Drainage pipe network detection system based on spider bionic robot

Abstract

The invention relates to a drainage pipe network detection system based on a spider bionic robot, and the system comprises the spider bionic robot which continuously carries out the detection in a drainage pipeline, and collects the internal environment parameters and images of the pipeline to obtain detection data; the inspection well base is arranged on the inner wall of the drainage pipeline and provides charging and data transmission support for the spider bionic robot; the processor is connected with the inspection well base, receives and analyzes the detection data from the spider bionic robot, and generates a drainage pipeline health assessment report at the same time; the detection data comprises the temperature and humidity of the drainage pipeline, an internal image of the drainage pipeline, the thickness and point cloud data of the inner wall of the drainage pipeline, and gas components and water flow in the drainage pipeline. By the adoption of the drainage pipe network detection system based on the spider bionic robot, efficient, intelligent and comprehensive drainage pipe network detection can be achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection of municipal drainage pipe networks, and in particular to a drainage pipe network inspection system based on a spider-inspired robot. Background Technology

[0002] As a crucial component of urban infrastructure, urban drainage networks play a vital role in collecting and transporting urban sewage and rainwater. Their operational status directly impacts the city's water environment quality, flood control and drainage capacity, and residents' quality of life. With rapid urban development, drainage networks are expanding in scale and becoming increasingly complex. However, these networks, operating in harsh underground environments, face numerous challenges, such as sewage corrosion, groundwater infiltration, geological subsidence, and external pressure. These factors lead to corrosion, deformation, siltation, cracks, collapses, and leaks, severely affecting the normal operation of the drainage network.

[0003] Traditional drainage pipe networks are mainly inspected manually or via closed-circuit television (CCTV). Manual inspection relies on manual entry into manholes, which is not only labor-intensive and inefficient but also poses safety risks. Furthermore, limitations in human vision and physical strength make it difficult to detect hidden defects, resulting in lower accuracy and reliability of inspection results. While CCTV inspection can acquire images of the inside of the pipes, it requires prior dredging of the pipes, making the inspection process slow. The subsequent analysis of large amounts of video data also consumes significant manpower and time, making it impossible to assess the health status of the pipes in a timely and accurate manner.

[0004] Therefore, with the development of robotics technology in recent years, some inspection robots have been gradually applied to drainage pipe network inspection to improve inspection efficiency and intelligent judgment. However, existing robot-based drainage pipe network inspection systems are still not comprehensive enough in data analysis, and their intelligent assessment accuracy is low.

[0005] Therefore, developing an efficient, intelligent, and comprehensive drainage pipe network detection system is of great practical significance for timely detection of drainage pipe network defects and ensuring the safe and stable operation of the drainage pipe network. Summary of the Invention

[0006] Based on this, the purpose of this invention is to provide a drainage pipe network detection system based on a spider-inspired robot. By integrating autonomous navigation and multimodal perception into the spider-inspired robot, and combining intelligent charging and cloud collaboration functions, it can achieve efficient, intelligent and comprehensive drainage pipe network detection.

[0007] A drainage pipe network inspection system based on a spider-inspired robot includes:

[0008] A spider-inspired robot continuously inspects inside drainage pipes, collecting environmental parameters and images to obtain inspection data.

[0009] The inspection well base, installed on the inner wall of the drainage pipe, provides charging and data transmission support for the spider-like robot.

[0010] The processor, connected to the manhole base, receives and analyzes detection data from the spider-like robot, and generates a health assessment report for the drainage pipes.

[0011] The detection data includes the temperature and humidity of the drainage pipe, internal images of the drainage pipe, the thickness and point cloud data of the inner wall of the drainage pipe, the gas composition and water flow rate inside the drainage pipe.

[0012] The processor includes:

[0013] The identification module is used to identify features in the detection data to obtain various disease characteristics;

[0014] The analysis module is used to combine multiple disease characteristics with point cloud data to construct a three-dimensional model of the disease and accurately calculate the parameters of the corresponding disease.

[0015] The assessment module is used to calculate the final health index of the drainage pipe based on the characteristics of the disease, disease parameters and detection data, and to assess the health status of the drainage pipe and generate a health assessment report.

[0016] A three-level evaluation index system was constructed using the analytic hierarchy process (AHP), including structural safety, functional integrity, and environmental risk. The structural safety index was calculated by multiplying the defect type weight matrix with geometric parameters. The functional integrity index was based on flow velocity and flow rate data, using Bernoulli's equation to calculate the pipeline's transport efficiency decay. The environmental risk index was calculated by predicting the remaining wall thickness of the metal pipeline using a corrosion rate model, and combining this with gas concentration data to derive the final health index calculation formula.

[0017]

[0018] Where w i Let v be the weight of the i-th indicator. i v is the measured value. i,max The threshold is the upper limit; the structural safety index accounts for 50% of the weight, and the geometric parameters of the disease are calculated from the three-dimensional coordinates of the point cloud data; the system implements a four-level response based on the HI value, and triggers a red warning when HI < 30, and automatically generates a drainage pipeline health assessment report based on the data analysis results.

[0019] Compared with existing technologies, the drainage pipe network inspection system based on spider bionic robots of the present invention can acquire comprehensive inspection data on drainage pipes, perform in-depth analysis and processing of the corresponding inspection data, generate health assessment reports, predict potential risks of pipes, provide a scientific basis for the operation and maintenance management of drainage pipe networks, greatly reduce operation and maintenance costs, and improve the reliability and safety of urban drainage systems.

[0020] Furthermore, the data processing system of the analysis module follows a three-level flow architecture from multi-source heterogeneous data to edge layer feature extraction, and then to cloud-based deep analysis. The multi-source heterogeneous data adopts parallel processing logic, and the original image, point cloud, sonar and environmental parameters are processed by independent algorithms to generate joint feature vectors. The processing flow of each branch has no time dependence and synchronous scheduling is achieved through hardware interrupt mechanism.

[0021] The edge layer feature extraction uses the robot's built-in embedded computing unit to perform preliminary processing on the detection data while adopting a parallel processing architecture. Each branch executes independently to generate intermediate feature vectors.

[0022] The image data undergoes nonlocal mean filtering to remove high-contour features; the lidar point cloud data is processed using the RANSAC algorithm to remove outliers and downsampled based on the voxel grid method; the sonar echo signal is decomposed using wavelet packet decomposition to extract energy feature vectors; the environmental parameters are constructed by fusing multi-source data through extended Kalman filtering to create a pipeline environmental state vector; the prediction formula for the EKF state is:

[0023]

[0024] in represents the predicted value of state x at time k-1; f represents the state function; This represents the state estimate at time k-1; u k This represents the control input or external input at time k;

[0025] The cloud-based deep analysis achieves the classification and calculation of disease parameters by sequentially processing the detection data through a three-level process: data governance, feature engineering, and intelligent decision-making. This includes:

[0026] The data governance layer uses a distributed stream computing framework to achieve real-time cleaning, detects data anomalies through a sliding window algorithm, fills in missing values ​​using interpolation, and constructs a standardized dataset.

[0027] The feature engineering layer uses graph neural networks to model the pipeline topology and combines an attention mechanism to extract spatial correlation features of diseases. Point cloud data is hierarchically organized through an octree structure, and sonar data is reduced in dimension through a convolutional autoencoder.

[0028] The intelligent decision-making layer simultaneously classifies defects based on a multi-task learning model, categorizing defects into 14 types, and uses a Poisson surface reconstruction algorithm to generate a high-precision mesh model of the pipe inner wall.

[0029] Furthermore, the graph neural network adopts a three-layer architecture, and its message passing mechanism is defined as m j →i=σ(W·X j +b), where m j →i represents the message passed from node j to node i, containing the features of j and its association with i; σ(·) represents the activation function; W and b represent the learnable weights and biases; X j The original characteristics of node j are represented, including data such as the diameter, material, and historical defects of the pipeline segment. The feature propagation between nodes is achieved through the adjacency matrix.

[0030] Furthermore, the spider-like bionic robot includes a shell, multiple walking legs, motion sensors, integrated detection components, and a control chip; the multiple walking legs are symmetrically arranged on both sides of the shell, driving the shell to move; the motion sensors are installed on the walking legs to detect the movement state and force conditions of the walking legs in real time; the integrated detection component is located at the front end of the shell to continuously detect the external environment and obtain detection data; the chip is located inside the shell to receive data from the motion sensors and control the movement of the multiple walking legs.

[0031] Furthermore, the integrated detection components include a temperature sensor, a humidity sensor, a high-definition camera, a sonar, a lidar, an environmental parameter sensor, and a flow meter. The temperature sensor uses a high-precision thermistor to measure temperature changes within the drainage pipe, providing data support for analyzing the physical and chemical processes inside the pipe. The humidity sensor monitors the humidity within the drainage pipe in real time to determine if leaks or other problems exist. The high-definition camera uses a global shutter CMOS sensor and is equipped with an infrared LED matrix illumination system. The LED light source has a wavelength of 850nm, and the illumination uniformity is ≥95%. The focal length of the high-definition camera can automatically switch within the range of 5-50mm to achieve optical imaging of 1mm-level defects on the inner wall of the pipe in a low-light environment of 0.1 lux. The sonar system... The system employs a sound wave with a frequency range of 200-500kHz to penetrate 30cm of silt and detect changes in the wall thickness of drainage pipes. Time-frequency domain features are extracted from the echo signal using a short-time Fourier transform. The lidar utilizes a multi-line rotating structure equipped with 64 laser beams and a scanning frequency of 10-20Hz. It constructs point cloud data of the drainage pipe's inner wall using a time-of-flight method, generating a 3D coordinate matrix of 100,000 points per second. The environmental parameter sensor integrates a high-precision temperature and humidity composite probe and an electrochemical gas sensor array, with detection limits of 0.1ppm for hydrogen sulfide and 1ppm for methane. The flow meter tracks the trajectory of floating objects using the Lucas-Kanad optical flow algorithm and, combined with the drainage pipe's cross-sectional area parameters, calculates the real-time flow rate using a continuity equation.

[0032] Furthermore, the control chip includes an action module, a power detection module, and a navigation module. The action module controls the movement of multiple walking legs to enable the spider-like bionic robot to move within the drainage pipe. The power detection module continuously monitors the power level of the power supply component and sends an automatic charging command to the navigation module when the power level of the power supply component is below 20%. The navigation module plans the shortest path to the nearest manhole base and sends it to the action module. The action module controls the spider-like bionic robot to travel to the manhole base for charging based on the shortest path.

[0033] Furthermore, the navigation module combines inertial navigation, visual navigation, and geomagnetic navigation technologies. Inertial navigation technology measures the angular velocity and acceleration of the spider-like robot using angle and accelerometer sensors on its walking legs to calculate the robot's posture. Visual navigation technology uses a high-definition camera to acquire image information of the surrounding environment and determines the position and orientation of the spider-like robot through image recognition and matching algorithms. Geomagnetic navigation technology determines the orientation of the spider-like robot based on changes in the Earth's magnetic field.

[0034] Furthermore, the processor also includes a data storage module for classifying and storing the large amount of received detection data through a combination of a distributed file system and a relational database.

[0035] Furthermore, the processor also includes an early warning module for real-time monitoring of health assessment reports and sending early warning information when serious diseases or abnormalities are detected.

[0036] Furthermore, it also includes a display screen that is connected to the processor to receive and present health assessment reports and warning information from the processor to the user in real time.

[0037] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the spatial structure of the drainage pipe network detection system based on a spider-inspired robot according to the present invention, located in a drainage pipe.

[0039] Figure 2 This is a schematic diagram of the spider-inspired robot structure in the drainage pipe network detection system based on the spider-inspired robot of the present invention.

[0040] Figure 3 This is a schematic diagram of the walking legs in the spider-inspired robot of the present invention.

[0041] Figure 4 This is a schematic diagram of the manhole base in the drainage pipe network inspection system based on a spider-inspired robot of the present invention. Detailed Implementation

[0042] The inventors carefully analyzed existing robot-based drainage network inspection systems and found that their shortcomings in data processing, analysis, and evaluation stemmed from difficulties in data acquisition, resulting in insufficient inspection data and an inability to fully utilize the data for in-depth mining and analysis, hindering accurate prediction and effective management of pipeline health. Therefore, this invention attempts to use a spider-inspired robot as a data acquisition vehicle and conducts further analysis and evaluation based on the data acquired by the spider-inspired robot to achieve comprehensive and accurate inspection of drainage networks. Simultaneously, the invention designs a manhole base for the spider-inspired robot and improves its automatic charging and data transmission capabilities, achieving high-efficiency automation and intelligence.

[0043] Based on this, please refer to Figure 1 The present invention provides a drainage pipe network inspection system based on a spider bionic robot, including a spider bionic robot 1, an inspection well base 2, a processor (not shown) and a display screen (not shown).

[0044] The spider-like bionic robot 1 continuously performs inspections inside the drainage pipe, collecting environmental parameters and images to obtain inspection data. Please refer to [link / reference needed]. Figure 2 The spider-inspired robot 1 includes a shell 11, multiple walking legs 12, motion sensors 13 and integrated detection components 14, a charging interface 15, a power supply assembly (not shown), and a control chip (not shown). The multiple walking legs 12 are symmetrically arranged on the left and right sides of the shell 11. The walking legs 12 are flexibly connected to the shell 11 via hinges or other means. The connection point between the walking legs 12 and the shell 11 can serve as a pivot point, enabling localized vertical and horizontal rotation as well as forward and backward rotation.

[0045] Please see Figure 3 The walking leg 12 includes a first leg arm 121, a second leg arm 122, a fixed rotation axis 123, and a movable rotation axis 124. The front end of the first leg arm 121 is connected to the machine housing 11 via the fixed rotation axis 123, allowing the first leg arm 121 to rotate up and down and forward and backward relative to the machine housing 11 with the fixed rotation axis 123 as a fixed point. The middle part of the second leg arm 122 is connected to the end of the first leg arm 121 via the movable rotation axis 124, allowing the second leg arm 122 to rotate up and down relative to the first leg arm 121 with the movable rotation axis 124 as a fixed point, enhancing the balance and flexibility of the spider bionic robot 1. The front end of the second leg arm 122 protrudes from the first leg arm 121, providing the ability to automatically flip back to the correct position when the machine housing 11 is inverted. The end of the second leg arm 122 contacts the ground, providing support and friction for the spider bionic robot 1 during movement. The number of walking legs 12 is preferably six. Compared with three walking legs 12, six walking legs 12 are symmetrically distributed on the left and right sides of the machine housing 11 in an even number, which will not affect the shooting field of view or charging position at the front and rear ends of the machine housing 11. Compared with two pairs of walking legs 12, the extra pair of walking legs 12 can greatly improve the turning ability and climbing ability of the spider bionic robot 1.

[0046] The design of the six walking legs 12 mimics the structure and movement of a spider's legs, employing a multi-joint, multi-degree-of-freedom design. Each walking leg 12 is made of high-strength, corrosion-resistant materials, such as titanium alloy or high-strength engineering plastics, thereby ensuring that the walking legs 12 have sufficient strength and durability in harsh drainage pipe environments.

[0047] The motion sensor 13 is located at the movable rotation axis 124 of the walking leg 12 and includes a pressure sensor, an angle sensor and an acceleration sensor. It provides real-time feedback to the control chip on the motion state and force conditions of the walking leg 12, thereby accurately controlling the motion of the walking leg 12. This enables the spider bionic robot 1 to flexibly crawl in drainage pipes of different diameters, shapes and complexities, adapting to various complex environments, such as narrow pipes, bends, slope changes and areas with obstacles.

[0048] The integrated detection component 14 is located at the front end of the machine housing 11 and continuously monitors the external environment to obtain detection data. The integrated detection component 14 includes a temperature sensor, a humidity sensor, a high-definition camera, a sonar, a lidar, an environmental parameter sensor, and a flow meter. Specifically, the temperature sensor uses a high-precision thermistor with an accuracy of ±0.1℃, accurately measuring temperature changes within the drainage pipe and providing data support for analyzing the physical and chemical processes inside the pipe. The humidity sensor is based on capacitive humidity sensing, with a measurement error of less than ±2%RH, real-time monitoring of humidity within the drainage pipe to determine if leaks or other problems exist. The high-definition camera uses a global shutter CMOS sensor with a resolution ≥4096×3072 pixels, equipped with an infrared LED matrix illumination system. The LED light source wavelength is 850nm, with illumination uniformity ≥95%. Combined with an automatic zoom lens, its focal length can automatically switch within the range of 5-50mm, enabling optical imaging of 1mm-level defects on the inner wall of the pipe even in low-light environments of 0.1 lux. The sonar system, based on phased array technology, emits sound waves at a frequency range of 200-500kHz, penetrating 30cm of silt to detect changes in the wall thickness of the drainage pipe. The echo signal is extracted for time-frequency domain features via Short-Time Fourier Transform (STFT). The lidar employs a multi-line rotating structure, equipped with a 64-line laser beam, a scanning frequency of 10-20Hz, and a ranging accuracy of ±2mm. It constructs point cloud data of the drainage pipe's inner wall using Time-of-Flight (ToF) technology, generating a 3D coordinate matrix of 100,000 points per second. The environmental parameter sensor integrates a high-precision temperature and humidity composite probe and an electrochemical gas sensor array. The detection limits for hydrogen sulfide and methane are 0.1ppm and 1ppm, respectively, with a response time <10s. Multi-sensor data is synchronously acquired via an I2C bus with a synchronization error <10ms. The flow meter uses a dual-camera stereo vision solution, tracking the trajectory of floating objects based on the Lucas-Kanad optical flow algorithm. Combined with the drainage pipe cross-sectional area parameters, a continuity equation is used to calculate the real-time flow rate. This enables the detection of material parameters, structural parameters, and surface parameters of the inner wall of drainage pipes, as well as the detection of temperature and humidity, ambient gas, and ambient water flow inside the drainage pipes, resulting in comprehensive testing data about the drainage network.

[0049] Preferably, the integrated detection component 14 is set on the fixed rotation axis 123 of the front walking foot 12. It can not only continuously detect relevant data of the environment in front, but also change the viewing angle in real time as it moves, so as to obtain the field of view on both sides in addition to the front, and can detect environmental data more comprehensively.

[0050] In one embodiment, the integrated detection component 14 employs a dynamic adaptive mechanism for detection. Under normal operating conditions, it uses a low-frequency mode, where the image frame rate captured by the high-definition camera is 5fps and the scanning frequency of the lidar is 5Hz, in order to reduce the power consumption of the integrated detection component 14. However, when a defect feature is detected, such as a crack width in a drainage pipe > 3mm, a high-frequency mode is triggered. In this mode, the image frame rate captured by the high-definition camera is increased to 30fps and the scanning frequency of the lidar is increased to 20Hz, thereby accurately and comprehensively detecting the relevant parameters of the defect feature.

[0051] The spider-inspired robot 1 is charged by connecting to an external power source through the charging interface 15. Preferably, the charging interface 15 is located at the rear end of the machine housing 11, which can avoid affecting the normal operation of the walking foot 12 without affecting the detection of the front environment by the integrated detection component.

[0052] The power supply component is located inside the machine housing 11 and is connected to the charging interface 15. It uses a rechargeable battery as its energy source, such as a high-performance lithium battery, which has advantages such as high energy density, high charging and discharging efficiency, and long cycle life.

[0053] The control chip is located inside the machine housing 11 and is connected to the power supply component, which supplies power to the robot. The control chip includes a movement module, a power detection module, and a navigation module. The movement module controls the movement of multiple walking legs 12 to enable the spider-like robot 1 to move within the drainage pipe. The power detection module continuously monitors the power supply component's charge level and sends an automatic charging command to the navigation module when the power supply component's charge level drops below 20%. The navigation module plans the shortest path to the nearest manhole base 2 and sends it to the movement module, which then controls the spider-like robot 1 to travel to the manhole base 2 for charging based on the shortest path.

[0054] Specifically, the navigation module integrates inertial navigation, visual navigation, and geomagnetic navigation technologies. Inertial navigation measures the angular velocity and acceleration of the spider-like robot 1 using angle and accelerometer sensors on its walking legs 12 to calculate its posture. Visual navigation uses a high-definition camera to acquire image information of the surrounding environment and determines the position and orientation of the spider-like robot 1 through image recognition and matching algorithms. Geomagnetic navigation determines the orientation of the spider-like robot 1 based on changes in the Earth's magnetic field. The navigation module uses these three navigation technologies to complement each other, improving navigation accuracy and ensuring that the spider-like robot 1 can accurately locate the nearest inspection well base 2 for charging.

[0055] The inspection well base 2 is located at the inspection port on the inner wall of the drainage pipe and at the upper part of the drainage pipe, thereby reducing the possibility of it being wetted by water flow and providing charging and data transmission support for the spider-like robot. The inspection well base 2 includes a power supply component 21 and a data transmission component 22. The bottom of the power supply component 21 is fixed to the upper wall of the drainage pipe, and a power supply interface 211 is provided at one end of the power supply component 21. The power supply interface 211 adopts an automatic docking design, combining electromagnetic induction, optical positioning, and mechanical locking design to achieve precise matching and connection with the charging interface 15 of the spider-like robot 1, thereby providing charging to the spider-like robot 1. Specifically, when the spider-like robot 1 gradually approaches the inspection well base 2, the power supply component 21 uses electromagnetic induction signals to guide the charging interface 15 of the spider-like robot 1 to initially align with the power supply interface 211 of the power supply component 21; at the same time, optical positioning is used for precise alignment; after docking, the spider-like robot 1 is fixed by mechanical locking design to ensure the stability of the charging process and achieve rapid charging.

[0056] The data transmission component 22 is fixedly connected to the power supply component 21, suspended inside the drainage pipe, and connected to an external data line or power supply line. It can simultaneously employ wired or wireless transmission methods, such as fiber optic, cable, 4G / 5G, Wi-Fi, etc. The data transmission component 22 includes a preprocessing module, a detection module, a transmission module, and a buffer module. The preprocessing module compresses and encrypts the detection data to reduce the amount of data transmitted, improve transmission efficiency, and ensure the security of the detection data. The detection module automatically detects the network status. When the network signal is strong, the detection module controls the transmission module to transmit signals wirelessly, achieving real-time transmission of the detection data. When the network signal is weak, the detection module controls the transmission module to transmit signals wiredly, ensuring that the detection data is not lost. When the network signal is insufficient for wired transmission, the detection module temporarily stores the detection data from the spider-like robot 1 in the buffer module.

[0057] The processor is connected to the inspection well base 2, receives and analyzes detection data from the spider-like robot, and generates a drainage pipe health assessment report. Specifically, the processor includes a data storage module, an identification module, an analysis module, an assessment module, and an early warning module.

[0058] The data storage module is used to classify and store the large amount of received detection data through a combination of a distributed file system and a relational database. Specifically, the distributed file system features high scalability, high reliability, and high performance, such as the Ceph Distributed Storage System, which can store massive amounts of images, videos, and files. The relational database is used to store structured data, such as MyStructured Query Language, which can store basic information about drainage pipes, processor detection results, and evaluation reports, facilitating data retrieval and management.

[0059] The identification module is used to perform feature recognition on the detection data in the data storage module to obtain various disease characteristics. Specifically, the identification module is based on machine vision technology using deep learning algorithms. It trains the deep learning model with a large number of labeled samples, enabling the model to accurately identify various disease characteristics such as corrosion, deformation, siltation, cracks, collapse, and leakage in drainage pipes.

[0060] The analysis module is used to construct a 3D model of the disease by combining multiple disease features with point cloud data from lidar, and to accurately calculate the parameters of the corresponding disease, including length, width, depth, area, and volume. Specifically, the data processing system of the analysis module follows a three-level flow architecture of "multi-source heterogeneous data → edge layer feature extraction → cloud-based deep analysis". The multi-source heterogeneous data adopts parallel processing logic, and the original image, point cloud, sonar, and environmental parameters are processed by independent algorithms to generate a joint feature vector. The processing flow of each branch has no time dependence and synchronous scheduling is achieved through a hardware interrupt mechanism.

[0061] The edge layer feature extraction is performed by the robot's built-in embedded computing unit. The detection data is initially processed using a parallel processing architecture, and each branch is executed independently to generate an intermediate feature vector.

[0062] Specifically, the image data is processed using non-local mean filtering (NLM) to remove high-contour features. Drainage pipe images often contain noise and motion blur. NLM achieves noise reduction through pixel block similarity weighting, improving computational speed by 1.8 times compared to conventional Gaussian filtering and bilateral filtering, while significantly preserving subtle features such as cracks. Alternatively, the BM3D algorithm can be used to remove high-contour features, improving the peak signal-to-noise ratio by 0.5 dB compared to the NLM method, but increasing computational complexity by 300%. Limited by the embedded platform, this approach sacrifices instrument dependence to improve the signal-to-noise ratio.

[0063] The lidar point cloud data is processed using the RANSAC algorithm to remove outliers and downsampled based on the voxel grid method. Lidar point clouds are prone to outliers due to reflections from water accumulation in pipes. The RANSAC method removes mismatched points through random sampling consistency, achieving a noise reduction rate of up to 98.7%, which is more accurate for fitting pipe surfaces compared to conventional statistical outlier filtering. Furthermore, the PROSAC algorithm can be used to further remove outliers, improving the clarity of outlier distribution patterns.

[0064] The sonar echo signal is subjected to wavelet packet decomposition to extract energy feature vectors. The sonar echo contains multi-band signals reflected from sediment. WPD achieves adaptive frequency band division through binary tree decomposition, which is more suitable for non-stationary signal analysis compared to conventional Fourier transform and short-time Fourier transform.

[0065] The environmental parameters are fused from multi-source data using an Extended Kalman Filter (EKF) to construct a pipeline environmental state vector. Since multi-sensor data for environmental parameter detection exhibits nonlinear coupling—for example, temperature can cause drift in gas sensors—the EKF employs Taylor expansion for linearization, achieving a higher balance between accuracy and efficiency compared to conventional Kalman and particle filters. The EKF state prediction formula is as follows:

[0066]

[0067] in represents the predicted value of state x at time k-1; f represents the state function; This represents the state estimate at time k-1; u k This represents the control input or external input at time k. Based on the "historical best state" and the "current input," the mathematical model predicts the "current state" of the system, which is a priori estimate of the dynamic changes of the system.

[0068] The cloud-based deep analysis achieves the classification and calculation of disease parameters by sequentially processing the detection data through a three-level process: a data governance layer, a feature engineering layer, and an intelligent decision-making layer. Specifically, the data governance layer uses a distributed stream computing framework for real-time cleaning, detects data anomalies using a sliding window algorithm, fills in missing values ​​using interpolation, and constructs a standardized dataset.

[0069] The feature engineering layer uses graph neural networks (GNN) to model the pipeline topology and combines an attention mechanism to extract spatial correlation features of the disease. The lidar point cloud is hierarchically organized through an octree structure, and the sonar data is reduced in dimension through a convolutional autoencoder (CAE).

[0070] The intelligent decision-making layer simultaneously classifies defects based on a multi-task learning model (MTL), categorizing them into 14 types with an average classification accuracy of 94.7%. The 3D reconstruction module uses the Poisson Surface Reconstruction algorithm to generate a high-precision mesh model of the pipe's inner wall. The graph neural network employs a three-layer architecture, with its message passing mechanism defined as m... j →i=σ(W·X j +b), where m j →i represents the message passed from node j to node i, containing the features of j and its association with i; σ(·) represents the activation function; W and b represent the learnable weights and biases; X j The original features of node j are represented, including data such as the diameter, material, and historical defects of the pipeline segment. Feature propagation between nodes is achieved through an adjacency matrix. The multi-task learning model is based on the ResNet-50 backbone network, with a shared feature layer accounting for 60%. Experiments have verified that compared to the single-task model, the inference speed is improved by 25% and the memory usage is reduced by 40%.

[0071] The assessment module is used to calculate the final health index of the drainage pipeline based on the characteristics, parameters, and detection data, and to generate a health assessment report. Specifically, the module incorporates mainstream domestic and international assessment standards, such as the "Technical Specification for Inspection and Assessment of Urban Drainage Pipelines (CJJ181)," the Water Research Centre (WRC), and the Pipeline Assessment Certification Program (PACP). Based on the identified disease types and severity, as well as relevant pipeline parameters such as pipe diameter, material, and service life, the module calculates the health status index of the drainage pipeline, assesses its health condition, and determines whether repair is needed and its priority. The generated health assessment report includes basic pipeline information, inspection time, inspection results, a detailed description of the disease type, location, and severity, a health status assessment conclusion, a clear definition of the pipeline's health level, maintenance recommendations, and specific maintenance measures and priorities based on the assessment results.

[0072] The assessment module utilizes big data analytics, machine learning, and artificial intelligence algorithms to deeply mine the detection data. By learning from historical data, a health prediction model for drainage pipelines is established, such as a Long Short-Term Memory (LSTM) network model based on deep learning, to predict the development trend and potential risks of drainage pipeline defects. A three-level assessment index system is constructed using the Analytic Hierarchy Process (AHP), including structural safety, functional integrity, and environmental risk. Specifically, the structural safety index is calculated by multiplying the defect type weight matrix with geometric parameters; the functional integrity index is based on flow velocity and flow rate data, using Bernoulli's equation to calculate the pipeline's transport efficiency decay; and the environmental risk index predicts the remaining wall thickness of the metal pipeline using a corrosion rate model, and combines this with gas concentration data to construct an explosion risk probability model based on a Bayesian network. The final health index (HI) calculation formula is as follows:

[0073]

[0074] Where w i Let v be the weight of the i-th indicator. i v is the measured value. i,max The threshold is the upper limit. The structural safety index accounts for 50% of the weight. The geometric parameters of the defects are calculated from the three-dimensional coordinates of the lidar point cloud data. The system implements a four-level response based on the HI value. When HI < 30, a red warning is triggered. The system automatically calls the GIS-T module to generate an emergency repair route with traffic control suggestions and automatically generates a drainage pipeline health assessment report based on the data analysis results.

[0075] The early warning module is used to monitor health assessment reports in real time and send early warning information when serious diseases or abnormalities are detected. Specifically, when serious diseases or abnormalities are detected, the early warning module immediately triggers a maintenance work order, assigns the maintenance task to relevant departments or personnel, and sends an early warning notification to the display screen via SMS, email, APP push, etc., to remind operators to handle the situation in a timely manner.

[0076] The display screen is connected to the processor, receiving and presenting information from the processor to the user in real time. Specifically, the display screen has a graphical user interface (GUI), through which operators can monitor the working status, location information, and detection data of the spider-like bionic robot 1 in real time. The working status includes the battery level, speed, and posture of the spider-like bionic robot 1. The location information is displayed visually on a map showing the specific location of the spider-like bionic robot 1 in the drainage pipe network. The detection data displays temperature, humidity, gas concentration, and disease information in real time. Operators can remotely control the robot's movement speed, direction, and start and stop of the detection equipment through the display screen, flexibly adjusting the detection strategy according to actual detection needs. At the same time, the display screen receives health assessment reports and early warning information sent by the processor. Operators can formulate maintenance plans, assign maintenance tasks, and track maintenance progress based on the report content, achieving effective management of the drainage pipe network.

[0077] The drainage pipe network inspection system based on a spider-inspired robot of the present invention mimics the movement characteristics of a spider and integrates advanced detection technology to operate stably in complex pipe environments, acquire comprehensive inspection data about drainage pipes, and perform in-depth analysis and processing of the corresponding inspection data to generate health assessment reports, predict potential pipe risks, provide a scientific basis for the operation and maintenance management of drainage pipe networks, reduce operation and maintenance costs, and improve the reliability and safety of urban drainage systems.

[0078] The embodiments described above merely illustrate the preferred implementation of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A drainage pipe network inspection system based on a spider-inspired robot, characterized in that: include: A spider-inspired robot continuously inspects inside drainage pipes, collecting environmental parameters and images to obtain inspection data. The inspection well base, installed on the inner wall of the drainage pipe, provides charging and data transmission support for the spider-like robot. The processor, connected to the manhole base, receives and analyzes detection data from the spider-like robot, and generates a health assessment report for the drainage pipes. The detection data includes the temperature and humidity of the drainage pipe, internal images of the drainage pipe, the thickness and point cloud data of the inner wall of the drainage pipe, the gas composition and water flow rate inside the drainage pipe. The processor includes: The identification module is used to identify features in the detection data to obtain various disease characteristics; The analysis module is used to combine multiple disease characteristics with point cloud data to construct a three-dimensional model of the disease and accurately calculate the parameters of the corresponding disease. The assessment module is used to calculate the final health index of the drainage pipe based on the characteristics of the disease, disease parameters and detection data, and to assess the health status of the drainage pipe and generate a health assessment report. A three-level evaluation index system was constructed using the analytic hierarchy process (AHP), including structural safety, functional integrity, and environmental risk. The structural safety index was calculated by multiplying the defect type weight matrix with geometric parameters. The functional integrity index was based on flow velocity and flow rate data, using Bernoulli's equation to calculate the pipeline's transport efficiency decay. The environmental risk index was calculated by predicting the remaining wall thickness of the metal pipeline using a corrosion rate model, and combining this with gas concentration data to derive the final health index calculation formula. Where w i Let v be the weight of the i-th indicator. i v is the measured value. i,max The threshold is the upper limit; the structural safety index accounts for 50% of the weight, and the geometric parameters of the disease are calculated from the three-dimensional coordinates of the point cloud data; the system implements a four-level response based on the HI value, and triggers a red warning when HI < 30, and automatically generates a drainage pipeline health assessment report based on the data analysis results.

2. The drainage pipe network detection system based on a spider-inspired robot according to claim 1, characterized in that: The data processing system of the analysis module follows a three-level flow architecture from multi-source heterogeneous data to edge layer feature extraction, and then to cloud-based deep analysis. The multi-source heterogeneous data adopts parallel processing logic, and the original image, point cloud, sonar and environmental parameters are processed by independent algorithms to generate joint feature vectors. The processing flow of each branch has no time dependence and synchronous scheduling is achieved through hardware interrupt mechanism. The edge layer feature extraction uses the robot's built-in embedded computing unit to perform preliminary processing on the detection data while adopting a parallel processing architecture. Each branch executes independently to generate intermediate feature vectors. The image data is filtered to remove high contour features using nonlocal mean filtering; the lidar point cloud data is filtered to remove outliers using the RANSAC algorithm and downsampled using the voxel grid method; the sonar echo signal is decomposed by wavelet packet decomposition to extract energy feature vectors. The environmental parameters are fused from multi-source data using extended Kalman filtering to construct a pipeline environmental state vector; the prediction formula for the EKF state is: in represents the predicted value of state x at time k-1; f represents the state function; This represents the state estimate at time k-1; u k This represents the control input or external input at time k; The cloud-based deep analysis achieves the classification and calculation of disease parameters by sequentially processing the detection data through a three-level process: data governance, feature engineering, and intelligent decision-making. This includes: The data governance layer uses a distributed stream computing framework to achieve real-time cleaning, detects data anomalies through a sliding window algorithm, fills in missing values ​​using interpolation, and constructs a standardized dataset. The feature engineering layer uses graph neural networks to model the pipeline topology and combines an attention mechanism to extract spatial correlation features of diseases. Point cloud data is hierarchically organized through an octree structure, and sonar data is reduced in dimension through a convolutional autoencoder. The intelligent decision-making layer simultaneously classifies defects based on a multi-task learning model, categorizing defects into 14 types, and uses a Poisson surface reconstruction algorithm to generate a high-precision mesh model of the pipe inner wall.

3. The drainage pipe network detection system based on a spider-inspired robot according to claim 2, characterized in that: The graph neural network adopts a three-layer architecture, and its message passing mechanism is defined as m j →i=σ(W·X j +b), where m j →i represents the message passed from node j to node i, containing the features of j and its association with i; σ(·) represents the activation function; W and b represent the learnable weights and biases; X j The original characteristics of node j are represented, including data such as the diameter, material, and historical defects of the pipeline segment. The feature propagation between nodes is achieved through the adjacency matrix.

4. The drainage pipe network detection system based on a spider-inspired robot according to claim 3, characterized in that: The spider-inspired robot includes a shell, multiple walking legs, motion sensors, integrated detection components, and a control chip. The multiple walking legs are symmetrically arranged on both sides of the shell, driving the shell's movement. The motion sensors are mounted on the walking legs to detect their motion and force in real time. The integrated detection component is located at the front of the shell, continuously monitoring the external environment to obtain data. A control chip is located inside the shell, receiving data from the motion sensors and controlling the movement of the multiple walking legs.

5. The drainage pipe network inspection system based on a spider-inspired robot according to claim 4, characterized in that: The integrated detection components include a temperature sensor, a humidity sensor, a high-definition camera, a sonar, a lidar, an environmental parameter sensor, and a flow meter. The temperature sensor uses a high-precision thermistor to measure temperature changes within the drainage pipe, providing data support for analyzing the physical and chemical processes inside the pipe. The humidity sensor monitors the humidity within the drainage pipe in real time to determine if leaks or other problems exist. The high-definition camera uses a global shutter CMOS sensor and is equipped with an infrared LED matrix illumination system. The LED light source has a wavelength of 850nm, and the illumination uniformity is ≥95%. The focal length of the high-definition camera can automatically switch within the range of 5-50mm to achieve optical imaging of 1mm-level defects on the inner wall of the pipe in a low-light environment of 0.1 lux. The sonar system emits radio frequency... The system employs sound waves with a frequency range of 200-500kHz to penetrate 30cm of silt to detect changes in the wall thickness of drainage pipes. Time-frequency domain features are extracted from the echo signals using a short-time Fourier transform. The lidar utilizes a multi-line rotating structure equipped with 64 laser beams and a scanning frequency of 10-20Hz. It constructs point cloud data of the drainage pipe's inner wall using a time-of-flight method, generating a 3D coordinate matrix of 100,000 points per second. The environmental parameter sensor integrates a high-precision temperature and humidity composite probe and an electrochemical gas sensor array, with detection limits of 0.1ppm for hydrogen sulfide and 1ppm for methane. The flow meter tracks the trajectory of floating objects using the Lucas-Kanad optical flow algorithm and, combined with the drainage pipe's cross-sectional area parameters, calculates the real-time flow rate using a continuity equation.

6. The drainage pipe network detection system based on a spider-inspired robot according to claim 5, characterized in that: The control chip includes an action module, a power detection module, and a navigation module. The action module controls the movement of multiple walking legs to enable the spider-like bionic robot to move inside the drainage pipe. The power detection module continuously monitors the power level of the power supply component and sends an automatic charging command to the navigation module when the power level of the power supply component is below 20%. The navigation module plans the shortest path to the nearest manhole base and sends it to the action module. The action module controls the spider-like bionic robot to go to the manhole base for charging based on the shortest path.

7. The drainage pipe network detection system based on a spider-inspired robot according to claim 6, characterized in that: The navigation module combines inertial navigation, visual navigation, and geomagnetic navigation technologies. The inertial navigation technology measures the angular velocity and acceleration of the spider-like robot using angle and accelerometer sensors on its walking legs to calculate the robot's posture. The visual navigation technology uses a high-definition camera to acquire image information of the surrounding environment and determines the position and orientation of the spider-like robot through image recognition and matching algorithms. The geomagnetic navigation technology determines the orientation of the spider-like robot based on changes in the Earth's magnetic field.

8. The drainage pipe network inspection system based on a spider-inspired robot according to any one of claims 1-7, characterized in that: The processor also includes a data storage module, which is used to classify and store the large amount of received detection data through a combination of a distributed file system and a relational database.

9. The drainage pipe network detection system based on a spider-inspired robot according to claim 8, characterized in that: The processor also includes an early warning module for real-time monitoring of health assessment reports and sending early warning information when serious diseases or abnormalities are detected.

10. The drainage pipe network detection system based on a spider-inspired robot according to claim 9, characterized in that: It also includes a display screen that connects to the processor to receive and present health assessment reports and warning information from the processor to the user in real time.

Images