-to-noise ratio is lower than 50dB, the video compression CRF value is 30; when the signal-to-noise ratio is 30dB to 60dB, the CRF value is 28; when the signal-to-noise ratio is 60dB to 70dB, the CRF value is 26; when the signal-to-noise ratio is above 70dB, the CRF value is 24; S44. After the video compression is completed, the NVIDIA Jetson Xavier NX embedded computer transmits the compressed video data to the onboard Raspberry Pi CM4-5G expansion board; S45, after receiving the data, the onboard Raspberry Pi CM4-5G expansion board transmits the data to the ground Raspberry Pi CM4-5G expansion board through the established link.

[0014] Preferably, the process of selecting the video transmission link in step 5 is as follows: S51. When the program starts, define the communication address. The self-organizing network link uses a fixed IPv4 address, and the addresses at both ends are 192.168.17.100 and 192.168.17.101 respectively. For the 5G link, due to the particularity of the 5G network, the IPv6 address of the link is randomly assigned each time the system is turned on, so the addresses at both ends are not fixed. S52, after the system is started, read the IPv6 address of the local end and record it for communication; S53, set a timer function, the function sends a connection establishment message to the other party's address every m seconds, called a "heartbeat packet"; S54, real-time capture of the heartbeat packets sent by the other party, and statistics of the received heartbeat packets, each time a new heartbeat packet is received, the counter is reset to keep the value of the counter at 0; S55. If the system does not receive the heartbeat packet sent by the other party within the predetermined time, the value of the counter will gradually increase; S56, when the value of the counter reaches a preset threshold (for example, no heartbeat packet is received for several seconds in a row), it is determined that the current communication link fails; S57, when the link fails, automatically print the link failure information, and trigger the link switching mechanism to switch the data transmission to the standby link that can work normally; S58. If the ad hoc network link fails, the system attempts to use the 5G link for data transmission; vice versa; S59: After detecting link failure, the system immediately switches the link, redefines the communication address, re-establishes the connection with the other party, and continues to send and receive data; S60: During the switching process, the system continuously monitors the status of the new link and makes further adjustments based on the actual situation; S61. If the failed link returns to normal after a period of time, the system will recapture the heartbeat packet sent by the other party and re-establish the connection of the link; S62: Determine whether to switch data transmission back to the restored link according to the link priority and current communication conditions, so as to optimize data transmission efficiency.

[0015] Therefore, the present invention adopts the above-mentioned low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system and method, which has the following beneficial effects: (1) Sensing, sensing, computing and control are integrated to adapt to different sensing loads. The present invention installs an embedded AI computer on a drone and uses the onboard embedded AI computer to perform preliminary detection and identification, which is expected to reduce the amount of returned data while improving processing accuracy and reliability. At the same time, in response to the problem that traditional drone inspection systems rely on manual operation, this system is combined with a drone intelligent path planning system to improve the intelligence of the entire system. Ultimately, sensing, sensing, computing and control are integrated to effectively solve problems such as insufficient real-time transmission and limited onboard computing power; (2) Multiple communication methods are integrated to ensure efficient data transmission and improve security. Considering the advantages of longer transmission distance, no reliance on ground infrastructure, and flexible deployment, the present invention integrates 5G public network and self-organizing network into drones, and uses self-organizing network as a supplement to 5G public network to meet the requirements of transmission distance and signal coverage, thereby ensuring that drones are online almost 24 hours a day; (3) Intelligent perception switching. Faced with increasingly complex scenarios, the processing and management capabilities that rely on the traditional manual rule pre-definition and execution have gradually been unable to meet the needs. Therefore, the present invention organically integrates ad hoc networks with 5G, enabling both ends to perceive the connection quality of the two communication links in real time, and use automated and intelligent systems and means to intelligently switch the current communication mode based on the quality information, thereby achieving a high degree of connection between devices; (4) Intelligent perception and processing of on-board data, and adaptive compression of video data after initial detection. The present invention carries an embedded intelligent computing system on the drone, and adds a pre-processing function on the basis of the original information collection and transmission functions of the drone. By sensing the real-time connection status of the link, the video data obtained by the optical pod camera is adaptively compressed after initial detection, thereby reducing the amount of returned data, reducing the bandwidth consumed by transmission, and improving the accuracy of information transmission; (5) Autonomous path planning for drones. By optimizing flight routes, the efficiency of mission execution is improved, and flight time and energy consumption are reduced. At the same time, flight safety can be enhanced, allowing drones to automatically avoid obstacles and reduce collision risks. While effectively reducing dependence on manual operations, drones can adapt to environmental changes and emergencies in real time and adapt to a variety of application scenarios.

[0016] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the multi-modal communication dynamic switching and adaptive compression system of the low-altitude intelligent network of the present invention; Figure 2 It is an overall flow chart of the method for dynamic switching and adaptive compression of multi-modal communication of low-altitude intelligent network of the present invention; Figure 3 is a link switching flow chart of an embodiment of the present invention; Figure 4 This is an application level diagram of the present invention. DETAILED DESCRIPTION

[0018] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1, a low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system, including airborne and ground ends; the airborne end includes the NVIDIA Jetson Xavier NX embedded computing platform, the airborne Raspberry Pi CM4-5G expansion board, the airborne ad hoc network radio, the flight control computer and the optical pod camera; this part uses a network switch to connect the NVIDIA NX development board, the airborne Raspberry Pi CM4-5G expansion board and a wireless transceiver module for connecting the ad hoc network into a system using a network port. The video compression part of this system uses the Jetson NX development board produced by NVIDIA to provide AI computing capabilities for the drone system. Its AI computing power reaches 100TOPS, with strong performance, suitable for various unmanned system perception scenarios, and can realize online detection and recognition of drones and adaptive compression of videos by sensing link status. At the same time, it is equipped with an onboard Raspberry Pi CM4-5G expansion board to connect the 5G communication module and the self-organizing network module. The 5G signal is received by the 5G antenna integrated on the onboard Raspberry Pi CM4, and the self-organizing network information is received by the wireless receiving module connected to the onboard Raspberry Pi CM4-5G expansion board through a switch. This module mainly plays the role of a transfer station. During the working process, it can sense and judge the connection status of the two links in real time, and independently select the transmission link according to the connection status, providing our equipment with high-speed and stable wireless connection, improving data transmission speed and network response performance, and realizing direct communication between devices and network self-management.

[0020] The ground side includes the ground Raspberry Pi CM4-5G expansion board and the ground ad hoc network radio; the two parts are connected through the network port. Its function is similar to that of the Raspberry Pi CM4-5G expansion board of the airborne part. The expansion board transmits and exchanges 5G data through the integrated 5G antenna, and the ground ad hoc network data is received by the ad hoc network radio connected by the network port. The ground ad hoc network radio is paired with the wireless receiving device carried by the airborne side to ensure end-to-end data transmission between the two. After receiving the compressed video data, the ground Raspberry Pi CM4-5G expansion board can be used to reorganize the data packets and restore the compressed image video. The NVIDIA Jetson Xavier NX embedded computing platform is equipped with a target detection model, a video compression module, a network status perception module, a data transmission module, and an intelligent route planning module. The intelligent route planning module can comprehensively consider factors such as arrival time, fuel consumption, threats, and flyable areas to plan the optimal or suboptimal flight trajectory for the drone's subsequent flight. Combining the intelligent algorithm based on artificial intelligence technology with autonomous route planning can effectively avoid the shortcomings of traditional trajectory planning algorithms and achieve dynamic response; the target detection model adopts the YOLOv5 network structure, which is trained by several drone aerial image samples. The YOLOv5 network detects targets by dividing the input image into grids of different sizes and predicting the bounding box and category of the target in each grid; during the training process, according to the feedback of the loss function, the YOLOv5 network adjusts its weights to optimize the performance and accuracy of the model. During the detection process, the detection video is output for subsequent processing. The video compression module compresses the video by calling the ffmpeg tool in the onboard Raspberry Pi CM4-5G expansion board, and adjusts the compression rate by adjusting the CRF value when transmitting the video. The CRF parameter is used to control the video quality of H.264 encoding. CRF is the abbreviation of Constant Rate Factor. It is a quantization parameter used to control the compression rate and quality of video encoding. The value range of CRF is between 0 and 51, where 0 represents lossless compression (i.e., original video quality) and 51 represents the lowest quality and highest compression rate. The value of CRF is inversely proportional to the video quality and file size. Specifically, under normal circumstances, the smaller the CRF value, the higher the video quality, but the larger the file size; the larger the CRF value, the lower the video quality, and the smaller the file size.The ground Raspberry Pi CM4-5G expansion board is used to establish an ad hoc network and 5G connection between the ground and the drone, sense the connection status of the current link in real time, switch the data transmission link based on the status, and transmit the current transmission quality to the NVIDIA Jetson Xavier NX embedded computer, so that it can perform video compression based on the value; and the ground Raspberry Pi CM4-5G expansion board uses the integrated 5G module to complete the connection of the 5G link, and at the same time connects the ground ad hoc network radio and the airborne ad hoc network radio through the network port to complete the connection of the ad hoc network link, and uses the airborne ad hoc network radio to complete the reading of the transmitted video data for real-time playback. The airborne Raspberry Pi CM4-5G expansion board is used to sense the connection quality of the two links, and then transmit the data to the NVIDIA Jetson Xavier NX embedded computer. The NVIDIA Jetson Xavier NX embedded computer compresses the video, and after compression, it is transmitted to the ground end through the airborne Raspberry Pi CM4-5G expansion board. The two links are the ad hoc network link and the 5G link. The problem of large signal fluctuation range in the drone scene is fully considered. At present, 5G communication has disadvantages such as small base station coverage, low overall coverage rate, and high price. Wireless ad hoc network has the advantages of longer transmission distance, no dependence on ground infrastructure, and flexible deployment. It can be used as a supplementary communication means for 5G public network, thereby ensuring the real-time online of drones. Ground ad hoc network radio and airborne ad hoc network radio are both used to establish ad hoc network links on the airborne end and the ground end.

[0021] See also Figure 2 , a method for dynamic switching and adaptive compression of multi-modal communication in low-altitude intelligent network, comprising the following steps: Step 1: The airborne end and the ground end establish a communication connection, and the airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board perceive the quality information of the two communication links in real time; Step 2: Specify the starting and ending points of the drone flight. The NVIDIA Jetson Xavier NX embedded computer will autonomously plan the flight path based on the environment and transmit the path information to the flight control computer. Step 3: During the drone cruise, the optical pod camera on board captures video data in real time, and then transmits the video data to the NVIDIA Jetson Xavier NX embedded computer; Step 4: The NVIDIA Jetson Xavier NX embedded computer obtains the link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board, determines the compression rate of the transmitted video based on the data, compresses the video data captured in step 3, and transmits the video data to the onboard Raspberry Pi CM4-5G expansion board after the compression is completed; the specific process is as follows: S41, the optical pod camera transmits the collected video data to the NVIDIA Jetson Xavier NX embedded computer; The S42, NVIDIA Jetson Xavier NX embedded computer obtains link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board and uses this data to determine the compression rate of the video transmission; S43, the video compression rate is divided into four gears. When the link signal-to-noise ratio is lower than 50dB, the video compression CRF value is 30; when the signal-to-noise ratio is 30dB to 60dB, the CRF value is 28; when the signal-to-noise ratio is 60dB to 70dB, the CRF value is 26; when the signal-to-noise ratio is above 70dB, the CRF value is 24; S44. After the video compression is completed, the NVIDIA Jetson Xavier NX embedded computer transmits the compressed video data to the onboard Raspberry Pi CM4-5G expansion board; S45, after receiving the data, the onboard Raspberry Pi CM4-5G expansion board transmits the data to the ground Raspberry Pi CM4-5G expansion board through the established link.

[0022] Step 5: The airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board select a video transmission link according to the connection status of the two current links, and transmit the video data from the selected link to the ground Raspberry Pi CM4-5G expansion board; the process of selecting the video transmission link is as follows: Figure 3 As shown, the details are as follows: S51. When the program starts, define the communication address. The self-organizing network link uses a fixed IPv4 address, and the addresses at both ends are 192.168.17.100 and 192.168.17.101 respectively. For the 5G link, due to the particularity of the 5G network, the IPv6 address of the link is randomly assigned each time the system is turned on, so the addresses at both ends are not fixed. S52, after the system is started, read the IPv6 address of the local end and record it for communication; S53, set a timer function, the function sends a connection establishment message to the other party's address every m (m=1) seconds, called a "heartbeat packet"; S54, real-time capture of the heartbeat packets sent by the other party, and statistics of the received heartbeat packets, each time a new heartbeat packet is received, the counter is reset to keep the value of the counter at 0; S55. If the system does not receive the heartbeat packet sent by the other party within the predetermined time, the value of the counter will gradually increase; S56, when the value of the counter reaches a preset threshold (for example, no heartbeat packet is received for several seconds in a row), it is determined that the current communication link fails; S57, when the link fails, automatically print the link failure information, and trigger the link switching mechanism to switch the data transmission to the standby link that can work normally; S58. If the ad hoc network link fails, the system attempts to use the 5G link for data transmission; vice versa; S59: After detecting link failure, the system immediately switches the link, redefines the communication address, re-establishes the connection with the other party, and continues to send and receive data; S60: During the switching process, the system continuously monitors the status of the new link and makes further adjustments based on the actual situation; S61. If the failed link returns to normal after a period of time, the system will recapture the heartbeat packet sent by the other party and re-establish the connection of the link; S62: Determine whether to switch data transmission back to the restored link according to the link priority and current communication conditions, so as to optimize data transmission efficiency.

[0023] Step 6: After the ground end receives the compressed video data, the ground Raspberry Pi CM4-5G expansion board decompresses the compressed video and plays it in real time.

[0024] like Figure 4 As shown, in specific applications, the entire system is divided into three layers, including application layer, function layer, and basic layer; the specific contents are as follows: Application layer: Based on the basic data and operational data provided by the infrastructure layer, this layer can provide various business applications, including disaster monitoring and rescue, agricultural plant protection, geological exploration, and urban planning.

[0025] Functional layer: This layer implements various functions during system operation, including intelligent analysis (analyzing collected data), real-time online network access (ensuring network stability), efficient data return (low bandwidth consumption) and beyond-line-of-sight transmission.

[0026] Basic layer: This layer provides basic support for the system's various business operations and internal and external information exchanges. It includes software resources such as network status perception module, data transmission module, data collection module and intelligent data analysis module, as well as hardware resources such as 5G public network communication module, Raspberry Pi, NVIDA chip and self-organizing network module.

[0027] Therefore, the present invention adopts the above-mentioned low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system and method, integrating the basic public loads of "perception", i.e. perception of transmission link quality, "transmission", i.e. 5G+advanced network link dual-link transmission, "computation", i.e. intelligent video compression algorithm, and "control", i.e. autonomous path planning of drones, etc., and utilizes 5G technology, ad hoc network technology, and airborne embedded AI computers to coordinate and combine the advantages of multiple aspects, thereby reducing the amount of backhaul data, reducing the bandwidth consumed by transmission, improving the accuracy of information transmission, and adapting to a variety of application scenarios.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system, characterized by: It includes the airborne side and the ground side; the airborne side includes the NVIDIA Jetson Xavier NX embedded computing platform, the airborne Raspberry Pi CM4-5G expansion board, the airborne ad hoc network radio, the flight control computer and the optical pod camera; the ground side includes the ground Raspberry Pi CM4-5G expansion board and the ground ad hoc network radio.

2. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 1 is characterized by: The NVIDIA Jetson Xavier NX embedded computing platform is equipped with a target detection model, video compression module, network status perception module, data transmission module and route intelligent planning module.

3. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 2 is characterized by: The target detection model adopts the YOLOv5 network structure and is trained with several drone aerial image samples. The YOLOv5 network performs target detection by dividing the input image into grids of different sizes and predicting the bounding box and category of the target in each grid.

4. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 3 is characterized by: The video compression module compresses the video by calling the ffmpeg tool in the onboard Raspberry Pi CM4-5G expansion board. The compression rate is adjusted by adjusting the CRF value when transmitting the video. The CRF parameter is used to control the video quality of H.264 encoding. The CRF value range is between 0 and 51, where 0 indicates lossless compression and 51 indicates the lowest quality and highest compression rate. The CRF value is inversely proportional to the video quality and file size.

5. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 4 is characterized by: The ground Raspberry Pi CM4-5G expansion board is used to establish an ad hoc network and 5G connection between the ground and the drone, sense the connection status of the current link in real time, switch the data transmission link based on the status, and transmit the current transmission quality to the NVIDIA Jetson Xavier NX embedded computer, so that it can compress the video according to the value; and the ground Raspberry Pi CM4-5G expansion board uses the integrated 5G module to complete the connection of the 5G link, and connects to the ground ad hoc network radio through the network port to complete the connection of the ad hoc network link with the airborne ad hoc network radio, and uses the ground-end Raspberry Pi CM4-5G expansion board to complete the reading and decompression of the transmitted video data for real-time playback.

6. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 5 is characterized by: The onboard Raspberry Pi CM4-5G expansion board is used to sense the connection quality of the two links, and then transmit the data to the NVIDIA Jetson Xavier NX embedded computer. The NVIDIA Jetson Xavier NX embedded computer compresses the video and transmits it to the ground end through the onboard Raspberry Pi CM4-5G expansion board. The two links are the ad hoc network link and the 5G link.

7. The low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 6 is characterized by: Both the ground ad hoc network radio and the airborne ad hoc network radio are used to establish ad hoc network links on the airborne side and the ground side.

8. A method for dynamic switching and adaptive compression of multi-modal communication in low-altitude intelligent network according to any one of claims 1 to 7, characterized in that: The following steps are involved: Step 1: The airborne end and the ground end establish a communication connection, and the airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board perceive the quality information of the two communication links in real time; Step 2: Specify the starting and ending points of the drone flight. The NVIDIA Jetson Xavier NX embedded computer will autonomously plan the flight path based on the environment and transmit the path information to the flight control computer. Step 3: During the drone cruise, the optical pod camera on board captures video data in real time, and then transmits the video data to the NVIDIA Jetson Xavier NX embedded computer; Step 4: The NVIDIA Jetson Xavier NX embedded computer obtains the link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board, compresses the video data captured in step 3, and transmits the video data to the onboard Raspberry Pi CM4-5G expansion board after the compression is completed; Step 5. The airborne Raspberry Pi CM4-5G expansion board and the ground Raspberry Pi CM4-5G expansion board both select a video transmission link according to the connection status of the two current links, and transmit the video data from the selected link to the ground Raspberry Pi CM4-5G expansion board; Step 6: After the ground end receives the compressed video data, the ground Raspberry Pi CM4-5G expansion board decompresses the compressed video and plays it in real time.

9. The method of low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 8 is characterized in that: The specific process of step 4 is as follows: S41, the optical pod camera transmits the collected video data to the NVIDIA Jetson Xavier NX embedded computer; The S42, NVIDIA Jetson Xavier NX embedded computer obtains link quality information transmitted by the onboard Raspberry Pi CM4-5G expansion board and uses this data to determine the compression rate of the video transmission; S43, the video compression rate is divided into four gears. When the link signal-to-noise ratio is lower than 50dB, the video compression CRF value is 30; when the signal-to-noise ratio is 30dB to 60dB, the CRF value is 28; when the signal-to-noise ratio is 60dB to 70dB, the CRF value is 26; when the signal-to-noise ratio is above 70dB, the CRF value is 24; S44. After the video compression is completed, the NVIDIA Jetson Xavier NX embedded computer transmits the compressed video data to the onboard Raspberry Pi CM4-5G expansion board; S45, after receiving the data, the onboard Raspberry Pi CM4-5G expansion board transmits the data to the ground Raspberry Pi CM4-5G expansion board through the established link.

10. The method of low-altitude intelligent network multi-modal communication dynamic switching and adaptive compression system according to claim 9 is characterized in that: The process of selecting the video transmission link in step 5 is as follows: S51. When the program starts, the communication address is defined. The self-organizing network link uses a fixed IPv4 address. For the 5G link, the IPv6 address of the link is randomly assigned. S52, after the system is started, read the IPv6 address of the local end and record it; S53, set a timer function, the function sends a connection establishment message to the other party's address every m seconds, called a "heartbeat packet"; S54, real-time capture of the heartbeat packets sent by the other party, and statistics of the received heartbeat packets, each time a new heartbeat packet is received, the counter is reset to keep the value of the counter at 0; S55. If the system does not receive the heartbeat packet sent by the other party within the predetermined time, the value of the counter will gradually increase; S56, when the value of the counter reaches a preset threshold, determining that the current communication link fails; S57, when the link fails, automatically print the link failure information, and trigger the link switching mechanism to switch the data transmission to the standby link that can work normally; S58. If the ad hoc network link fails, the system attempts to use the 5G link for data transmission; vice versa; S59: After detecting link failure, the system immediately switches the link, redefines the communication address, re-establishes the connection with the other party, and continues to send and receive data; S60: During the switching process, the system continuously monitors the status of the new link and makes further adjustments based on the actual situation; S61. If the failed link returns to normal after a period of time, the system will recapture the heartbeat packet sent by the other party and re-establish the connection of the link; S62: Determine whether to switch data transmission back to the restored link according to the link priority and current communication conditions, so as to optimize data transmission efficiency.

Images