Gate groove sill foreign matter cleaning method and system based on intelligent sensing

Abstract

The invention discloses a gate groove sill foreign matter cleaning method and system based on intelligent sensing. The method comprises the following steps that the cleaning system is fixed to a gate through a self-adaptive hook beam; the multi-source sensing module collects underwater foreign matter and environment data and constructs a sill three-dimensional environment map; the decision processing unit analyzes and processes the preprocessed data to generate a high-quality foreign matter feature map, and fuses the high-quality foreign matter feature map with a three-dimensional environment map constructed in the earlier stage to generate a sill foreign matter distribution map; the decision processing unit identifies the foreign matter type and marks the cleaning priority, and meanwhile, the decision processing unit plans an optimal cleaning path; according to the dirt type, the decision processing unit sends a cleaning device conversion instruction to the dirt cleaning device, and the dirt cleaning device executes cleaning operation according to the instruction; a high-precision torque sensor arranged in each joint of the mechanical arm collects joint force application data in real time and transmits the joint force application data to the decision processing unit; the collecting device synchronously recovers chippings generated in the cleaning process, and the communication module achieves two-way data transmission. And after the cleaning operation is completed, the multi-source sensing module scans the sill again, and the cleanliness is verified through secondary scanning. According to the cleaning method and system, efficient and automatic cleaning of the gate groove sill of the hydropower station gate can be achieved.

Description

Technical Field

[0001] This invention belongs to the technical field of gate cleaning in hydropower stations, and specifically relates to a method and system for cleaning foreign objects at the bottom sill of the gate slot based on intelligent sensing. Background Technology

[0002] During the operation of hydropower station gates, the gate slot bottom sill, as a key fitting part for gate closure, is prone to malfunctions such as incomplete gate closure and jamming due to the accumulation of foreign objects such as metal fragments, tree branches, and concrete blocks, which affect the gate's sealing performance and operational safety.

[0003] The current cleanup solutions mainly include two types: one is manual cleanup, which requires divers to dive underwater to manually remove foreign objects. However, the underwater environment is complex and affected by factors such as turbid water quality and water flow, resulting in low cleanup efficiency, high cost, and extremely high personal safety risks. The other type is mechanical cleanup devices, which are mostly fixed scrapers or high-pressure water guns. The cleanup path of these devices depends on preset programs and cannot accurately identify the location, type, and material of foreign objects, resulting in incomplete cleanup. Furthermore, rigid operation can easily damage the bottom sill structure. Summary of the Invention

[0004] This invention provides a method and system for cleaning foreign objects at the bottom sill of gate slots based on intelligent sensing, in order to solve the problem of low cleaning efficiency at the bottom sill of gate slots in hydropower stations.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: On the one hand, the method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing includes the following steps: Step 1: Secure the cleaning system to the gate using the adaptive hook beam and adjust the working position to complete the installation initialization; Step 2: After the system self-check is successful, the multi-source sensing module collects underwater foreign objects and environmental data and transmits them to the decision processing unit to construct a three-dimensional environmental map of the bottom sill. Step 3: The decision processing unit analyzes and processes the preprocessed data to generate a high-quality foreign object feature map, and merges it with the previously constructed 3D environment map to generate a foreign object distribution map of the bottom sill. Step 4: The decision processing unit calls the built-in CNN convolutional neural network model to segment and classify the preprocessed foreign object feature map, identify the type of foreign object and mark the cleaning priority. At the same time, the decision processing unit plans the optimal cleaning path. Step 5: Based on the type of contaminant, the decision-making unit sends a cleaning device switching command to the cleaning device, and the cleaning device performs the cleaning operation according to the command; Step 6: The high-precision torque sensors built into each joint of the robotic arm collect the force data applied to the joints in real time and transmit it to the decision processing unit; Step 7: During the cleaning process, the collection device simultaneously recovers the debris generated by the cleaning device; the communication module enables bidirectional data transmission between the decision processing unit and the remote monitoring terminal. Step 8: After the cleaning operation is completed, the multi-source sensing module scans the bottom sill again to generate a three-dimensional model of the cleaned bottom sill. The cleanliness is verified by a second scan, forming a closed-loop cleaning process.

[0006] Furthermore, in step two, the data collected by the multi-source sensing module includes bottom sill image data collected by an underwater high-precision optical camera and bottom sill three-dimensional coordinate data collected by a laser rangefinder. The decision processing unit fuses the two types of data to construct a 1:1 scale three-dimensional environmental map of the bottom sill.

[0007] Furthermore, in step three, the preprocessed data includes the foreign object contour image acquired by the underwater high-precision optical camera, the foreign object spectral reflectance characteristic data acquired by the multi-source spectral sensor, and the foreign object's three-dimensional coordinate information acquired in real time by the laser rangefinder.

[0008] Furthermore, in step four, when the decision processing unit plans the optimal cleaning path, it starts the SAC reinforcement learning algorithm and plans the optimal cleaning path by combining the 3D environment map, the foreign object distribution map and the current operating status of the robotic arm. When the water flow causes the foreign object to shift, the environmental data is updated in real time through the sliding window algorithm to regenerate an appropriate cleaning strategy.

[0009] Furthermore, during the cleaning process of the cleaning device in step five, the cleaning bucket pump continuously sucks up the silt from the bottom of the sill using the screw conveyor shaft, the robotic arm switches to the electromagnetic adsorber to adsorb and fix metal fragments, or switches to the high-pressure water jet nozzle to break up concrete blocks, or switches to the flexible gripper to hold organic matter.

[0010] Furthermore, in step six, the decision processing unit compares the collected force data with the preset threshold of the compressive strength limit of the bottom sill concrete in real time. When the contact pressure is detected to be close to the threshold, the robotic arm is instructed to reduce its movement speed and adjust the range of motion.

[0011] Furthermore, during the operation of the collection device in step seven, the negative pressure pump inside the collection device is activated, and the adsorption port is aligned with the debris generated during cleaning. The debris is sucked into the filter collection container through the negative pressure adsorption pipe. The clean water after solid-liquid separation is returned to the reservoir water area through the return pipe. The separated debris is stored in the container and will be centrally discharged after the operation is completed.

[0012] Furthermore, in step eight, “verifying cleanliness,” the decision processing unit registers and compares the three-dimensional models before and after cleaning, calculates the amount of foreign matter residue on the sill, and determines that the cleaning is qualified when the residue is ≤0.1kg / m²; if the residue does not meet the standard, the system automatically generates a targeted secondary cleaning strategy and re-executes the cleaning process from step three to step seven until the cleanliness of the sill meets the preset standard.

[0013] On the other hand, the intelligent sensing-based door sill foreign object removal system is used to perform the aforementioned intelligent sensing-based door sill foreign object removal method, including: The multi-source sensing module includes an underwater high-precision optical camera for acquiring underwater image information, a multispectral sensor for acquiring spectral reflectance data of dirt, and a laser rangefinder for acquiring three-dimensional coordinate information. A cleaning device for cleaning debris from the bottom sill, including a cleaning bucket pump and a multi-degree-of-freedom robotic arm; Collection device, used to suck up, filter and collect dirt generated during the cleaning process, including negative pressure adsorption pipe, filter collection container and return pipe; The decision processing unit is the core control component of the system, used to process data information; The communication module, connected to the decision processing unit, is used to transmit information to the remote monitoring terminal and receive control commands sent by the remote terminal. The adaptive hook beam is installed inside the gate slot to provide an installation platform for the cleaning device.

[0014] Furthermore, the robotic arm includes a freely replaceable modular tool head, which includes an electromagnetic adsorber for adsorbing metallic debris, a high-pressure water jet nozzle for breaking hard objects, and a flexible gripper for holding organic foreign objects.

[0015] The present invention can achieve the following beneficial effects: 1. The cleaning method described in this application is a full-link foreign object cleaning process for gate sills, including installation initialization, multi-source data acquisition, decision planning, adaptive cleaning, debris recycling, real-time monitoring, and closed-loop verification. Through the coordinated operation of adaptive hook beams, multi-source sensing modules, decision processing units, cleaning devices, collection devices, and communication modules, it replaces the traditional manual underwater cleaning method, eliminating the personal safety risks such as water flow impact and water turbidity faced by diving operations. At the same time, it avoids the defects of existing mechanical cleaning paths and indiscriminate operations, realizing full automation and standardization of foreign object cleaning. The cleaning time of a single gate is shortened compared with traditional manual methods, significantly reducing the downtime maintenance losses of hydropower stations and improving the overall efficiency and operational reliability of gate sill cleaning.

[0016] 2. The cleaning system of this application is equipped with a multi-source sensing module that integrates an underwater high-precision optical camera, a multispectral sensor and a laser rangefinder. It is paired with a decision processing unit with a built-in CNN convolutional neural network and SAC reinforcement learning algorithm. It can achieve multi-dimensional accurate acquisition of foreign object contours, material characteristics and three-dimensional coordinates, as well as accurate classification of foreign object types, priority marking and dynamic optimal path planning. This effectively improves the accuracy of cleaning operations, reduces collision damage to sensitive structures such as gate sealing surfaces and waterstops, and ensures the safety of gate operation.

[0017] 3. This application sets up a cleaning device including a cleaning bucket pump and a multi-degree-of-freedom robotic arm, equipped with three types of modular tool heads: an automatically switchable electromagnetic adsorber, a high-pressure water jet nozzle, and a flexible gripper. It is also equipped with a high-precision torque sensor for the robotic arm joints and a collection device with infrared sensing and solid-liquid separation functions. It can perform targeted cleaning of different types of foreign objects such as silt, metal fragments, concrete blocks, and branches. Through real-time force feedback and dynamic adjustment, it avoids damage to the bottom concrete structure due to excessive force. At the same time, it completes debris recovery and solid-liquid separation simultaneously. The filtered clean water is returned to the reservoir without causing water pollution. It solves the problems of existing mechanical cleaning strategies being single, easily damaging structures, and easily causing secondary pollution. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the foreign object removal method for the bottom sill of a door slot based on intelligent sensing according to the present invention; Figure 2 This is a structural framework diagram of the intelligent sensing-based foreign object removal system for door slot bottom sills according to the present invention. Detailed Implementation

[0019] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings, which illustrate embodiments of the present application. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of this application will be thorough and complete.

[0020] In one specific embodiment, the gate sill is located 15m underwater, with a length of 18m and a width of 0.8m. Due to long-term water flow impact, the sill surface retains various foreign objects. Preliminary sampling in shallow water and manual exploration revealed that the sill contains concrete blocks with a diameter of 5-15cm, tree branches with a length of 20-50cm, irregular metal fragments, and a small amount of silt covering. The water turbidity in this area is approximately 65 NTU, and the water flow velocity is 0.8m / s. Traditional manual cleaning would require downtime for more than 3 days and poses safety risks associated with diving operations. Therefore, the gate sill foreign object cleaning method and system based on intelligent sensing of this invention are used for automated cleaning operations.

[0021] like Figure 1 and Figure 2 As shown, this invention discloses a method and system for cleaning foreign objects at the bottom sill of a doorway based on intelligent sensing. The foreign object cleaning system at the bottom sill is used to perform the foreign object cleaning method at the bottom sill of the doorway. The foreign object cleaning system at the bottom sill of the doorway includes a multi-source sensing module, a decision processing unit, a cleaning device, a collection device, a communication module, and an adaptive hook beam.

[0022] The adaptive hook beam is installed within the gate slot to provide a platform for the cleaning device. Specifically, the adaptive hook beam has a symmetrical frame structure and is detachably connected to the pre-set mounting position on the gate crossbeam via bolts. A horizontal slide rail is installed at the bottom of the adaptive hook beam. The multi-source sensing module and the cleaning device are slidably connected to the adaptive hook beam via the horizontal slide rail. The multi-source sensing module and the cleaning device can adjust their operating positions according to the width of the sill to adapt to different gate slot structures.

[0023] The multi-source sensing module integrates an underwater high-precision optical camera, a multispectral sensor, and a laser rangefinder, all housed in a single waterproof casing. The casing is made of corrosion-resistant engineering plastic with a scratch-resistant coating and water flow channels to minimize the impact of water flow on the accuracy of the internal instruments. The underwater high-precision optical camera incorporates a low-light enhancement chip and an image noise reduction module. Its lens features a waterproof and dustproof coating and supports dynamic exposure adjustment, enabling it to clearly capture the outlines of foreign objects in turbid water with a turbidity greater than 50 NTU. The multispectral sensor selects a detection unit in the 400-1000 nm wavelength band and incorporates eight spectral filters, accurately acquiring spectral reflectance characteristics of foreign objects of different materials to ensure real-time material identification. The laser rangefinder uses a pulsed laser emission and reception structure with a measurement range of 0.5-20 m, quickly obtaining the three-dimensional coordinates of the bottom sill and foreign objects. A standard data interface is located on the side of the multi-source sensing module's casing, connecting to the decision processing unit via a waterproof cable to transmit the acquired images, spectra, and three-dimensional coordinate data to the decision processing unit in real time.

[0024] The cleaning device includes a cleaning bucket pump and a multi-degree-of-freedom robotic arm. The robotic arm's end is equipped with automatically switchable modular tool heads. The cleaning bucket pump includes a bucket and a auger conveyor shaft within the bucket, used to suction sludge from the bottom of the sill. The robotic arm is detachably mounted on the bucket's side wall and employs a six-joint rotating structure. Each joint is equipped with a high-precision harmonic reducer and a waterproof stepper motor, thus covering all areas of the sill for gripping debris. The robotic arm's end has a standardized tool head interface, which is detachably connected to a modular tool head via an electromagnetic clutch. The modular tool head includes an electromagnetic suction device, a high-pressure water jet nozzle, and flexible grippers. The electromagnetic suction device is suitable for cleaning metal fragments, the high-pressure water jet nozzle is suitable for breaking up hard foreign objects such as concrete blocks, and the flexible grippers have silicone fingertips to avoid damaging organic foreign objects such as branches. All three tool heads are waterproof, and their control circuitry is integrated inside the robotic arm. Communication is established with the decision processing unit through the robotic arm's control interface to receive tool head switching and motion control commands.

[0025] The collection device includes a negative pressure adsorption pipe, a filter collection container, and a return pipe. One end of the negative pressure adsorption pipe has an adjustable-angle adsorption port, with an infrared sensor module installed at the edge of the port to detect debris location and guide the adsorption direction. The other end of the pipe is sealed to the inlet of the filter collection container. The filter collection container has a slag discharge port at the bottom and a return port on the side. One end of the return pipe connects to the return port, and the other end extends into the reservoir area. The collection container has a built-in negative pressure pump, which is connected to a decision processing unit via a cable. The unit controls the pump's start / stop and adjusts the negative pressure intensity to ensure that debris generated during the cleaning process can be recovered in real time.

[0026] The decision processing unit is the core control component of the system. Its internal chassis houses a high-performance processor, a high-speed cache module, a data storage module, and multi-channel interface boards. The high-performance processor uses a multi-core industrial-grade chip, capable of simultaneously handling multiple tasks such as multi-source data preprocessing, foreign object identification and classification, and path planning. The high-speed cache module temporarily stores real-time acquired data and intermediate algorithm results. The data storage module uses a solid-state drive, capable of storing millions of underwater foreign object samples for training models and various data generated during the cleanup operation. The decision processing unit integrates a CNN convolutional neural network and SAC reinforcement learning algorithms. It establishes bidirectional communication with the multi-source sensing module, the cleaning device, and the communication module through interface boards. Its input interface supports the synchronous reception of multiple data types, including images, spectra, and 3D coordinates. Its output interface can send control commands such as tool head switching and robotic arm movement to the cleaning device. It also features data feedback reception, enabling real-time acquisition of robotic arm force data and foreign object cleaning status.

[0027] The communication module is connected to the decision processing unit and can transmit real-time data collected by the multi-source sensing module, robotic arm operating parameters, force data, cleaning status and other information to the remote monitoring terminal. At the same time, it can receive control commands sent by the remote terminal to realize two-way communication between the system and the remote terminal.

[0028] A method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing includes the following steps: Step 1: Installation and Initialization: Operators use lifting equipment to hoist the gate sill foreign object removal system to the vicinity of the target gate sill. Utilizing the symmetrical frame structure and bolted connection design of the adaptive hook beam, the cleaning system is connected and fixed to the pre-set installation position on the gate beam. Because the adaptive hook beam has a horizontal slide rail at its bottom, the installation position of the multi-source sensing module and the cleaning device on the horizontal slide rail can be adjusted according to the width of the gate sill during operation, ensuring compatibility with the current gate sill structure dimensions. The entire installation process requires no welding, drilling, or other modifications to the gate body.

[0029] Step 2, Link Connection and System Initialization: After installation, connect the underwater waterproof power supply line and communication link, start the system self-test program, and the decision processing unit sequentially tests the multi-source sensing module, cleaning device, collection device, and communication module. After confirming that each module responds normally and has no fault alarms, the multi-source sensing module is started to scan the entire area of ​​the gate sill. The laser rangefinder collects three-dimensional point cloud data of the gate sill in a pulsed transmission and reception mode, and the underwater high-precision optical camera simultaneously collects image data of the sill. The decision processing unit fuses and processes the two types of data to construct a 1:1 scale three-dimensional environmental map of the sill, completing the system initialization.

[0030] Step 3: After initialization, the underwater high-precision optical camera, through its built-in low-light enhancement chip and image noise reduction module, combined with dynamic exposure adjustment, captures the outline of foreign objects underwater. A multi-source spectral sensor accurately collects spectral reflectance characteristic data of foreign objects of different materials, ensuring real-time material identification. A laser rangefinder collects the three-dimensional coordinate information of each foreign object in real time. The collected images, spectra, and three-dimensional coordinate data are transmitted via a waterproof cable to the preprocessing module of the decision processing unit. The data undergoes median filtering for noise reduction, histogram equalization enhancement, and multispectral fusion processing to generate a high-quality foreign object feature map. Simultaneously, the laser rangefinder data is combined to extract the three-dimensional coordinates of the foreign objects, which are then fused with the previously constructed three-dimensional environmental map to generate a foreign object distribution map of the sill, providing complete data support for subsequent decision-making.

[0031] Step 4: The decision processing unit invokes the built-in CNN convolutional neural network model. This model, trained on millions of underwater foreign object samples, segments and classifies the preprocessed foreign object feature map, accurately identifying foreign object types such as metal fragments, concrete blocks, branches, and silt. Based on the size, sharpness, and material characteristics of the foreign objects, it automatically assigns a cleaning priority. For example, sharp metal fragments are marked as "Level 1 Emergency Treatment," concrete blocks as "Level 2 Treatment," and branches and silt as "Level 3 Treatment." Simultaneously, the decision processing unit activates the SAC reinforcement learning algorithm, combining the 3D environment map, foreign object distribution map, and the current operating status of the robotic arm to plan the optimal cleaning path within 100ms, avoiding sensitive structures such as gate sealing surfaces and waterstops. When water flow causes the foreign object's position to shift, the sliding window algorithm updates the environmental data in real time, regenerating an adapted cleaning strategy within 50ms to ensure the cleaning path dynamically adapts to environmental changes.

[0032] Step 5: The cleaning device initiates a coordinated cleaning operation based on the instructions from the decision-making unit: the cleaning bucket pump starts synchronously, and the bucket adheres to the surface of the sill, continuously sucking up the silt from the bottom of the sill using the built-in spiral conveyor shaft to achieve efficient silt removal; the six-joint rotating robotic arm installed on the side wall of the bucket moves according to the planned path, automatically switching modular tool heads via an electromagnetic clutch. For metal fragments, an electromagnetic adsorber is used for adsorption and fixation; for concrete blocks, a high-pressure water jet nozzle is used and the spray angle is adjusted for crushing; for organic materials such as branches, flexible grippers with silicone fingertips are used for stable clamping. Then, the various cleaned foreign objects are precisely transferred to the negative pressure adsorption port of the collection device, completing the adaptive cleaning of the foreign objects.

[0033] Step Six: High-precision torque sensors built into each joint of the robotic arm collect joint force data in real time and transmit it to the decision processing unit. The decision processing unit compares the collected force data with the preset threshold of the compressive strength limit of the bottom sill concrete in real time. When the contact pressure is detected to be close to the threshold, the robotic arm is immediately instructed to reduce its movement speed and adjust the range of motion to avoid damage to the bottom sill structure due to excessive force.

[0034] Step Seven: During the cleaning process, the collection device and the cleaning device work in tandem. The negative pressure pump of the collection device starts, and through an adjustable-angle suction port equipped with an infrared sensor module, it accurately detects and aligns with the debris generated during cleaning. The debris is then sucked into the filter collection container through the negative pressure suction pipe. After solid-liquid separation, the clean water flows back to the reservoir through a return pipe. The separated debris remains in the container and is centrally processed through the bottom slag discharge port after the operation is completed, with no wastewater discharge. Simultaneously, the communication module achieves bidirectional data transmission between the decision processing unit and the remote monitoring terminal with an ultra-low latency of ≤20ms. It synchronizes real-time images collected by the multi-source sensing module, robotic arm operating parameters, force data, and cleaning status information to the terminal, allowing operators to monitor the entire process visually. When the system detects an unknown foreign object, it automatically triggers an alarm, which is transmitted to the decision processing unit via the communication module, enabling manual emergency intervention.

[0035] Step 8: After the initial cleaning operation is completed, the decision processing unit instructs the multi-source sensing module to scan the entire sill area again, repeating the data acquisition and preprocessing process in Step 3 to generate a cleaned 3D model of the sill. The decision processing unit registers and compares the 3D models before and after cleaning, and combines CNN secondary image scanning and laser ranging data to accurately calculate the amount of foreign matter remaining on the sill. When the amount of remaining matter is ≤0.1kg / m², the cleaning is deemed qualified; if the amount of remaining matter does not meet the standard, the system automatically generates a targeted secondary cleaning strategy and re-executes the cleaning process from Step 3 to Step 7 until the cleanliness of the sill meets the preset standard.

[0036] After the cleaning is completed and the target standard is met, the operators stop the operation of each module of the system through the command in the central control room, remove the fixing bolts between the adaptive hook beam and the gate, and use the lifting equipment to lift the cleaning system away from the work area, thus completing the foreign object cleaning operation at the bottom of the gate slot.

[0037] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing, characterized in that, Includes the following steps: Step 1: Secure the cleaning system to the gate using the adaptive hook beam and adjust the working position to complete the installation initialization; Step 2: After the system self-check is successful, the multi-source sensing module collects underwater foreign objects and environmental data and transmits them to the decision processing unit to construct a three-dimensional environmental map of the bottom sill. Step 3: The decision processing unit analyzes and processes the preprocessed data to generate a high-quality foreign object feature map, and merges it with the previously constructed 3D environment map to generate a foreign object distribution map of the bottom sill. Step 4: The decision processing unit calls the built-in CNN convolutional neural network model to segment and classify the preprocessed foreign object feature map, identify the type of foreign object and mark the cleaning priority. At the same time, the decision processing unit plans the optimal cleaning path. Step 5: Based on the type of contaminant, the decision-making unit sends a cleaning device switching command to the cleaning device, and the cleaning device performs the cleaning operation according to the command; Step 6: The high-precision torque sensors built into each joint of the robotic arm collect the force data applied to the joints in real time and transmit it to the decision processing unit; Step 7: During the cleaning process, the collection device simultaneously recovers the debris generated by the cleaning device; the communication module enables bidirectional data transmission between the decision processing unit and the remote monitoring terminal. Step 8: After the cleaning operation is completed, the multi-source sensing module scans the bottom sill again to generate a three-dimensional model of the cleaned bottom sill. The cleanliness is verified by a second scan, forming a closed-loop cleaning process.

2. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step two, the data collected by the multi-source sensing module includes bottom sill image data collected by an underwater high-precision optical camera and bottom sill three-dimensional coordinate data collected by a laser rangefinder. The decision processing unit fuses the two types of data to construct a 1:1 scale three-dimensional environmental map of the bottom sill.

3. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step three, the preprocessed data includes the outline image of the foreign object acquired by the underwater high-precision optical camera, the spectral reflection characteristic data of the foreign object acquired by the multi-source spectral sensor, and the three-dimensional coordinate information of the foreign object acquired in real time by the laser rangefinder.

4. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step four, when the decision processing unit plans the optimal cleaning path, it starts the SAC reinforcement learning algorithm and plans the optimal cleaning path by combining the 3D environment map, the foreign object distribution map and the current operating status of the robotic arm. When the water flow causes the foreign object to shift, the environmental data is updated in real time through the sliding window algorithm to regenerate an appropriate cleaning strategy.

5. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step five, when the cleaning device is cleaning up the dirt, the cleaning bucket pump uses the screw conveyor shaft to continuously suck up the silt at the bottom of the sill. The robotic arm switches to the electromagnetic adsorber to adsorb and fix metal fragments, or switches to the high-pressure water jet nozzle to break up concrete blocks, or switches to the flexible gripper to hold organic matter.

6. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step six, the decision processing unit compares the collected force data with the preset threshold of the compressive strength limit of the bottom sill concrete in real time. When the contact pressure is detected to be close to the threshold, the robotic arm is instructed to reduce its movement speed and adjust the range of motion.

7. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: When the collection device is running in step seven, the negative pressure pump inside the collection device starts, aligns the adsorption port with the debris generated during cleaning, and sucks the debris into the filter collection container through the negative pressure adsorption pipe. After solid-liquid separation, the clean water flows back to the reservoir water area through the return pipe. The separated debris is left in the container and will be centrally discharged after the operation is completed.

8. The method for cleaning foreign objects from the bottom sill of a doorway based on intelligent sensing according to claim 1, characterized in that: In step eight, "verifying cleanliness", the decision processing unit will register and compare the three-dimensional models before and after cleaning, calculate the amount of foreign matter residue on the bottom sill, and determine that the cleaning is qualified when the residue is ≤0.1kg / m². If the residue does not meet the standard, the system will automatically generate a targeted secondary cleaning strategy and re-execute the cleaning process from step three to step seven until the cleanliness of the bottom sill meets the preset standard.

9. A door sill foreign object removal system based on intelligent sensing, used to perform the door sill foreign object removal method based on intelligent sensing as described in any one of claims 1-8, characterized in that, include: The multi-source sensing module includes an underwater high-precision optical camera for acquiring underwater image information, a multispectral sensor for acquiring spectral reflectance data of dirt, and a laser rangefinder for acquiring three-dimensional coordinate information. A cleaning device for cleaning debris from the bottom sill, including a cleaning bucket pump and a multi-degree-of-freedom robotic arm; Collection device, used to suck up, filter and collect dirt generated during the cleaning process, including negative pressure adsorption pipe, filter collection container and return pipe; The decision processing unit is the core control component of the system, used to process data information; The communication module, connected to the decision processing unit, is used to transmit information to the remote monitoring terminal and receive control commands sent by the remote terminal. The adaptive hook beam is installed inside the gate slot to provide an installation platform for the cleaning device.

10. The intelligent sensing-based foreign object removal system for door slot bottom sills according to claim 9, characterized in that: The robotic arm includes a freely replaceable modular tool head, which includes an electromagnetic adsorber for adsorbing metallic debris, a high-pressure water jet nozzle for breaking hard objects, and a flexible gripper for holding organic foreign objects.

Images