High-precision stereoscopic visual perception system of inspection robot

Abstract

The utility model discloses an inspection robot high-precision stereoscopic visual perception system which comprises a shell, a data analysis and processing device is installed in the shell, an induction radar, a depth camera and a pan-tilt camera are arranged on the shell, and the relative positions of the induction radar, the depth camera and the pan-tilt camera are kept unchanged. The induction radar and the depth camera are arranged in front of the shell and used for detecting objects in front of the inspection robot and generating images to guide the inspection robot to move, the pan-tilt camera is arranged at the top of the shell and used for collecting reading information of an instrument panel, and the induction radar, the depth camera and the pan-tilt camera are connected with the data analysis and processing device. The sensing radar and the depth camera are used for collecting data for real-time processing and analysis, a three-dimensional space map can be rapidly constructed, the inspection robot can flexibly avoid various obstacles and efficiently and safely shuttle among all devices, and therefore the cloud deck camera is used for collecting reading information of an instrument panel, and the inspection efficiency is improved. And the inspection task is automatically and efficiently completed.

Description

Technical Field

[0001] This application relates to the field of robotics technology, and in particular to a high-precision stereo vision perception system for an inspection robot. Background Technology

[0002] As the hub of a nuclear power plant's water circulation system, the pump house undertakes a series of crucial functions, including cooling, lubrication, and pressure regulation. Many critical pieces of equipment within the plant, such as the nuclear reactor and turbines, rely on the pump house for a stable and reliable water circulation system. A malfunction in the pump house can lead to serious problems such as equipment overheating and pressure imbalances, ultimately jeopardizing the safe operation of the entire nuclear power plant. Therefore, real-time monitoring and maintenance of the pump house equipment is of paramount importance for ensuring the normal operation of a nuclear power plant.

[0003] With the rapid development of technology, robotics has been widely applied and promoted in the industrial field. Robots possess advantages such as high precision, high reliability, repeatability, and adaptability to harsh environments, effectively compensating for the shortcomings of traditional manual inspections. In the inspection scenario of nuclear power plant pump rooms, the introduction of inspection robots has become an inevitable choice to improve operation and maintenance efficiency and safety. Equipped with advanced sensors and intelligent control systems, inspection robots can achieve automated and precise inspections of pump room equipment, collect equipment operation data in real time, and promptly detect potential faults, providing strong support for the safe operation of nuclear power plants. In the early explorations of nuclear power plant pump room inspections, traditional visual perception technology was attempted, but its limitations gradually became apparent. Traditional two-dimensional vision systems can only acquire planar image information of equipment, making it difficult to accurately reconstruct the complex spatial structure of equipment and the relative positional relationships between components. For example, when identifying minor wear on pump impellers or hidden cracks at pipe connections, the lack of depth information easily leads to misjudgments and missed detections. Moreover, the lighting inside the pump room is complex and variable, with areas of direct sunlight and shadows. When dealing with such environments, traditional vision technology is prone to overexposure or underexposure of images, resulting in the loss of key features and seriously affecting detection accuracy.

[0004] Therefore, it is necessary to develop a high-precision stereo vision perception system for inspection robots. Utility Model Content

[0005] To address one of the technical problems existing in the prior art, this application proposes a high-precision stereo vision perception system for inspection robots used in nuclear power plant pump rooms. This system can adapt to the harsh environment of nuclear power plant pump rooms and aims to automate and intelligentize pump room inspections, thereby improving inspection efficiency and safety.

[0006] According to some embodiments of this application, a high-precision stereo vision perception system for an inspection robot is provided. The perception system is installed above the inspection robot and includes a housing. A data analysis and processing device is installed inside the housing. A sensor radar, a depth camera, and a gimbal camera are disposed on the housing, and the relative positions of the sensor radar, the depth camera, and the gimbal camera remain unchanged. The sensor radar and the depth camera are disposed in front of the housing to detect objects in front of the inspection robot and generate images to guide the movement of the inspection robot. The gimbal camera is disposed on the top of the housing to collect reading information from the instrument panel. The sensor radar, the depth camera, and the gimbal camera are respectively connected to the data analysis and processing device.

[0007] In some embodiments of this application, the sensing radar is the omnidirectional stereo perception mid360 radar developed by Lanwo Technology.

[0008] In some embodiments of this application, the depth camera is a RealSense D435 depth camera, which includes two infrared sensors, an infrared laser emitter, and a color camera. The two infrared sensors are respectively disposed on the left and right sides of the depth camera.

[0009] In some embodiments of this application, the gimbal camera is a SIYIA8mini camera gimbal.

[0010] In some embodiments of this application, an inclined surface is provided in front of the housing, and the angle between the inclined surface and the horizontal plane is between 40° and 90°; the sensing radar is disposed on the inclined surface, and the hemispherical window of the sensing radar is exposed outside the housing.

[0011] In some embodiments of this application, the depth camera is embedded in front of the housing, and the height of the depth camera is lower than the height of the sensing radar.

[0012] In some embodiments of this application, the gimbal camera is provided with a rotating mechanism, which drives the gimbal camera to rotate.

[0013] In some embodiments of this application, the gimbal camera is provided with a pitch mechanism, which drives the gimbal camera to perform pitch movements.

[0014] In some embodiments of this application, the housing is provided with a connecting frame made using 3D printing technology, and the sensing radar, the depth camera, and the gimbal camera are all mounted on the connecting frame.

[0015] In some embodiments of this application, the connecting frame is provided with an adjustable joint structure corresponding to the gimbal camera.

[0016] The beneficial effects of this application include: The high-precision stereo vision perception system for inspection robots provided in this application uses a combination of inductive radar, depth camera, and gimbal camera. The inductive radar and depth camera are set at the front of the shell to detect obstacles in front of the inspection robot. The inductive radar emits high-frequency electromagnetic waves or lasers and receives signals reflected back from target objects, enabling it to quickly and accurately obtain distance information of objects in the surrounding environment. Moreover, the detection of the inductive radar is not limited by lighting conditions and can work stably in dimly lit corners or areas with insufficient temporary lighting due to equipment maintenance. The detection of the inductive radar also has a certain penetrating ability to water vapor and smoke, and can still obtain reliable environmental data even in special situations such as steam leaks in pump rooms. More importantly, the radar mapping technology demonstrates excellent stability and anti-interference ability when facing radiation environments. The depth camera works in conjunction with the sensing radar to collect data. By processing and analyzing the large amount of distance data collected by the sensing radar and depth camera in real time, the data analysis and processing device can quickly build an accurate three-dimensional spatial map of the pump room. This provides the inspection robot with clear and accurate environmental perception information, enabling it to accurately plan its own movement path in the complex pump room environment, flexibly avoid various obstacles, and efficiently and safely move between various devices. It also uses the pan-tilt camera to collect the reading information of the instrument panel and automatically and efficiently completes the inspection task.

[0017] Other features and advantages of this application will be set forth in detail in the following description, and will be apparent in part from the following description, or may be learned by practicing this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solution of this application, the following description will be provided in conjunction with the accompanying drawings and embodiments. It should be understood that the following drawings only show some embodiments of this application and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort. In the drawings:

[0019] Figure 1 This is a structural schematic diagram of the high-precision stereo vision perception system for the inspection robot provided in this application.

[0020] Figure 2 yes Figure 1 A schematic diagram of some of the components.

[0021] Explanation of reference numerals in the attached figures:

[0022] Housing 100, connecting frame 110, radar mounting hole 111, camera mounting hole 112, tilting surface 113, gimbal camera 120, vision device 121, rotation mechanism 122, pitch mechanism 123, sensing radar 130, depth camera 140. Detailed Implementation

[0023] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments. Similar elements in different embodiments are referred to by related similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the present application. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present application are not shown or described in the specification. This is to avoid obscuring the core parts of the present application with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0024] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0025] The serial numbers assigned to components in this document, such as "first" and "second," are used only to distinguish the described objects and have no sequential or technical meaning. The terms "connection" and "linkage" used in this application, unless otherwise specified, include both direct and indirect connections (linkages).

[0026] The following is combined Figures 1 to 2 The provided embodiments further illustrate the high-precision stereo vision perception system for inspection robots proposed in this application.

[0027] like Figure 1 and Figure 2As shown, some embodiments of this application provide a high-precision stereo vision perception system for an inspection robot. The perception system is installed above the inspection robot and includes a housing 100. A data analysis and processing device is installed inside the housing 100. A sensor radar 130, a depth camera 140, and a gimbal camera 120 are disposed on the housing 100, and the relative positions of the sensor radar 130, depth camera 140, and gimbal camera 120 remain unchanged. The sensor radar 130 and depth camera 140 are disposed in front of the housing 100 to detect objects in front of the inspection robot and generate images to guide the movement of the inspection robot. The gimbal camera 120 is disposed on the top of the housing 100 to collect reading information from the dashboard. The sensor radar 130, depth camera 140, and gimbal camera 120 are respectively connected to the data analysis and processing device. The high-precision stereo vision perception system for the inspection robot provided in this application employs a combination of a sensor radar 130, a depth camera 140, and a gimbal camera 120. The sensor radar 130 and depth camera 140 are positioned in front of the housing 100 to detect objects in front of the inspection robot and generate images to guide its movement. By processing and analyzing the large amount of data collected by the sensor radar 130 and depth camera 140 in real time, the data analysis and processing device can quickly construct a precise three-dimensional spatial map of the nuclear power plant's pump room, providing the inspection robot with clear and accurate environmental perception information. This allows the inspection robot to accurately plan its movement path in the complex pump room environment, flexibly avoid various obstacles, and efficiently and safely move between different pieces of equipment. The gimbal camera 120 can be used to collect reading information from the instrument panels of the equipment inside the nuclear power plant's pump room. Furthermore, the sensing radar 130, by emitting high-frequency electromagnetic waves or lasers and receiving signals reflected back from target objects, can quickly and accurately acquire distance information of objects in the surrounding environment. Moreover, the sensing radar 130's detection is not limited by lighting conditions; it can work stably in dimly lit corners or areas with insufficient temporary lighting due to equipment maintenance. The sensing radar 130 also has a certain penetrating ability to water vapor and smoke, and can still acquire reliable environmental data even in special situations such as steam leaks in pump rooms. More importantly, when facing radiation environments, radar mapping technology demonstrates excellent stability and anti-interference capabilities, ensuring detection accuracy and better guiding the movement of inspection robots. The inspection robot can flexibly and precisely adjust the shooting angle and focal length through the PTZ camera 120, enabling it to quickly and accurately align with dashboards of different locations and sizes. The PTZ camera 120 is connected to and works with a data analysis and processing device, which uses advanced image recognition algorithms and deep learning technology to quickly process and analyze the collected dashboard images, accurately identify the position of pointers, scale values, and various warning signs, and realize the real-time and automatic collection of equipment operating parameters.The application of this technology not only greatly improves the efficiency and accuracy of data collection and effectively avoids errors that may occur during manual reading, but more importantly, by replacing manual operation with robots, it significantly reduces the exposure time of inspection personnel in the radiation environment, effectively protecting their health and safety. At the same time, the real-time and accurate collection of equipment operating parameters also provides reliable data support for the equipment status monitoring and fault early warning system of nuclear power plants, helping to promptly identify potential equipment failure hazards and take countermeasures in advance to ensure the safe and stable operation of nuclear power plants.

[0028] like Figure 1 As shown, in some embodiments of this application, the sensing radar 130 is preferably the omnidirectional stereo perception mid360 radar developed by Lanwo Technology. Developed by Lanwo Technology and launched in 2023, the mid360 radar is specifically designed for robots. It adopts a new architecture and achieves 360° horizontal omnidirectional stereo perception using self-developed hybrid solid-state technology. In the radar mapping technology system of inspection robots, the mid360 radar has unique advantages. In complex environments such as nuclear power plant pump rooms, the advantages of the mid360 radar are particularly evident. Pump rooms have crisscrossing pipes, dense equipment, complex spatial structures, and are subject to radiation and electromagnetic interference. The mid360 radar incorporates an active anti-interference design, effectively addressing the situation of multiple radar signals coexisting in nuclear power plant pump rooms. Even in complex electromagnetic environments, it can operate stably and continuously provide reliable environmental spatial information for the inspection robot. Furthermore, the mid360 radar is compact, with typical dimensions of 65mm in length, 65mm in width, and 60mm in height. This provides more options for the structural design of inspection robots within limited spaces, facilitating flexible deployment. In practical applications, the Mid360 radar can be used in conjunction with various mapping algorithms. Through Mid360 radar mapping, inspection robots can accurately plan their paths within the pump room of a nuclear power plant, moving between various pieces of equipment to efficiently complete inspection tasks and effectively ensure the safe and stable operation of nuclear power plant equipment.

[0029] like Figure 1 As shown, in some embodiments of this application, the horizontal field of view of the sensing radar 130 is 360°, and the vertical field of view is greater than 50°. The sensing radar 130 of this application employs a mid360 radar, which exhibits excellent field of view performance and a longer detection range compared to traditional single-line sensing radars and depth cameras. This ultra-wide field of view and long range advantage enables the inspection robot to quickly collect sufficient environmental feature points through the sensing radar 130, efficiently completing mapping, localization, and obstacle avoidance tasks. Furthermore, the non-repeating scanning mode in the mid360 radar can obtain a higher density point cloud through time integration of multiple frames of point cloud. As time accumulates, the angular resolution of the point cloud continuously improves, further enhancing the perception capability of environmental details.

[0030] like Figure 1 As shown, in some embodiments of this application, the sensing radar 130 is installed in front of the housing 100 by embedding, and the hemispherical window of the sensing radar 130 is exposed on the housing 100. When the sensing radar 130 is embedded, the exposed hemispherical window has a diameter of only 45mm and a height of 25mm, which is small in size, easy to install, and can be flexibly deployed.

[0031] like Figure 2 As shown, in some embodiments of this application, an inclined surface 113 is provided at the front of the housing 100, and a radar mounting hole 111 is formed on the inclined surface 113. A hemispherical window is disposed on the inclined surface 113 and embedded in the radar mounting hole 111. The angle between the inclined surface 113 and the horizontal plane is between 40° and 90°. The sensing radar 130 is deployed at the front of the housing 100, and preferably placed at an angle of 40° to 90° to better utilize the ultra-wide field of view of the sensing radar 130, acquire detailed point cloud data in front of the inspection robot, and help the inspection robot construct a three-dimensional image of the surrounding environment.

[0032] like Figure 1 As shown, in some embodiments of this application, the depth camera 140 is preferably a RealSense D435 depth camera, which includes two infrared sensors, an infrared laser emitter, and a color camera. The RealSense D435 is a powerful depth camera 140 with a maximum binocular depth resolution of 1280×720 and an RGB resolution of 1920×1080, a depth detection range of 0.2m-10m, and a maximum depth video stream of 90FPS. The RealSense D435 depth camera uses ActiveStereoscopic depth technology to generate 3D point clouds, where each point has (x,y,z) coordinates corresponding to the position of an object's surface in the scene. When using the RealSense D435 depth camera for 3D mapping, point cloud data must first be acquired. Using the API in the RealSense SDK, depth and color images from the depth camera 140 can be obtained. Converting the depth images into point cloud data yields the 3D point cloud from the camera's perspective. Subsequently, the 3D model can be reconstructed using the point cloud data and reconstruction algorithms. In practical applications, it is often used in conjunction with the RTAB-Map (Real-time Appearance Mapping) software package, a powerful software for 3D mapping and navigation using depth cameras, with corresponding software packages available for seamless integration. RTAB-Map utilizes depth images received by the depth camera 140 to perform graph-based SLAM (Simultaneous Localization and Mapping), generating dense color point clouds and camera odometry. It can also perform real-time loop closure detection, thereby achieving high-precision 3D mapping.

[0033] Furthermore, in some embodiments of this application, two infrared sensors are respectively positioned on the left and right sides of the depth camera 140. By using the two infrared sensors to collect depth data and with the aid of an infrared dot matrix emitter as an auxiliary light source, a precise three-dimensional point cloud can be generated, so that each point has (x, y, z) coordinates, corresponding to the position of an object surface in the scene.

[0034] like Figure 1 and Figure 2 As shown, in some embodiments of this application, a camera mounting hole 112 is provided at the front of the housing 100, and the depth camera 140 is installed in the camera mounting hole 112 by embedding, and the front end of the depth camera 140 is exposed on the housing 100, so that it can be flexibly deployed.

[0035] like Figure 1 and Figure 2 As shown, in some embodiments of this application, the gimbal camera 120 is a SIYIA8mini camera gimbal. This gimbal possesses excellent stability and flexibility, and its lightweight design facilitates integration into the inspection robot without affecting the robot's overall mobility. Accurate detection of the instrument panel is crucial during the inspection of a nuclear power plant pump room. Therefore, at the algorithm level, the SIYIA8mini camera gimbal uses an improved YOLOv5 algorithm. The original YOLOv5 algorithm has already demonstrated powerful performance in object detection, capable of quickly identifying various target objects in images. For the instrument panel detection scenario in a nuclear power plant pump room, the improved YOLOv5 algorithm further optimizes the recognition accuracy of key features such as the instrument panel pointer position, scale values, and warning signs. By introducing an attention mechanism, the algorithm can focus more intently on the instrument panel area, reducing interference from other irrelevant background information and improving detection speed and accuracy. In addition, to address issues such as blurring and glare that may occur in instrument panel images under complex lighting conditions in pump rooms, the improved algorithm has implemented targeted processing in the data augmentation stage, enhancing the model's adaptability to different lighting environments and ensuring stable and accurate detection of the instrument panel status under various operating conditions.

[0036] like Figure 1 and Figure 2 As shown, in some embodiments of this application, the gimbal camera 120 is provided with a rotation mechanism 122, which drives the gimbal camera 120 to rotate. Furthermore, the gimbal camera 120 is also provided with a pitch mechanism 123, which drives the gimbal camera 120 to perform pitch movements. By setting the rotation mechanism 122 and the pitch mechanism 123, the gimbal camera 120 possesses high-precision rotation and pitch control capabilities, ensuring that the mounted vision device 121 can always accurately focus on the dashboard, acquiring clear and distortion-free images of the dashboard regardless of the inspection robot's complex operating posture.

[0037] like Figure 1 and Figure 2 As shown, in some embodiments of this application, a connecting frame 110 manufactured using 3D printing technology is provided on the housing 100. The sensing radar 130, depth camera 140, and gimbal camera 120 are all mounted on the connecting frame 110. The connecting frame 110 is specially designed to adapt to the sensing radar 130, depth camera 140, and corresponding gimbal camera 120, fully considering the needs of their collaborative operation and the complex environmental characteristics of the nuclear power plant pump room, ensuring a stable and accurate connection between the sensing radar 130, depth camera 140, and gimbal camera 120. Firstly, for the sensing radar 130, since the sensing radar 130 needs to stably acquire environmental information during operation, and the pan-tilt camera 120 needs to flexibly adjust its visual direction to detect devices such as dashboards, the connecting frame is designed with a precise docking structure. Specifically, in some embodiments, the connection between the connecting frame 110 and the sensing radar 130 is provided with radar mounting holes 111, which are customized according to the dimensions of the mid360 radar (65mm long, 65mm wide, and 60mm high). Through the tightly fitting slots and screw fixing points, it is ensured that the sensing radar 130 does not become loose after being installed on the connecting frame 110. There will be some shaking, which is to prevent vibration from affecting the accuracy of radar data acquisition. Then, a camera mounting hole 112 is opened at the front of the connecting frame 110. The position of the camera mounting hole 112 is lower than that of the radar mounting hole 111. The depth camera 140 is embedded and fixed inside the camera mounting hole 112 to ensure the stability of the depth camera 140. Finally, for the gimbal camera 120, several screw holes are provided at the top of the connecting frame. The gimbal camera 120 can be precisely fixed to the connecting frame through the screw holes, ensuring that the positions of the sensing radar 130, the depth camera 140 and the gimbal camera 120 are relatively fixed. In addition, the use of 3D printing technology to manufacture the connecting frame 110 has many advantages. On the one hand, 3D printing can realize the one-piece molding manufacturing of complex structures. For the radar and gimbal connecting frame 110, its internal structure may need to be designed with reinforcing ribs, weight reduction holes and other structures to reduce weight while ensuring strength. 3D printing technology can easily realize these complex designs, which are difficult to achieve with traditional manufacturing processes. On the other hand, 3D printing has a very high degree of customization. In the environment of a nuclear power plant pump room, there may be special circumstances such as space constraints. 3D printing can quickly customize the frame according to the shape and size of the actual installation space, ensuring that the frame can perfectly fit the installation environment and facilitate the flexible deployment of inspection robots in limited spaces.

[0038] Furthermore, in some embodiments of this application, the connecting frame 110 is also equipped with an adjustable joint structure corresponding to the gimbal camera 120, allowing the gimbal camera 120 to flexibly adjust its pitch and rotation angles within a certain range to meet the needs of instrument panel detection at different positions. Simultaneously, it ensures that the gimbal camera 120 maintains a relatively stable positional relationship with the sensing radar 130 and depth camera 140 during rotation, enabling precise spatial matching between the environmental information acquired by the sensing radar 130 and depth camera 140 and the visual information of the gimbal camera 120. In practical applications, the connecting frame 110 greatly ensures the collaborative work of the sensing radar 130, depth camera 140, and gimbal camera 120. During the inspection of a nuclear power plant pump room, the inspection robot relies on the sensing radar 130 and depth camera 140 to construct an environmental map. Based on the provided environmental information, the gimbal camera 120 quickly and accurately adjusts to a suitable angle to inspect the equipment instrument panel. The stable structural design of the connecting frame 110 enables the sensing radar 130, depth camera 140 and gimbal camera 120 to work stably and collaboratively in the complex pump room environment, facing the vibration of the inspection robot's movement and the interference generated by the equipment operation. This continuously provides the inspection robot with reliable environmental perception and visual information acquisition capabilities, effectively promoting the efficient and accurate completion of inspection tasks and safeguarding the safe and stable operation of nuclear power plant equipment.

[0039] It is understood that the above embodiments only illustrate preferred embodiments of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that, for those skilled in the art, without departing from the concept of this application, the above technical features can be freely combined, and several modifications and improvements can be made, all of which fall within the protection scope of this application. Therefore, all equivalent transformations and modifications made within the scope of the claims of this application should fall within the coverage of the claims of this application.

Claims

1. A high-precision stereo vision perception system for an inspection robot, characterized in that, The sensing system is installed above the inspection robot. The sensing system includes a housing, inside which a data analysis and processing device is installed. The housing is equipped with a sensor radar, a depth camera, and a gimbal camera, and the relative positions of the sensor radar, the depth camera, and the gimbal camera remain unchanged. The sensor radar and the depth camera are positioned in front of the housing to detect objects in front of the inspection robot and generate images to guide the inspection robot's movement. The gimbal camera is positioned on the top of the housing to collect readings from the instrument panel. The sensor radar, the depth camera, and the gimbal camera are respectively connected to the data analysis and processing device.

2. The high-precision stereo vision perception system for inspection robots as described in claim 1, characterized in that, The sensing radar is a mid360 radar with omnidirectional three-dimensional perception developed by Lanwo Technology.

3. The high-precision stereo vision perception system for inspection robots as described in claim 1, characterized in that, The depth camera is a RealSense D435 depth camera, which includes two infrared sensors, an infrared laser emitter, and a color camera. The two infrared sensors are respectively located on the left and right sides of the depth camera.

4. The high-precision stereo vision perception system for inspection robots as described in claim 1, characterized in that, The gimbal camera is a SIYIA8mini camera gimbal.

5. The high-precision stereo vision perception system for inspection robots as described in claim 1, characterized in that, An inclined surface is provided at the front of the housing, and the angle between the inclined surface and the horizontal plane is between 40° and 90°. The sensing radar is disposed on the inclined surface, and the hemispherical window of the sensing radar is exposed outside the housing.

6. The high-precision stereo vision perception system for inspection robots as described in claim 5, characterized in that, The depth camera is embedded in the front of the housing, and the height of the depth camera is lower than the height of the sensing radar.

7. The high-precision stereo vision perception system for inspection robots as described in claim 4, characterized in that, The gimbal camera is equipped with a rotating mechanism, which drives the gimbal camera to rotate.

8. The high-precision stereo vision perception system for inspection robots as described in claim 4, characterized in that, The gimbal camera is equipped with a pitch mechanism, which drives the gimbal camera to perform pitch movements.

9. The high-precision stereo vision perception system for inspection robots as described in any one of claims 1 to 8, characterized in that, The housing is provided with a connecting frame made using 3D printing technology, and the sensing radar, the depth camera, and the gimbal camera are all mounted on the connecting frame.

10. The high-precision stereo vision perception system for inspection robots as described in claim 9, characterized in that, The connecting frame is equipped with an adjustable joint structure corresponding to the gimbal camera.

Images