Underwater visual perception and edge end data processing module with high computing power

Abstract

The invention relates to the technical field of underwater intelligent equipment application, in particular to an underwater visual perception and edge end data processing module with high computing power, which is suitable for universal underwater visual perception and edge end data processing of various underwater robots carrying high computing power platforms. The whole module is of an underwater sealed cabin structure, the sensor assembly is used for collecting underwater scene data and is connected to the computing power platform assembly through a data line, and the computing power platform assembly is used for processing and analyzing the data of the sensor assembly and receiving the data of the sensor assembly through the data line; the communication and switch assembly is responsible for data transmission with an external robot and equipment control, and is connected to the computing power platform assembly through a data line; a passive heat dissipation mode is adopted, and heat of internal hardware is conducted to an external water environment through a heat conduction material structural part. According to the invention, on the basis of NVIDIAJetson embedded edge computing equipment, extra high-computing-power perceptual information processing capability is provided for various underwater robots.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater intelligent equipment applications, and particularly relates to an underwater vision perception and edge - end data processing module with high computing power, which is applicable to the general underwater vision perception and edge - end data processing of high - computing - power platforms carried by various underwater robots. Background Art

[0002] In recent years in the ocean field, with the continuous growth of demands such as underwater operations, ocean monitoring, and seabed exploration, the development and utilization of underwater unmanned vehicles (UUVs) have become an important trend in ocean development. Among them, the perception system is an important part of underwater robots, which can help robots effectively perceive the underwater environment to achieve underwater operations. Currently, the perception function of underwater robots mainly collects various information of the environment, targets, and the robot itself by carrying various sensors such as sonars, optical cameras, inertial measurement units, depth gauges, etc. After the robot obtains the input information and processes it, it provides corresponding feedback for downstream related tasks. For the processing of perception information, there are usually relatively high requirements for hardware, including sufficient computing power. The information can be transmitted to the on - shore processing unit through wired or wireless communication methods, or the information can be stored and processed offline after being brought ashore. Therefore, autonomous and online perception underwater is a challenge faced by underwater robots, and enhancing the underwater perception ability of robots has become one of the important requirements in the current ocean equipment field.

[0003] Visual perception, as an efficient robot perception method, can particularly well meet the requirements of real - time and high - precision in the perception of underwater close - range environments, and the sensing devices of the visual system are small in volume and low in power consumption. However, in combination with the development status of computer vision technology on land, when facing complex tasks, there is usually a necessary requirement for the computing power of the processing hardware. The complex underwater environment makes it relatively difficult to equip a high - computing - power platform for underwater operations. Especially for small underwater robots, the robot hardware structure is simple and lacks scalability for the perception function. Most of the existing underwater robot perception modules are integrated with modules such as drive control, communication, and power supply in the dry cabin of the robot body to form the whole robot. However, the high integration brings some limitations to the performance of the hardware, such as the lack of high - performance computing capabilities supporting deep learning, slow heat dissipation in the enclosed space, poor versatility of the perception module, etc. Furthermore, in terms of software, the mature perception algorithms on land are limited by the robot hardware and are difficult to be directly migrated and applied to the underwater environment, so there are certain limitations and impacts on the underwater operation ability of the robot. Summary of the Invention

[0004] The present invention aims to make up for the deficiencies in the design of high-computing-power perception hardware in existing marine technology equipment, and proposes an underwater vision perception and edge-side data processing module with high computing power. The high-computing-power perception computing module based on NVIDIA Jetson embedded edge computing devices, as a small and independent hardware module similar to an "underwater brain", realizes providing external perception computing and decision-making capabilities with high computing power for various types of underwater robots, and can run artificial intelligence (AI) algorithms such as deep learning and large models, thereby assisting in improving the operation capabilities of underwater robots based on scene understanding and cognitive decision-making.

[0005] The technical solution of the present invention is as follows:

[0006] An underwater vision perception and edge-side data processing module with high computing power. The whole module is an underwater sealed cabin structure with waterproof and pressure-resistant properties. Inside the sealed cabin, there are a sensor component, a computing power platform component, a communication and switch component, and an internal support of the cabin. The specific connection relationships are as follows: The sensor component is used to collect underwater scene data and is connected to the computing power platform component in a data cable manner. The computing power platform component is used to process and analyze the data of the sensor component, receives the data of the sensor component through the data cable, and sends the processing results through the communication module. The communication and switch component is responsible for data transmission and device control and is connected to the computing power platform component through the data cable. The internal support of the cabin is used to fix and support the sensor component, the computing power platform component, and the communication and switch component. The sensor component, the computing power platform component, and the communication and switch component are installed on the internal support of the cabin.

[0007] The underwater vision perception and edge-side data processing module with high computing power is based on NVIDIA Jetson embedded edge computing devices and provides additional perception information processing capabilities with high computing power for various underwater robots.

[0008] For the underwater vision perception and edge-side data processing module with high computing power, the sensor component includes: a USB camera, an IMU attitude sensing module, a camera fixing bracket, and an IMU fixing bracket. The specific connection relationships are as follows: The USB camera is installed on the camera fixing bracket. The camera fixing bracket and the IMU fixing bracket are hinged so that the USB camera can pitch and rotate to adjust the lens angle. The IMU attitude sensing module is installed on the IMU fixing bracket, and the IMU fixing bracket is fixedly connected to the front flange bracket of the internal support of the cabin.

[0009] The underwater vision perception and edge-side data processing module with high computing power uses a USB camera as the main body vision sensor and is assisted by an IMU attitude sensing module. The USB camera adjusts the pitch viewing angle through the IMU fixing bracket.

[0010] The underwater vision perception and edge - end data processing module with high computing power, the computing power platform components include: NVIDIA Jetson embedded edge computing device, power conversion module, battery, and heat - conducting material structure. The specific connection relationship is as follows: The power interface of the NVIDIA Jetson embedded edge computing device is connected to the output end of the power conversion module, and the data interface of the NVIDIA Jetson embedded edge computing device is connected to external devices; the input end of the power conversion module is connected to the positive and negative electrodes of the battery, and the output end of the power conversion module is connected to the power interface of the NVIDIA Jetson embedded edge computing device.

[0011] For the underwater vision perception and edge - end data processing module with high computing power, the outer side of the fan metal shell of the NVIDIA Jetson embedded edge computing device is closely attached to one side of the heat - conducting material structure. The other side of the heat - conducting material structure is closely attached to the inner side of the aluminum alloy hatch cover and flange. Thermal grease or phase - change sheets are added to the two attachment gaps to reduce contact voids. The heat - conducting material structure is clamped and fixed by the nut of the waterproof bolt, nut, and plug installed on the aluminum alloy hatch cover and flange; a passive heat dissipation method is adopted to conduct the internal hardware heat to the external water environment through the heat - conducting material structure.

[0012] For the underwater vision perception and edge - end data processing module with high computing power, the communication and switch components include: two waterproof bolt, nut, and plugs, two watertight connectors, and a waterproof rotary switch, all installed on the flange of the aluminum alloy hatch cover and flange. The function of the waterproof rotary switch is to short - circuit the interface pins of the NVIDIA Jetson embedded edge computing device, serving as the power - on switch button of the perception module. One path of the watertight connector is a gigabit 8 - core network cable connected to the network port of the NVIDIA Jetson embedded edge computing device, and the other path is a 2 - core wire connected to the battery charging port.

[0013] The described underwater vision perception and edge - side data processing module with high computing power. The external structure of this module includes an acrylic transparent hemispherical cover, a hemispherical cover pressing plate and flange, an acrylic sealed cabin tube, an external fixed connector of the module, an aluminum alloy cabin cover and flange, which are arranged in sequence along the same axis direction. The specific connection relationships are as follows: The hemispherical cover pressing plate and flange are a combined structure where the pressing plate and flange are fixedly connected by bolts. The acrylic transparent hemispherical cover is fixedly connected with an interference fit between the pressing plate and flange of the hemispherical cover pressing plate and flange. The pressing plate and flange clamp the edge of the acrylic transparent hemispherical cover to ensure a tight fit. One end of the acrylic sealed cabin tube is connected with the flange of the hemispherical cover pressing plate and flange through an interference fit with a rubber sealing ring. The aluminum alloy cabin cover and flange are a combined structure where the cabin cover and flange are fixedly connected by bolts. The other end of the acrylic sealed cabin tube is connected with the flange of the aluminum alloy cabin cover and flange through an interference fit with a rubber sealing ring and is fixedly connected with the cabin cover of the aluminum alloy cabin cover and flange by bolts. The external fixed connector of the module is a ring - shaped two - half split connection structure. Two groups of external fixed connectors of the module are fixedly installed on the acrylic sealed cabin tube. The two external fixed connectors of the module are located at the front and rear ends of the acrylic sealed cabin tube. Threaded holes are opened at the bottom of the external fixed connector of the module as installation holes for the underwater robot mounting combination.

[0014] The design concept of the present invention is:

[0015] The core design concept of the present invention lies in constructing an intelligent modular underwater high - computing - power perception system that can be decoupled from the underwater robot body to solve key problems such as insufficient computing power of underwater robot perception hardware, limited heat dissipation, and poor versatility.

[0016] 1. The adaptability of the high - performance underwater computing platform and deep - learning algorithms enables complex AI models (such as convolutional neural networks) to be directly deployed in the underwater environment. Based on its GPU acceleration architecture, efficient processing of visual data (such as underwater image enhancement, target recognition, SLAM) can be achieved. Different from traditional centralized shore - based processing, edge - side computing significantly reduces communication latency, meets the real - time decision - making requirements of underwater robots, and breaks through the bottleneck of traditional underwater robots relying on offline processing.

[0017] 2. The collaborative optimization of the waterproof sealed cabin and heat - dissipation design. Through the passive contact - type heat - dissipation design of the aluminum alloy cabin cover and heat - conducting material structures (such as thermal conductive silicone grease, phase - change sheets), the heat generated by the GPU hardware is conducted to the cabin shell, and external flowing water is used for liquid - cooling heat dissipation. This design takes into account both the requirements of waterproof pressure resistance and efficient heat dissipation. The low - thermal - resistance characteristics of the heat - dissipation structure ensure the continuous performance of the hardware under full load (such as preventing the computing power platform from reducing frequency due to overheating).

[0018] 3. The module is independently powered and has a standardized communication interface (gigabit Ethernet port). Through the design of universal mounting holes in external fixed connectors, it can be quickly adapted to various underwater robots (such as ROVs and AUVs). The "plug and play" feature enables the module to operate independently of the robot body (such as only performing underwater observation tasks), or to be connected to the robot's main control system via an Ethernet cable to form a distributed computing architecture (dynamic allocation of computing power resources), greatly enhancing the system flexibility.

[0019] The advantages and beneficial effects of the present invention are:

[0020] 1. The present invention proposes an underwater vision perception module equipped with a high-computing-power platform. The overall design is a waterproof and pressure-resistant sealed cabin. Based on NVIDIA Jetson embedded edge computing devices, it can provide additional high-computing-power perception information processing capabilities for various underwater robots.

[0021] 2. The present invention mainly mounts a USB wide-angle camera as the main body vision sensor, assisted by a nine-axis IMU attitude sensor. The camera can adjust the pitch angle through the mounting bracket.

[0022] 3. The present invention adopts a passive contact type solid heat conduction design inside the cabin, and uses the flowing water outside the cabin for liquid cooling heat dissipation of the cabin cover section. The heat dissipation method of the NVIDIA Jetson hardware inside the underwater sealed cabin uses high thermal conductivity materials to contact and connect to conduct the heat generated by the hardware to the metal part of the sealed cabin shell (such as the aluminum alloy cabin cover). The shell is in direct contact with water to conduct the heat to the external water environment, with relatively high heat dissipation efficiency to optimize the operating performance of the NVIDIA Jetson computing power platform.

[0023] 4. As an independent module for underwater perception (underwater brain), the present invention has a self-powered power supply and communication function, can be installed on various types of underwater robot operation platforms, etc., and can communicate with the robot via an Ethernet cable and provide the perception information processed by the computing power platform for the robot. Description of the Drawings

[0024] Figure 1 is the overall external structure schematic diagram of the underwater vision perception and edge-end data processing module provided by the present invention; in the figure, 1-1, acrylic transparent hemispherical cover; 1-2, hemispherical cover pressing plate and flange; 1-3, acrylic sealed cabin tube; 1-4, external fixed connector of the module; 1-5, aluminum alloy cabin cover and flange.

[0025] Figure 2 is the internal installation structure schematic diagram of the module provided by the present invention; in the figure, 2, sensor assembly; 3, computing power platform assembly; 4, communication and switch assembly; 5, internal support of the cabin.

[0026] Figure 3It is a schematic structural diagram of the module sensor component provided by the present invention; in the figure, 2-1 is a USB camera, 2-2 is a nine-axis IMU attitude sensing module, 2-3 is a camera fixing bracket, and 2-4 is an IMU fixing bracket.

[0027] Figure 4 It is a schematic structural diagram of the module core computing power platform component provided by the present invention; in the figure, 3-1 is an NVIDIA Jetson embedded edge computing device, 3-2 is a power conversion module, 3-3 is a battery, 3-4 is a heat-conducting material structure, 5-1 is a front flange bracket, 5-2 is an upper bracket, 5-3 is a lower bracket, and 5-4 is a rear flange bracket.

[0028] Figure 5 It is a schematic structural diagram of the module communication and switch component provided by the present invention; in the figure, 4-1 is a waterproof bolt, nut and plug, 4-2 is a watertight connector, and 4-3 is a waterproof knob switch. Specific embodiments

[0029] Next, the present invention will be further described in detail with reference to the accompanying drawings and specific embodiments.

[0030] As Figure 1 shown, the underwater vision perception and edge-end data processing module provided by the present invention is in the form of a waterproof and pressure-resistant underwater sealed cabin as a whole. The external structure includes an acrylic transparent hemispherical cover 1-1, a hemispherical cover pressing plate and a flange 1-2, an acrylic sealed cabin tube 1-3, a module external fixing connector 1-4, an aluminum alloy cabin cover and a flange 1-5 arranged in sequence along the same axis direction. The specific connection relationship is as follows: The hemispherical cover pressing plate and the flange 1-2 are a combined structure of a pressing plate and a flange fixed by bolts. The acrylic transparent hemispherical cover 1-1 is fixedly connected with the pressing plate and the flange of the hemispherical cover pressing plate and the flange 1-2 through a transition fit. The pressing plate and the flange clamp the edge of the acrylic transparent hemispherical cover 1-1 to ensure a tight fit. One end of the acrylic sealed cabin tube 1-3 is connected with the flange of the hemispherical cover pressing plate and the flange 1-2 through an interference fit with a rubber sealing ring. The aluminum alloy cabin cover and the flange 1-5 are a combined structure of a cabin cover and a flange fixed by bolts. The other end of the acrylic sealed cabin tube 1-3 is connected with the flange of the aluminum alloy cabin cover and the flange 1-5 through an interference fit with a rubber sealing ring and is fixedly connected with the cabin cover of the aluminum alloy cabin cover and the flange 1-5 by bolts. In addition, the characteristic position of the module external fixing connector 1-4: The module external fixing connector 1-4 is a ring-shaped two-half split connection structure. Two sets of module external fixing connectors 1-4 are fixedly installed on the acrylic sealed cabin tube 1-3. The two module external fixing connectors 1-4 are located at the front and rear ends of the acrylic sealed cabin tube 1-3. The bottom of the module external fixing connector 1-4 is provided with an internal threaded hole as an installation hole for the underwater robot carrying combination.

[0031] As Figure 2As shown in the figure, the interior of the sealed cabin includes: a sensor component 2, a computing power platform component 3, a communication and switch component 4, and an interior bracket 5 of the cabin. The specific connection relationships are as follows: The sensor component 2 is used to collect underwater scene data and is connected to the computing power platform component 3 by means of a data cable. The computing power platform component 3 is used to process and analyze the data of the sensor component 2, receives the data of the sensor component 2 through the data cable, and sends the processing results through the communication module, and interacts with other systems or devices through the communication and switch component 4. The communication and switch component 4 is responsible for data transmission and device control, is connected to the computing power platform component 3 through a data cable, and is connected to a switch or other control device through a control cable. The interior bracket 5 of the cabin is used to fix and support the sensor component 2, the computing power platform component 3, and the communication and switch component 4. The sensor component 2, the computing power platform component 3, and the communication and switch component 4 are installed on the interior bracket 5 of the cabin by means of screws, nuts, buckles, etc.

[0032] As Figure 3 shown, the sensor component 2 includes: a USB camera 2-1, a nine-axis IMU attitude sensing module 2-2, a camera fixing bracket 2-3, and an IMU fixing bracket 2-4. The specific connection relationships are as follows: The USB camera 2-1 is installed on the camera fixing bracket 2-3. The camera fixing bracket 2-3 and the IMU fixing bracket 2-4 are hinged so that the USB camera 2-1 can pitch and rotate to adjust the lens angle. The nine-axis IMU attitude sensing module 2-2 is installed on the IMU fixing bracket 2-4. The IMU fixing bracket 2-4 is fixedly connected to the front flange bracket 5-1 of the interior bracket 5 of the cabin. The circumferential structure of the IMU fixing bracket 2-4 fits the flange to ensure that there is no relative displacement between the nine-axis IMU attitude sensing module 2-2 and the entire module.

[0033] As Figure 4As shown in the figure, the computing power platform component 3 includes: NVIDIA Jetson embedded edge computing device 3-1, power conversion module 3-2, battery 3-3, and thermal conductive material structure 3-4. The specific connection relationship is as follows: The power interface of the NVIDIA Jetson embedded edge computing device 3-1 is connected to the output end of the power conversion module 3-2, and the data interface of the NVIDIA Jetson embedded edge computing device 3-1 is connected to an external device. The input end of the power conversion module 3-2 is connected to the positive and negative electrodes of the battery 3-3, and the output end of the power conversion module 3-2 is connected to the power interface of the NVIDIA Jetson embedded edge computing device 3-1. In addition, the characteristic positions of the thermal conductive material structure 3-4 are as follows: The outer side of the fan metal shell of the NVIDIA Jetson embedded edge computing device 3-1 is closely attached to one side of the thermal conductive material structure 3-4, and the other side of the thermal conductive material structure 3-4 is closely attached to the inner side of the aluminum alloy hatch cover of the aluminum alloy hatch cover and flange 1-5. Thermal conductive silicone grease or phase change sheets can be added to the two attachment gaps to reduce the contact voids. The thermal conductive material structure 3-4 is clamped and fixed by the nuts of two waterproof bolt and nut plugs 4-1 installed on the aluminum alloy hatch cover and flange 1-5. After the aluminum alloy hatch cover and flange 1-5 are installed, both sides of the thermal conductive material structure 3-4 are closely attached.

[0034] As Figure 4 shown in the figure, the internal support 5 of the cabin body includes: front flange support 5-1, upper support 5-2, lower support 5-3, and rear flange support 5-4. The specific connection relationship is as follows: The upper support 5-2 and the lower support 5-3 are arranged opposite to each other up and down. One side of them is connected to the front flange support 5-1, and the other side is connected to the rear flange support 5-4.

[0035] As Figure 4 shown in the figure, the characteristic positions of the computing power platform component 3 and the internal support 5 of the cabin body are as follows: The positioning holes at both ends on one side of the fan on the NVIDIA Jetson embedded edge computing device 3-1 and the positioning holes of the rear flange support 5-4 are fixed by bolts and nuts. The power conversion module 3-2 and the upper support 5-2 are connected and fixed by bolts and nuts. The upper support 5-2 is installed and fixed on the outer side of the rear flange support 5-4. The battery 3-3 is installed on the lower support 5-3 and fixed by interference fit with the rectangular frame of the lower support 5-3. The lower support 5-3 is installed and fixed on the inner side of the rear flange support 5-4. The front flange support 5-1 is fixed to the inner threaded hole on the flange of the hemisphere cover pressing plate and flange 1-2 by bolts. The rear flange support 5-4 is fixed to the inner threaded hole on the flange of the aluminum alloy hatch cover and flange 1-5 by bolts.

[0036] As Figure 5As shown in the figure, the communication and switch component 4 includes: two waterproof bolt nut plugs 4-1, two watertight connectors 4-2, and a waterproof rotary switch 4-3, all of which are installed on the flange of the aluminum alloy hatch cover and flange 1-5. The function of the waterproof rotary switch 4-3 is to short-circuit the interface pins PIN1 and PIN2 of the NVIDIA Jetson embedded edge computing device 3-1, serving as the power-on switch button for the sensing module. One path of the watertight connector 4-2 is a gigabit 8-core network cable connected to the network port of the NVIDIA Jetson embedded edge computing device 3-1, and the other path is a 2-core wire connected to the battery charging port.

[0037] In the present invention, the battery of the computing power platform component can be optionally selected with an output of 12V and a capacity of 5000mAh, which can support the module to run at full power for nearly two hours; the model of the NVIDIA Jetson embedded edge computing device can be optionally selected as Orin Nano or Orin Nx, etc.

[0038] The specific installation method of the module of the present invention on various underwater robots depends on the shape structure of the robot. It can be connected to the communication interface of the robot through network cable communication, or it can also be installed in combination for underwater observation and other sensing tasks only, thus not requiring a communication function to connect to the underwater robot. "Underwater Brain" is a conceptual name for the high-computing-power sensing module provided by the present invention, meaning that this module can provide a "brain" for underwater robots to intelligently process visual perception information and then feedback to the robot body to execute relevant underwater operation tasks.

[0039] The implementation results show that an underwater vision perception and edge-side data processing module (underwater brain) provided by the present invention provides a high-computing-power processing platform for the information collected by sensors such as visual cameras by encapsulating the NVIDIA Jetson embedded edge computing device and designing a corresponding sealed cabin heat dissipation structure for it. As a general-purpose underwater module, it can be independently powered and communicate, and thus can be carried on various underwater robots, especially small robots, to deploy and run sensing algorithms with certain computing power requirements, such as underwater image enhancement based on deep learning, underwater target recognition, underwater SLAM, underwater base model, etc. The present invention can expand and enhance the perception operation ability of underwater robots, and at the same time provides a modular solution for the brain-body decoupling of underwater robot intelligence.

Claims

1. An underwater vision perception and edge - side data processing module with high computing power, characterized in that, The whole module is an underwater sealed cabin structure with waterproof and pressure-resistant properties. Inside the sealed cabin, there are a sensor component, a computing power platform component, a communication and switch component, and internal brackets of the cabin body. The specific connection relationships are as follows: The sensor component is used to collect underwater scene data and is connected to the computing power platform component by means of a data cable. The computing power platform component is used to process and analyze the data of the sensor component, receives the data of the sensor component through the data cable, and sends the processing results through the communication module. The communication and switch component is responsible for data transmission and device control and is connected to the computing power platform component by a data cable. The internal brackets of the cabin body are used to fix and support the sensor component, the computing power platform component, and the communication and switch component. The sensor component, the computing power platform component, and the communication and switch component are installed on the internal brackets of the cabin body.

2. The underwater vision perception and edge - side data processing module with high computing power according to claim 1, characterized in that, Based on the NVIDIA Jetson embedded edge computing device, this module provides additional high-computing-power perception information processing capabilities for various underwater robots.

3. The underwater vision perception and edge - side data processing module with high computing power according to claim 1, characterized in that, The sensor component includes: a USB camera, an IMU attitude sensing module, a camera fixing bracket, and an IMU fixing bracket. The specific connection relationships are as follows: The USB camera is installed on the camera fixing bracket. The camera fixing bracket and the IMU fixing bracket are hinged so that the USB camera can pitch and rotate to adjust the lens angle. The IMU attitude sensing module is installed on the IMU fixing bracket, and the IMU fixing bracket is fixedly connected to the front flange bracket of the internal brackets of the cabin body.

4. The underwater vision perception and edge - side data processing module with high computing power according to claim 3, characterized in that, The USB camera is used as the main body vision sensor, and is assisted by the IMU attitude sensing module. The USB camera adjusts the pitch viewing angle through the IMU fixing bracket.

5. The underwater vision perception and edge - side data processing module with high computing power according to claim 1, characterized in that, The computing power platform component includes: an NVIDIA Jetson embedded edge computing device, a power conversion module, a battery, and a heat-conducting material structure. The specific connection relationships are as follows: The power interface of the NVIDIA Jetson embedded edge computing device is connected to the output end of the power conversion module. The data interface of the NVIDIA Jetson embedded edge computing device is connected to external devices. The input end of the power conversion module is connected to the positive and negative poles of the battery, and the output end of the power conversion module is connected to the power interface of the NVIDIA Jetson embedded edge computing device.

6. The underwater vision perception and edge - side data processing module with high computing power according to claim 5, characterized in that, The outer side of the fan metal shell of the NVIDIA Jetson embedded edge computing device is closely attached to one side of the heat-conducting material structure. The other side of the heat-conducting material structure is closely attached to the inner side of the aluminum alloy cabin cover and the flange of the aluminum alloy cabin cover. Thermal grease or a phase change sheet is added to the two attachment gaps to reduce the contact voids. The heat-conducting material structure is clamped and fixed by the nut of the waterproof bolt nut plug installed on the aluminum alloy cabin cover and the flange. The passive heat dissipation method is adopted to conduct the internal hardware heat to the external water environment through the heat-conducting material structure.

7. The underwater vision perception and edge - side data processing module with high computing power according to claim 1 or 6, characterized in that, The communication and switch component includes: two waterproof bolt-nut plugs, two watertight connectors, and a waterproof rotary switch, all of which are installed on the flange of the aluminum alloy hatch cover and flange. The function of the waterproof rotary switch is to short-circuit the interface pins of the NVIDIA Jetson embedded edge computing device and serve as the power-on switch button for the sensing module. One path of the watertight connector is a gigabit 8-core network cable connected to the network port of the NVIDIA Jetson embedded edge computing device, and the other path is a 2-core wire connected to the battery charging port.

8. The underwater vision perception and edge - side data processing module with high computing power according to claim 1 or 6, characterized in that, The external structure of this module includes an acrylic transparent hemispherical cover, a hemispherical cover pressing plate and flange, an acrylic sealed cabin tube, an external module fixing connector, an aluminum alloy hatch cover and flange arranged in sequence along the same axis direction. The specific connection relationship is as follows: The hemispherical cover pressing plate and flange are a combined structure of a pressing plate and a flange fixed by bolts. The acrylic transparent hemispherical cover is fixedly connected with a transition fit between the pressing plate and flange of the hemispherical cover pressing plate and flange, and the pressing plate and flange clamp the edge of the acrylic transparent hemispherical cover to ensure a tight fit; One end of the acrylic sealed cabin tube is connected to the flange of the hemispherical cover pressing plate and flange by an interference fit with a rubber sealing ring. The aluminum alloy hatch cover and flange are a combined structure of a hatch cover and a flange fixed by bolts. The other end of the acrylic sealed cabin tube is connected to the flange of the aluminum alloy hatch cover and flange by an interference fit with a rubber sealing ring and is fixedly connected to the hatch cover of the aluminum alloy hatch cover and flange by bolts; The external module fixing connector is a ring-shaped two-half split connection structure. Two groups of external module fixing connectors are fixedly installed on the acrylic sealed cabin tube. The two external module fixing connectors are located at the front and rear ends of the acrylic sealed cabin tube. Threaded holes are opened at the bottom of the external module fixing connector as the installation holes for the underwater robot carrying combination.

Images