A kind of inspection robot for coal mine underground special environment and use method

Abstract

The application discloses a kind of inspection robots for special environment in coal mine and use method, it is related to coal mine safety production technical field.A kind of inspection robots for special environment in coal mine, including walking track, further include: walking mechanism, main body frame, detection unit, power supply unit, communication unit and holder shield subassembly, wherein, the walking mechanism is cooperated with walking track and realizes movement, the detection unit is realized by holder shield subassembly instrument image acquisition, the power supply unit and communication unit are respectively for whole machine power supply and data transmission;The application ensures stable operation under high-risk environment by adopting explosion-proof safety design, realizes the high-precision positioning and identification of instrument by dynamic search algorithm and intelligent detection algorithm, has unmanned inspection, real-time data transmission function, its modular structure can be flexibly extended, can adapt to a variety of instrument detection needs, significantly improves coal mine inspection efficiency and intelligent level.

Description

Technical Field

[0001] This invention belongs to the field of coal mine safety production technology, specifically, it relates to an inspection robot designed for the special environment of underground coal mines. Background Technology

[0002] The underground working environment in coal mines is complex and dangerous, posing numerous safety hazards such as high humidity, high dust levels, strong electromagnetic interference, and flammable and explosive gases. In such an environment, accurate and timely meter readings of various equipment and instruments are crucial to ensuring safe production in coal mines. Traditional manual meter reading methods are inefficient, costly, and expose workers to hazardous environments for extended periods, posing significant safety risks.

[0003] While the application of track-mounted inspection robots has improved meter reading to some extent, existing inspection robots still have shortcomings in terms of structural design and detection algorithms. On the one hand, their structural design is difficult to fully adapt to the harsh environment underground in coal mines, and cannot fully guarantee the stable operation of robots in high-risk scenarios. On the other hand, intelligent instrument positioning and detection methods have defects such as low detection efficiency and insufficient accuracy when dealing with problems such as robot positioning errors, equipment position changes, diverse instrument shapes, and poor underground lighting conditions, making it difficult to meet the needs of coal mine safety production for high-precision and real-time acquisition of instrument data.

[0004] Therefore, there is an urgent need for an inspection robot system and instrument testing solution that is innovative and optimized in both structural design and testing methods for the special environment of underground coal mines. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide an inspection robot for the special environment of underground coal mines that can overcome or at least partially solve the above problems.

[0006] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows:

[0007] An inspection robot designed for the special environment of underground coal mines includes a walking track, a walking mechanism, a main frame, a detection unit, a power supply unit, a communication unit, and a gimbal cover assembly. The walking mechanism works in conjunction with the walking track to achieve movement, the detection unit acquires instrument images through the gimbal cover assembly, and the power supply unit and communication unit provide power and data transmission for the entire machine, respectively.

[0008] Preferably, the walking mechanism includes two symmetrically arranged drive wheels, two pairs of load-bearing wheel assemblies, two speed measuring wheel assemblies, two limit wheel assemblies, two pairs of side guard wheel assemblies, and a robot drive assembly mounted on the main frame. The robot drive assembly is used to drive the drive wheels. The load-bearing wheel assemblies are rolled into the inner bottom of the grooves on both sides of the walking track. The speed measuring wheel assemblies and the limit wheel assemblies are both pressed against the bottom surface of the walking track. The two pairs of side guard wheel assemblies are respectively locked on both sides of the lower end of the walking track.

[0009] Preferably, the main frame is provided with an explosion-proof main cavity, a battery cavity, an explosion-proof breathing and drainage valve and a proximity switch. A photoelectric distance sensor is installed on the outside of the main frame, and a control unit is installed inside the main frame. A lower protective cover assembly is fixedly provided at the bottom of the main frame.

[0010] Preferably, the detection unit includes a gimbal cover assembly and a front-facing camera mounted on the main frame. The gimbal cover assembly integrates a visible light camera, an infrared camera, a fill light, a horizontal rotation motor, and a vertical rotation motor.

[0011] Preferably, the power supply unit includes a battery pack installed inside the battery cavity, and an automatic charging component is installed on the battery pack.

[0012] Preferably, the communication unit includes a wiring cavity and a control unit mounted on the main frame, and the wiring cavity is provided with a network interface.

[0013] Furthermore, the gimbal cover assembly includes an annular cover mounted on the main frame, a column rotatably mounted inside the annular cover, a slot at the lower end of the column, a turntable rotatably mounted inside the slot, a visible light camera, an infrared camera, and a fill light distributed in a triangular shape on the outer wall of the turntable, a vertical rotation motor fixedly mounted on the outer wall of the column, the output shaft of the vertical rotation motor fixedly connected to the end of the turntable shaft, a horizontal rotation motor fixedly mounted inside the annular cover, and the output shaft of the horizontal rotation motor connected to the outer wall of the column through two meshing transmission gears.

[0014] Furthermore, the outer wall of the turntable is provided with a fan-shaped groove. The visible light camera and the infrared camera are both rotatably mounted in the fan-shaped groove via a rotating rod. A torsion spring is installed between the outer wall of the rotating rod and the inner wall of the fan-shaped groove. A rubber membrane is installed between the outer wall of the visible light camera and the infrared camera and the inner wall of the port of the fan-shaped groove. The top of the column is provided with a through hole. A cooling fan is fixedly installed in the through hole. A filter element is installed at the upper port of the through hole. The upper end of the turntable is provided with an air inlet groove aligned with the through hole. A connecting pipe is provided in the air inlet groove and extends into the fan-shaped groove. An exhaust hole is provided in the fan-shaped groove and extends into the outer wall of the turntable.

[0015] Furthermore, passive magnets are fixedly installed on the outer walls of both the visible light camera and the infrared camera. A circular groove is provided at the end of the turntable shaft, and a disc is provided in the groove. The disc is fixedly connected to the inner wall of the groove by a crossbar. Two symmetrically arranged active magnets are fixedly installed on the outer wall of the disc. L-shaped elastic plates are fixedly connected to the inner walls on both sides of the groove, and the other end of the L-shaped elastic plate abuts against the inner wall of the turntable.

[0016] A method for using an inspection robot includes the following steps:

[0017] S1. Equipment Start-up and Initialization: The control unit initializes the walking mechanism self-test, calibrates the detection unit, and performs communication tests;

[0018] S2. Path movement: According to the inspection route issued by the ground control center, the drive wheels move along the walking track, and the proximity switch monitors obstacles;

[0019] S3. Instrument Positioning: After reaching the detection point, the visible light camera and infrared camera are adjusted by the horizontal and vertical rotating motors. The visible light camera acquires images, and the instrument is positioned using a dynamic search algorithm for instrument position.

[0020] S4. Data Acquisition: High-magnification zoom is used to capture instrument images, and the scale is identified and pointer readings are calculated using the U-Net algorithm;

[0021] S5. Data Reporting: Upload the encrypted test results and return to standby mode after completing all locations;

[0022] S6. Anomaly Handling: Trigger a re-inspection for out-of-limit data; switch to redundancy mode to complete the task in case of equipment failure.

[0023] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0024] 1. High Safety and Stability: The explosion-proof and intrinsically safe inspection robot system for coal mines adopts an explosion-proof and intrinsically safe design. Each unit has excellent explosion-proof and protective performance, enabling safe and stable operation in the high-risk environment of underground coal mines. This effectively protects equipment and personnel safety and reduces the probability of accidents. Furthermore, the system's rational structural design and coordinated operation of all parts enhance the robot's reliability and anti-interference capabilities.

[0025] 2. Precise and Efficient Instrument Inspection: The intelligent instrument positioning and inspection method, combined with innovative algorithms, enables rapid and accurate instrument location and high-precision inspection of pointer instruments. The dynamic instrument position search algorithm solves the inspection challenges caused by robot positioning errors and equipment position changes, ensuring the acquisition of high-quality instrument images. The intelligent pointer instrument inspection algorithm is optimized for the special conditions in underground coal mines, effectively improving the accuracy and reliability of instrument reading detection and meeting the high-precision data acquisition requirements for safe coal mine production.

[0026] 3. High level of intelligence and automation: This invention automates the testing of instruments for underground equipment in coal mines, eliminating the need for manual intervention, greatly improving inspection efficiency and reducing labor costs. Simultaneously, through advanced communication technology, it enables real-time data transmission and remote control between the robot and the ground control center, allowing staff to promptly grasp the operating status of underground equipment and enhancing the intelligent management level of coal mine safety production.

[0027] 4. Strong adaptability and scalability: The modular design of the inspection robot system allows for flexible configuration and expansion of functional modules according to different working environments and inspection needs in coal mines. The intelligent instrument positioning and detection method can also adapt to the detection of different types of instruments by adjusting algorithm parameters and training data, exhibiting strong versatility and adaptability, providing powerful support for the intelligent development of coal mines.

[0028] 5. This invention utilizes a cooling fan to continuously exhaust air downwards from the lower port of the through-hole. As the air passes through the fan-shaped groove, it carries away some of the heat from the visible light and infrared cameras, thereby achieving heat dissipation for the visible light and infrared cameras and maintaining their good stability.

[0029] 6. This invention uses a vertical rotating motor to drive the turntable to rotate at a certain angle, which in turn causes the visible light camera and the infrared camera to rotate synchronously. The visible light camera and the infrared camera are then housed in the top of the slot to prevent the lenses from being contaminated by foreign objects.

[0030] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0031] In the attached diagram:

[0032] Figure 1 This is a schematic diagram of the front view structure of an inspection robot for the special environment of underground coal mines proposed in this invention.

[0033] Figure 2 This invention proposes an inspection robot designed for the unique environment of underground coal mines. Figure 1 Schematic diagram of part A in the middle;

[0034] Figure 3 This is a schematic diagram of a partial isometric structure of an inspection robot designed for the special environment of underground coal mines, as proposed in this invention. Figure 2 ;

[0035] Figure 4 This is a schematic diagram of a partial isometric structure of an inspection robot designed for the special environment of underground coal mines, as proposed in this invention. Figure 2 ;

[0036] Figure 5 This is a partial cross-sectional structural diagram of an inspection robot for the special environment of underground coal mines proposed in this invention.

[0037] Figure 6 This is a schematic diagram of the turntable cross-section structure of an inspection robot for the special environment of underground coal mines proposed in this invention.

[0038] Figure 7 This is a schematic diagram of the turntable isometric structure of an inspection robot designed for the special environment of underground coal mines, as proposed in this invention.

[0039] Figure 8 This is a schematic diagram of a C-shaped tube structure for an inspection robot designed for the special environment of underground coal mines, as proposed in this invention.

[0040] In the diagram: 1. Explosion-proof main cavity; 2. Lower protective cover assembly; 166. Drive wheel; 183. Load-bearing wheel assembly; 198. Speed ​​measuring wheel assembly; 209. Side guard wheel assembly; 210. Limit wheel assembly; 220. Traveling track; 222. Proximity switch; 3. Photoelectric distance sensor; 4. Vertical rotary motor; 5. Horizontal rotary motor; 6. Annular protective cover; 601. Supplemental light; 602. Visible light camera; 603. Infrared camera; 7. Air intake port; 8. Column; 9. Turntable; 10. Slot; 11. Through hole; 12. Cooling fan; 13. Air inlet slot. ; 14. Connecting pipe; 15. Fan-shaped groove; 16. Exhaust port; 17. Rotating rod; 18. Rubber membrane; 19. Circular groove; 20. Disc; 21. Crossbar; 22. Active magnet; 23. Explosion-proof breathing and drain valve; 24. Wiring chamber; 25. Battery chamber; 26. Gimbal cover assembly; 27. Passive magnet; 28. Filter element; 29. ​​End cap; 30. C-shaped tube; 31. Strip cover; 32. Air intake groove; 33. Guide wheel; 34. Horizontal plate; 35. Cleaning brush; 36. Transmission gear; 37. L-shaped elastic plate; 61. Automatic charging assembly; 95. Front camera. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0042] Example 1: Refer to Figures 1-8 An inspection robot designed for the special environment of underground coal mines includes a walking track 220 installed on the top of the coal mine shaft, and also includes a walking mechanism, a main frame, a detection unit, a power supply unit, a communication unit, and a gimbal cover assembly 26. The walking mechanism works in conjunction with the walking track 220 to achieve movement, the detection unit acquires instrument images through the gimbal cover assembly 26, and the power supply unit and communication unit provide power and data transmission for the whole machine, respectively.

[0043] Specifically, during use, the walking mechanism can move the main frame on the walking track 220, the detection unit can scan and detect the equipment, the power supply unit can continuously supply power to the entire device, and the communication unit transmits the detected data to the cloud server.

[0044] The main frame is equipped with an explosion-proof main cavity 1, a battery cavity 25, an explosion-proof breathing and drain valve 23, and a proximity switch 222. The proximity switch 222 acts as a switch to locate the movement position of the main frame. Both the explosion-proof main cavity 1 and the battery cavity 25 have cavities inside and are made of Q345 steel. The battery cavity 25 weighs 2.56 kg. There are two sets of explosion-proof breathing and drain valves 23. Four photoelectric distance sensors 3 are installed on the outside of the main frame. The main frame is equipped with a control unit, which includes a processor and a sensor interface. A lower protective cover assembly 2 is fixed at the bottom of the main frame to protect the bottom. The power supply unit includes a battery pack installed in the battery cavity 25. The battery pack is a 48V / 20Ah intrinsically safe lithium battery pack. An automatic charging assembly 61 is installed on the battery pack. The automatic charging assembly 61 is made of 304 stainless steel, weighs 0.17 kg, and is used to achieve contactless charging. The battery lasts for ≥8 hours on a single charge.

[0045] The aforementioned detection unit includes a gimbal shield assembly 26 and a front-facing camera 95 mounted on the main frame. The gimbal shield assembly 26 is an independent explosion-proof module. The gimbal shield assembly 26 integrates a visible light camera 602, an infrared camera 603, a supplementary light 601, a horizontal rotation motor 5, and a vertical rotation motor 4. The supplementary light 601 provides illumination compensation, the infrared camera 603 supports temperature detection, and the visible light camera 602 supports 10x optical zoom. All of these support the robot detection system (including the camera) at -20°C. ° C-+60 °Operating stably in environment C, the horizontal rotary motor 5 and the vertical rotary motor 4 are used to adjust the horizontal and vertical angles of the visible light camera 602, the infrared camera 603, and the supplementary light 601.

[0046] The aforementioned communication unit includes a wiring cavity 24 and a control unit installed on the main frame. The wiring cavity 24 is equipped with a network interface, which mainly includes fiber optic and cable interfaces. The wiring cavity 24 has an internal cavity, is made of Q345 steel, weighs 1.45kg, supports fiber optic wired communication (explosion-proof interface) and 5G wireless communication (explosion-proof module), with a transmission rate of ≥100Mbps, and can transmit images and detection data back in real time.

[0047] Example 2: Refer to Figures 1-2 An inspection robot designed for the special environment of underground coal mines is basically the same as that in Example 1, but with a further improvement:

[0048] The aforementioned walking mechanism includes two symmetrically arranged drive wheels 166 mounted on the main frame, two pairs of load-bearing wheel assemblies 183, two speed measuring wheel assemblies 198, two limit wheel assemblies 210, two pairs of sidewall wheel assemblies 209, and a robot drive assembly 158. The robot drive assembly 158 is explosion-proof and drives the drive wheels 166. The load-bearing wheel assemblies 183 are rolled into the inner bottom of the grooves on both sides of the walking track 220. The speed measuring wheel assemblies 198 and the limit wheel assemblies 210 are both pressed against the bottom surface of the walking track 220. The two pairs of sidewall wheel assemblies 209... The drive wheels 166 are made of 2A12 aluminum alloy and there are 2 of them. The robot drive assembly 158 is made of Q345 steel and there are 2 of it. The load-bearing wheel assembly 183 is made of 2A12 aluminum alloy and there are 4 of it, each weighing 115.32 kg. The speed measuring wheel assembly 198 is made of 6063-T4 aluminum alloy and there are 3 of it. The limit wheel assembly 210 is made of 04 stainless steel and there are 2 of it, each weighing 0.04 kg. The side guard wheel assembly 209 has 6 of it and weighs 0.01 kg.

[0049] Specifically, during use, the robot drive assembly 158 is used to drive the drive wheel 166 to move along the walking track 220. During the movement, the side guard wheel assemblies 209 on both sides are used to prevent the entire main frame from derailing. The two pairs of load-bearing wheel assemblies 183 are mainly used to bear the weight of the entire device. The two speed measuring wheel assemblies 198 are used to realize speed measurement. The limit wheel assembly 210 is used to limit the up and down bumping of the main frame.

[0050] Example 3: Reference Figures 3-7 An inspection robot designed for the special environment of underground coal mines is basically the same as that in Example 2, but with a further improvement:

[0051] The aforementioned gimbal cover assembly 26 includes an annular cover 6 mounted on the main frame. The annular cover 6 is made of ABS material with an IP67 protection rating. A column 8 is rotatably mounted inside the annular cover 6. A slot 10 is provided at the lower end of the column 8, and a turntable 9 is rotatably mounted inside the slot 10. A visible light camera 602, an infrared camera 603, and a supplementary light 601 are distributed in a triangular shape and mounted on the outer wall of the turntable 9. A vertical rotation motor 4 is fixedly mounted on the outer wall of the column 8, and the output shaft of the vertical rotation motor 4 is fixedly connected to the shaft end of the turntable 9. A horizontal rotation motor 5 is fixedly mounted inside the annular cover 6, and the output shaft of the horizontal rotation motor 5 is connected to the outer wall of the column 8 through two meshing transmission gears 36. The output wires of the horizontal rotation motor 5 and the vertical rotation motor 4 are sealed by explosion-proof glands. The overall weight of the assembly is 0.36 kg.

[0052] Specifically, when it is necessary to adjust the horizontal direction of the visible light camera 602, the infrared camera 603, and the supplementary light 601, the horizontal rotation motor 5 is started. The horizontal rotation motor 5 can drive the column 8 to rotate inside the annular cover 6 through two meshing transmission gears 36. The column 8 will then drive the visible light camera 602, the infrared camera 603, and the supplementary light 601 on the turntable 9 to rotate horizontally. When it is necessary to adjust the vertical direction of the visible light camera 602, the infrared camera 603, and the supplementary light 601, the vertical rotation motor 4 is started. The vertical rotation motor 4 can drive the turntable 9 to rotate through the output shaft. The turntable 9 can then drive the visible light camera 602, the infrared camera 603, and the supplementary light 601 to adjust their positions in the vertical direction.

[0053] Example 4: Reference Figures 3-8 An inspection robot designed for the special environment of underground coal mines is basically the same as that in Example 3, but with a further improvement:

[0054] The outer wall of the turntable 17 is provided with a fan-shaped groove 15. The visible light camera 602 and the infrared camera 603 are both rotatably mounted in the fan-shaped groove 15 via the rotating rod 17. A torsion spring is installed between the outer wall of the rotating rod 17 and the inner wall of the fan-shaped groove 15. A rubber membrane 18 is installed between the outer wall of the visible light camera 602 and the infrared camera 603 and the inner wall of the port of the fan-shaped groove 15. The top of the column 8 is provided with a through hole 11. A cooling fan 12 is fixedly installed in the through hole 11. A filter element 28 for filtering dust is installed at the upper port of the through hole 11. The upper end of the turntable 9 is provided with an air inlet groove 13 aligned with the through hole 11. A connecting pipe 14 is provided in the air inlet groove 13 and extends into the fan-shaped groove 15. An exhaust hole 16 extends into the outer wall of the turntable 9.

[0055] Specifically, during use, the cooling fan 12 can continuously exhaust air downwards from the lower port of the through hole 11, while drawing in outside air from the upper port. Since the openings on both sides of the slot 10 are blocked by two L-shaped elastic plates 37, the air exhausted downwards from the through hole 11 will enter the air inlet slot 13, then enter the fan-shaped slot 15 through the connecting pipe 14, and finally be discharged from the exhaust hole 16. When the air passes through the fan-shaped slot 15, it will carry away some of the temperature from the visible light camera 602 and the infrared camera 603, thereby achieving heat dissipation for the visible light camera 602 and the infrared camera 603 and maintaining good stability for the visible light camera 602 and the infrared camera 603.

[0056] The top of the column 8 is rotatably connected to an end cap 29 that communicates with the through hole 11. C-shaped tubes 30 are fixedly connected to both sides of the end cap 29. Strip covers 31 are fixedly connected to the ends of the two C-shaped tubes 30. Suction grooves 32 are opened on opposite sides of the two strip covers 31. Guide wheels 33 are rotatably installed on opposite sides of the two strip covers 31. Suction holes 7 are provided on the top of the end cap 29. A horizontal plate 34 is fixedly connected inside the end cap 29. A cleaning brush 35 that penetrates to the top of the suction groove 32 is fixedly connected on the horizontal plate 34.

[0057] Specifically, during use, the two strip covers 31 need to be fastened to the grooves on both sides of the travel track 220, with the air intake 7 facing the lower end face of the travel track 220. When the whole thing moves, the guide wheel 33 is used to prevent the strip covers 31 from scraping against the outer wall of the travel track 220, while the cleaning brush 35 will sweep along the bottom of the travel track 220. When the through hole 11 sucks air through the top, the end cap 29 can suck air through the C-shaped tube 30 to the strip cover 31. The strip cover 31 will suck up the dust on both sides of the travel track 220 through the air intake groove 32, so that the drive wheels 166 on both sides will not slip, and the air intake 7 will suck air to the bottom of the travel track 220, thereby sucking away the dust at the bottom of the travel track 220, so that the bottom limit wheel assembly 210 and the speed measuring wheel assembly 198 will not slip.

[0058] Example 5: Refer to Figures 5-6 An inspection robot designed for the special environment of underground coal mines is basically the same as that in Example 4, but with a further improvement:

[0059] The outer walls of the visible light camera 602 and the infrared camera 603 are both fixedly equipped with passive magnets 27. The shaft end of the turntable 9 is provided with a circular groove 19, and a disc 20 is provided in the circular groove 19. The disc 20 is fixedly connected to the inner wall of the slot 10 through a crossbar 21. Two symmetrically arranged active magnets 22 are fixedly installed on the outer wall of the disc 20. L-shaped elastic plates 37 are fixedly connected to the inner walls on both sides of the slot 10. The other end of the L-shaped elastic plate 37 is pressed against the inner wall of the turntable 9.

[0060] Specifically, when the inspection stops, the vertical rotary motor 4 drives the turntable 9 to rotate 180°, which in turn causes the visible light camera 602 and the infrared camera 603 to rotate synchronously. The visible light camera 602 and the infrared camera 603 are then housed inside the top of the slot 10 to prevent contamination of the lenses by foreign objects. The air inlet slot 13 is moved away from the lower end of the through hole 11, so the air at the lower end of the through hole 11 can only be discharged from both sides of the slot 10. This pushes the L-shaped elastic plate 37, creating a gap between it and the outer wall of the turntable 9. When the air is discharged from the gap, it carries away the visible light camera 602 and the infrared camera 603. The dust on the external camera 603 is automatically removed, thus cleaning the visible light camera 602 and the infrared camera 603. At this time, the passive magnet 27 is attracted, causing the visible light camera 602 and the infrared camera 603 to deflect via the rotating rod 17. In other words, the shooting direction of the visible light camera 602 and the infrared camera 603 will be perpendicular to the axis of the turntable 9. As a result, the lenses of the visible light camera 602 and the infrared camera 603 will be tangent to the outer circumference of the turntable 9. Thus, the flowing air can flow better over the lens surface of the visible light camera 602 and the infrared camera 603, improving the cleaning efficiency.

[0061] Example 6: Refer to Figures 1-8 A method for using an inspection robot includes the following steps:

[0062] S1. Equipment Start-up and Initialization: After the robot body is started via command from the ground control center or by operating the rotary switch, the explosion-proof power supply (integrated in the battery cavity 25) supplies power to the whole machine. The control unit (explosion-proof main cavity 1) initializes each module in sequence: First, the walking mechanism performs a self-test, using proximity switch 222 to detect the alignment status of the drive wheel 166, load-bearing wheel assembly 183, and other wheels with the track; then, the detection unit is calibrated, including the rotation accuracy test of the vertical rotary motor 4 and the horizontal rotary motor 5, the autofocus of the visible light camera 602 and the infrared camera 603, and the brightness adjustment of the supplementary light 601; at the same time, the fiber optic and 5G module connection test of the communication unit is completed to ensure stable data transmission; finally, the photoelectric distance sensor 3 collects environmental data in real time, and combines it with the gyroscope to complete the robot's initial position calibration.

[0063] S2. Start the inspection task and path movement: The ground control center sends out the inspection route containing the coordinates of the instrument points through the communication unit. The control unit parses the route and generates a path plan. The robot starts the track walking mode. The drive motor (integrated in the robot drive component 158) drives the drive wheel 166 to move along the robot walking track 220. The load-bearing wheel component 183 bears the weight of the whole machine. The speed measuring wheel component 198 provides real-time feedback of speed (default 1m / s). The limit wheel component 210 and the edge wheel component 209 work together to constrain the track boundary. The proximity switch 222 monitors obstacles in real time to prevent derailment. When the battery power is lower than 20%, the return charging process is automatically triggered. The automatic charging component 61 enables contactless charging and battery life.

[0064] S3. Arrival at the Point and Instrument Positioning Search: When the robot arrives at the preset detection point via the speed measuring wheel assembly 198 and the electronic map, the control unit sends an instruction to the detection unit: the horizontal rotary motor 5 and the vertical rotary motor 4 work together to adjust the gimbal cover assembly 26 to the target direction and the estimated height (default -30° to +30°), and the visible light camera 602 acquires wide-angle images at 3x low-magnification zoom for instrument positioning.

[0065] Traditional instrument positioning algorithms face two major bottlenecks in underground coal mines: ① Complex background interference leads to a high false detection rate based on the traditional YOLO model, and the model's computational load is too large (parameter count > 10M), making it difficult to run multiple algorithms in real time on embedded devices; ② Changes in instrument position and deviations in robot positioning require searching for the instrument's location. Random search strategies typically require more than 15 iterations to locate the target, taking over 8 seconds. This patent uses a dynamic instrument position search algorithm (step S301) to quickly identify the rough outline of the instrument. If corner features of the instrument are detected (through Harris corner detection), the actual position is predicted using Kalman filtering, combined with the instrument's geometric model (e.g., the aspect ratio of a rectangle), and the gimbal is controlled to fine-tune the viewing angle with a 0.5° step. If no features are detected, the robot moves slightly forward and backward (±0.5m) using its drive wheels to expand the search range until the instrument outline is locked (step S302).

[0066] S4. Data Acquisition and Instrument Numerical Calculation: After confirming the instrument's location, the visible light camera 602 switches to 10x high-magnification zoom, and the supplementary light 601 automatically adjusts its brightness (0-1000 lux) according to the ambient light to ensure that the instrument image resolution is ≥4032×3024. In the data processing stage, the visible light image uses the U-Net algorithm to identify the scale position and calculate the scale start and end points and scale values ​​(steps S401, 402). Then, the RANSAC-optimized least squares method is used to fit the center of the circle, and finally the pointer reading is calculated (step S403).

[0067] S5. Data Reporting and Task Cycle: Detection results (including images, readings, and temperature data) are encrypted and uploaded to the ground control center via fiber optic or 5G communication modules (transmission rate ≥100Mbps). After a single-point task is completed, the robot moves to the next detection point according to a preset path and repeats the process. After all points are detected, the robot returns to the charging base station or standby point to wait for subsequent instructions.

[0068] S6. Abnormal Handling Mechanism: If the detection data exceeds the limit or the equipment temperature is abnormal, the robot will alarm in real time through the communication unit and automatically trigger a partial re-inspection (such as repeating the positioning test 3 times) to eliminate false detections; if the gimbal or walking mechanism malfunctions (such as motor jamming), the control unit will automatically switch to the redundancy mode (such as single-axis drive), prioritize completing the current detection task and then return to maintenance to ensure the continuity of the detection process and the reliability of the data.

[0069] S301 Instrument Coarse Feature Rapid Detection

[0070] To address the issues of high false positive rates and excessive computational demands (>10M parameters) associated with traditional YOLO models due to complex background interference, making it difficult to run multiple algorithms in real-time on embedded devices, this patent employs the YOLOv11s framework for instrument detection, integrating knowledge distillation and structured pruning techniques. The teacher model uses YOLOv11x (54.7M parameters), while the student model uses YOLOv11s. The models have been lightweighted by reducing the number of YOLOv11s parameters from 9.4M to 4.3M through channel pruning compression and replacing standard convolutions with depthwise separable convolutions. On the NVIDIA Jetson TX2 platform, the measured inference speed reached 200FPS, a significant improvement over the original. The distillation loss function is defined as: L... distill =λ cls L cls +λ obj L obj +λ box L box

[0071] Where L distill L represents the total loss value of the distillation loss function. cls For the classification loss function, L obj Let L be the target detection loss function. box Let λ be the bounding box regression loss function. cls λ is the weighting coefficient for the classification loss, with a value of 0.5. obj λ represents the weighting coefficient for the target detection loss, with a value of 0.7. box , which is the weight coefficient of the bounding box regression loss, with a value of 0.3, and achieves knowledge transfer between feature layers through weight allocation.

[0072] The original formula for the computational complexity of a convolutional network is:

[0073] FLOPs1=K² * C_in * C_out * H_out * W_out

[0074] The optimized formula for network computation is:

[0075] FLOPs2=(K² * C_in + C_in * C_out) * H_out * W_out

[0076] Compared to the original network, the optimization reduces the network size by:

[0077] FLOPs2 / FLOPs1 = 1 / C_out + 1 / K²

[0078] The meanings of each parameter are as follows:

[0079] K: Kernel size

[0080] C_in: Number of channels for the input feature

[0081] C_out: Number of channels for the output feature

[0082] H_out, W_out: The height and width of the output feature map

[0083] S302 Adaptive Corner-Guided Search

[0084] Due to changes in the instrument's position and deviations in each robot positioning, the detection algorithm in S301 needs to be used to search for the instrument's location. However, due to the uncertainty of the instrument's position, the random search strategy requires an average of more than 15 iterations to locate the target, taking more than 8 seconds. This patent proposes an adaptive corner-guided search method to accelerate the instrument position search speed. The algorithm flow is as follows:

[0085] (1) Algorithm premise and input / output

[0086] Basic requirements:

[0087] 1. The S301 instrument detection algorithm has output the coarse feature detection results of the instruments in the current image (if the instrument outline exists, the ROI region is returned; otherwise, an empty region is returned).

[0088] 2. The gimbal has dual-axis control capability in horizontal (±180°) and vertical (-90°-+45°), with an angular resolution of 0.1° and a positioning error of ±0.5°.

[0089] enter:

[0090] Current gimbal angle: Horizontal angle α cur Vertical angle β cur.

[0091] S301 Detection Result: The coordinates of the instrument's ROI area (u,v,w,h) are either empty or null.

[0092] Output:

[0093] The angle of the gimbal at the next moment (α) next ,β next ).

[0094] Search status: Continue searching / Location successful / Switch to global scan.

[0095] (2) Core search logic flow

[0096] (2.1) State initialization and region partitioning

[0097] Preset search range:

[0098] Horizontal direction: (α) cur -60°, α cur +60° (Prioritize searching the area directly in front ±60°, which conforms to the installation rules of downhole instruments).

[0099] Vertical direction: (β) cur -30°, β cur +15° (This takes into account both ground and wall-mounted instruments, 30° downwards and 15° upwards).

[0100] Regional division strategy:

[0101] - Divide the search area into a central area (±30° horizontal, ±15° vertical) and an edge area (the remaining area), with the central area having higher priority than the edge area.

[0102] (2.2) Hierarchical response based on S301 results

[0103] Scenario 1: S301 detected the complete instrument ROI (area ≥ 80% of preset value)

[0104] Once the location is determined to be successful, the search stops and the S4 fine focusing process is triggered.

[0105] Scenario 2: S301 detected some instrument features (such as corners and contours ≥50%).

[0106] Extract the center pixel coordinates C(u) from the ROI. c ,v c Given the pixel coordinates of the image center point, calculate the biaxial deviation:

[0107] -Horizontal deviation: Δα=K u ×(u c -u center )

[0108] - Vertical deviation: Δβ=K v ×(v c -v center )

[0109] Among them, u c ,v c The center pixel coordinates of the instrument area, u center v center K represents the center pixel coordinates of the image. u K v These are the conversion coefficients (° / pixel) from pixel deviation in the horizontal and vertical directions to angular deviation, respectively, calculated as follows:

[0110] Then we can calculate the next gimbal adjustment strategy:

[0111] α next =α cur +Δα×η

[0112] β next =β cur +Δβ×η

[0113] α cur β cur α represents the current horizontal and vertical angles of the gimbal. next β next The horizontal and vertical angles that the gimbal needs to reach next.

[0114] η=0.8 is the adaptive damping coefficient to avoid overshoot (excessive movement may exceed the instrument's range). After adjustment, if the movement does not exceed the central area, it will move towards the central area first; if it exceeds the central area, it will be limited to the central area boundary.

[0115] Scenario 3: No instrument detected in S301 (ROI is empty)

[0116] Then, the edge region scanning strategy is executed, that is, the edge regions are scanned sequentially in a spiral expansion order, and the steps are as follows:

[0117] 1. Increase the horizontal angle by 15°, while keeping the vertical angle at its current value (expanding to the right).

[0118] 2. If not detected, the horizontal angle reverts to its initial value, and the vertical angle decreases by 10° (expanding downwards).

[0119] 3. If it is still not detected, the horizontal angle is reduced by 15°, while the vertical angle remains at its current value (expanding to the left).

[0120] 4. If no detection is detected, the horizontal angle reverts to its initial value, and the vertical angle increases by 10° (extending upwards).

[0121] After each scan, S301 detection is performed. If no detection is detected for 4 consecutive steps, the global scan mode (horizontal ±90°, vertical ±45°, step size 20°) is triggered.

[0122] (2.3) False detection elimination and termination conditions

[0123] False positive determination:

[0124] If an ROI is detected twice consecutively but its area is less than 30% of the preset value and it has no corner features, it is determined to be dust noise or a false target, and the search status is reset.

[0125] Termination conditions:

[0126] - Successful location (Case 1).

[0127] - The global scan is complete but no instrument is detected. The system returns "Search timed out" and awaits manual intervention.

[0128] This algorithm employs a spatial priority partitioning strategy, prioritizing the scanning of the central area (horizontal ±30°, vertical ±15°) that conforms to the probability distribution of instrument installation in underground coal mines, thereby reducing the invalid search range. It utilizes a dual-axis linked pixel-angle mapping calculation, directly converting the pixel deviation of the detected instrument ROI area into the horizontal and vertical angle adjustment of the gimbal, avoiding the blindness of traditional spiral searches. A hierarchical response mechanism is established, designing differentiated search logic for three states: complete detection, partial detection (corners / contours ≥50%), and no detection, improving robustness in complex scenarios. Simultaneously, by jointly verifying the ROI area threshold (≥30% preset value) and corner features, it effectively eliminates false detections caused by dust noise or pseudo-targets, achieving a dual improvement in search efficiency and accuracy.

[0129] The S301 output provides coarse detection results, requiring only an instrument presence determination and a rough location (ROI coordinates), without needing fine features, thus ensuring real-time performance. The S302 algorithm, through spatial priority partitioning, pixel-angle mapping, and a hierarchical response strategy, significantly improves the accuracy and efficiency of underground instrument searches in coal mines. It works in efficient synergy with the S301 coarse detection algorithm to meet both real-time and reliability requirements.

[0130] Intelligent detection algorithm for pointer instruments

[0131] Downhole pointer instrument detection faces three major challenges: ① Low-light environments cause pointer feature loss, with the IoU of traditional U-Net being only 63.2%; ② Dust interference causes scale line breakage, with morphological methods having a false detection rate of 22%; ③ Instrument panel obstruction causes center fitting error exceeding 5 pixels.

[0132] S401 scale division

[0133] First, adaptive histogram equalization and Gaussian filtering are performed on the input instrument ROI region to enhance low-light features and suppress dust noise. Then, the region is input into a U-Net segmentation network to obtain multiple pixel-level masks containing the scale markings. Next, multi-stage post-processing is used to accurately extract the effective scale markings: first, connected component analysis is performed to filter candidate regions with areas ranging from 10px to 5% of the dial area; then, ellipse fitting is used to constrain shape features, retaining connected components with eccentricity > 0.7 (slender characteristics) and rectangularity between 0.6 and 0.95 (geometric regularity). This yields all pointer scale markings. Simultaneously, the pointer target, i.e., the largest region, can be obtained, and its polar angle θ can be calculated.

[0134] S402 Starting Point Detection

[0135] Calculate the coordinates of the center points of all scale regions obtained in S401, and transform them to polar coordinates to obtain an ordered polar angle sequence of the scale:

[0136] {ϕ(1),ϕ(2),...,ϕ(m)}

[0137] ϕ is the polar coordinate corresponding to each scale mark. The polar coordinates are sorted, with the smallest polar coordinate corresponding to the starting point of the scale and the largest corresponding to the ending point.

[0138] S403 Circle Center Fitting Algorithm Formula

[0139] For all the above tick mark centers, RANSAC optimization is used to fit the least squares circle centers. First, for the candidate tick mark set {(x... i ,y i The steps for fitting the center of the circle are as follows:

[0140] Transform the standard equation of a circle into a linear form.

[0141] Where D, E, and F are all equation parameters. Center coordinates .

[0142] Establish a matrix equation for n candidate points:

[0143] x1, y1, xn, and yn are the coordinates of the selected candidate points. Substituting all these values ​​into the equations yields multiple sets of equations. Finally, the least squares method is used to solve for the center a and b of the circle.

[0144] S404 Solving pointer scale values

[0145] (1) Taking the center coordinates (a,b) as the origin, the starting pixel coordinates of the scale (u) are... s ,v s ) and endpoint pixel coordinates (u e,v e Convert to polar angle:

[0146] Starting polar angle: φs = arctan2(v s -b,u s -a)

[0147] Endpoint polar angle: φe = arctan2(v e -b,u e -a)

[0148] Where a, b are the coordinates of the center of the circle, u s ,v s The starting pixel coordinates of the scale, u e ,v e End point pixel coordinates

[0149] The angle is normalized: if the endpoint polar angle is less than the starting polar angle (φe < φs), the endpoint polar angle is increased by 360°; the pointer angle θ is synchronously processed: if θ < φs, θ is increased by 360°. This solves the problem of angle discontinuity when the range crosses 0°.

[0150] Simultaneously record the starting scale value Vs and the ending scale value Ve (the instrument scale range is known).

[0151] (2) Calculate the angle proportionality coefficient

[0152] Calculate the relative position ratio of the pointer within the range arc:

[0153] α=(θ-φs) / (φe-φs)

[0154] This coefficient represents the proportion of the pointer's deviation from the starting point to the total range angle, where φs is the starting polar angle, φe is the ending polar angle, and θ is the polar angle value calculated by dividing the pointer in S401.

[0155] Finally, the readings are calculated using linear interpolation:

[0156] Pointer V = Vs + α × (Ve - Vs)

[0157] Where V is the current pointer scale value, Vs is the starting scale value, Ve is the ending scale value, and α is the relative position ratio of the pointer in the range arc.

[0158] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been described above with reference to preferred embodiments, it is not intended to limit the present invention. Any modifications or alterations made by those skilled in the art without departing from the scope of the present invention using the above-described technical content can be considered as equivalent embodiments. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An inspection robot designed for the special environment of underground coal mines, comprising a walking track (220), characterized in that, Also includes: Walking mechanism, main frame, detection unit, power supply unit, communication unit and gimbal cover assembly (26). The walking mechanism works in conjunction with the walking track (220) to achieve movement, the detection unit acquires instrument images through the gimbal cover assembly (26), and the power supply unit and communication unit provide power and data transmission for the whole machine, respectively. The detection unit includes a gimbal cover assembly (26) and a front camera (95) mounted on the main frame. The gimbal cover assembly (26) integrates a visible light camera (602), an infrared camera (603), a fill light (601), a horizontal rotation motor (5), and a vertical rotation motor (4). The gimbal cover assembly (26) includes an annular cover (6) mounted on the main frame. A column (8) is rotatably mounted inside the annular cover (6). A slot (10) is provided at the lower end of the column (8). A turntable (9) is rotatably mounted inside the slot (10). The visible light camera (602), infrared camera (603), and fill light (601) are distributed in a triangular shape on the outer wall of the turntable (9). The vertical rotation motor (4) is fixedly mounted on the outer wall of the column (8). The output shaft of the vertical rotation motor (4) is fixedly connected to the shaft end of the turntable (9). The horizontal rotation motor (5) is fixedly mounted inside the annular cover (6). The output shaft of the horizontal rotation motor (5) is connected to the outer wall of the column (8) through two meshing transmission gears (36). The outer wall of the turntable (9) is provided with a fan-shaped groove (15). The visible light camera (602) and the infrared camera (603) are both rotatably mounted in the fan-shaped groove (15) via a rotating rod (17). A torsion spring is installed between the outer wall of the rotating rod (17) and the inner wall of the fan-shaped groove (15). A rubber membrane (18) is installed between the outer wall of the visible light camera (602) and the infrared camera (603) and the inner wall of the port of the fan-shaped groove (15). The top of the column (8) is provided with a through hole (11), a cooling fan (12) is fixedly installed in the through hole (11), a filter element (28) is installed at the upper port of the through hole (11), the upper end of the turntable (9) is provided with an air inlet groove (13) aligned with the through hole (11), a connecting pipe (14) is provided in the air inlet groove (13) and extends into the fan-shaped groove (15), and an exhaust hole (16) is provided in the fan-shaped groove (15) and extends into the outer wall of the turntable (9). Passive magnets (27) are fixedly installed on the outer walls of the visible light camera (602) and the infrared camera (603). A circular groove (19) is provided at the shaft end of the turntable (9). A disc (20) is provided in the circular groove (19). The disc (20) is fixedly connected to the inner wall of the slot (10) through a crossbar (21). Two symmetrically arranged active magnets (22) are fixedly installed on the outer wall of the disc (20). L-shaped elastic plates (37) are fixedly connected to the inner walls on both sides of the slot (10). The other end of the L-shaped elastic plate (37) is pressed against the inner wall of the turntable (9).

2. The inspection robot for the special environment of underground coal mines according to claim 1, characterized in that, The walking mechanism includes two symmetrically arranged drive wheels (166), two pairs of load-bearing wheel assemblies (183), two speed measuring wheel assemblies (198), two limit wheel assemblies (210), two pairs of side guard wheel assemblies (209), and a robot drive assembly (158) mounted on the main frame. The robot drive assembly (158) is used to drive the drive wheels (166). The load-bearing wheel assemblies (183) are rolled on the inner bottom of the grooves on both sides of the walking track (220). The speed measuring wheel assembly (198) and the limit wheel assembly (210) are both pressed against the bottom surface of the walking track (220). The two pairs of side guard wheel assemblies (209) are respectively locked on both sides of the lower end of the walking track (220).

3. The inspection robot for the special environment of underground coal mines according to claim 1, characterized in that, The main frame is provided with an explosion-proof main cavity (1), a battery cavity (25), an explosion-proof breathing and drainage valve (23) and a proximity switch (222). A photoelectric distance sensor (3) is installed on the outside of the main frame, and a control unit is installed inside the main frame. A lower protective cover assembly (2) is fixedly provided at the bottom of the main frame.

4. The inspection robot for the special environment of underground coal mines according to claim 1, characterized in that, The power supply unit includes a battery pack installed in the battery compartment (25), and an automatic charging assembly (61) is installed on the battery pack.

5. The inspection robot for the special environment of underground coal mines according to claim 1, characterized in that, The communication unit includes a wiring cavity (24) and a control unit installed on the main frame, and the wiring cavity (24) is provided with a network interface.

6. A method of using an inspection robot, characterized in that, The inspection robot described in any one of claims 1-5, designed for the special environment of underground coal mines, comprises the following steps: S1. Equipment Start-up and Initialization: The control unit initializes the walking mechanism self-test, calibrates the detection unit, and performs communication tests; S2. Path movement: According to the inspection route issued by the ground control center, the drive wheel (166) moves along the walking track (220), and the proximity switch (222) monitors obstacles; S3. Instrument positioning: After reaching the detection point, the visible light camera (602) and the infrared camera (603) are adjusted by the horizontal rotary motor (5) and the vertical rotary motor (4). The visible light camera (602) collects images and the instrument is positioned by the instrument position dynamic search algorithm. S4. Data Acquisition: High-magnification zoom is used to capture instrument images, and the scale is identified and pointer readings are calculated using the U-Net algorithm; S5. Data Reporting: Upload the encrypted test results and return to standby mode after completing all locations; S6. Anomaly Handling: Trigger a re-inspection for out-of-limit data; switch to redundancy mode to complete the task in case of equipment failure.

Images