Intelligent underground cavern wall surface roughness measuring system and method based on multi-sensor fusion

Abstract

The invention provides an underground cavity wall surface roughness intelligent measurement system and method based on multi-sensor fusion. The system comprises a multi-sensor fusion device, an environmental parameter acquisition sensor, an intelligent signal processing unit, a power management system and a cloud analysis platform. And the cloud analysis platform is used for comprehensively analyzing the roughness estimation values corresponding to the sampling intervals to generate a three-dimensional roughness model of the wall surface of the underground cavern. The invention provides an underground cavern wall surface roughness intelligent measurement system and method based on multi-sensor fusion, which is a wall surface roughness measurement system integrating various non-contact sensing technologies and having environment self-adaptive capability, intelligent signal processing capability and portable deployment characteristics, so that the measurement precision and efficiency are improved, and the measurement cost is reduced. And stable operation can be realized in a complex underground environment, real-time transmission and three-dimensional modeling of data are supported, and the requirements of intelligent underground engineering management and quality control in the future are met.

Description

Technical Field

[0001] The present invention relates to the technical field of roughness measurement, and particularly to an intelligent measurement system and method for the roughness of the wall surface of an underground chamber based on multi-sensor fusion. Background Art

[0002] In large-scale underground space projects, such as pumped storage power stations, underground powerhouses, and traffic tunnels, the roughness of the chamber wall surface not only affects the evaluation of structural stability, but also is directly related to the air flow distribution, resistance coefficient, and energy consumption level of the ventilation system. The currently widely used roughness measurement methods are mostly manual contact means, such as needle profilometers, contact probes, etc. Although they have a certain accuracy in flat areas, they are difficult to operate, have low efficiency, limited measurement range, and pose potential safety hazards to personnel in underground environments such as deep burial, narrowness, and insufficient lighting.

[0003] With the development of non-contact measurement technologies, measurement methods based on acoustic wave reflection have gradually been studied and applied to surface feature extraction. Acoustic wave measurement has advantages such as strong penetration and adaptability to low-light environments, and is particularly suitable for underground chambers with complex structures or harsh environments. However, most existing acoustic wave measurement devices are single-point sampling, rely on manual deployment, have low system integration, and are difficult to achieve continuous high-precision perception of the roughness of a large range of wall surfaces. At the same time, under the interference of complex and variable temperature, humidity, and background noise in the chamber, the measurement error is large, and it is difficult to perform real-time correction and big data analysis. Summary of the Invention

[0004] In view of the technical problems existing in the existing measurement of the roughness of the wall surface of underground chambers, such as difficult deployment, low measurement accuracy, and poor environmental adaptability, the present invention proposes an intelligent measurement system and method for the roughness of the wall surface of an underground chamber based on multi-sensor fusion, an intelligent measurement system that integrates acoustic wave reflection and multi-source sensing. The system adopts a handheld or semi-fixed deployment method, combines multi-source sensor integration and an adaptive signal processing algorithm to achieve rapid, stable, and high-precision perception of the roughness of the wall surface under complex terrains.

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

[0006] The present invention provides an intelligent measurement system for the roughness of the wall surface of an underground chamber based on multi-sensor fusion, including a multi-sensor fusion device, an environmental parameter acquisition sensor, an intelligent signal processing unit, a power management system, and a cloud analysis platform;

[0007] The multi-sensor fusion device is integrally installed with an ultrasonic sensor, a lidar scanning unit, and a structured light camera; the ultrasonic sensor, the lidar scanning unit, and the structured light camera are used to synchronously sample each sampling interval of the underground chamber wall surface, and respectively and real-time measure the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface of each sampling interval, and upload them to the intelligent signal processing unit in real time;

[0008] The environmental parameter acquisition sensor is used to collect the environmental parameters of the underground chamber in real time and upload them to the intelligent signal processing unit in real time;

[0009] The intelligent signal processing unit includes a signal preprocessing module, an environmental parameter compensation module, a feature extraction module, a data fusion module, and a roughness estimation module based on a neural network;

[0010] The signal preprocessing module is used to respectively perform data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface of each sampling interval to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface;

[0011] The environmental parameter compensation module is used to perform environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data;

[0012] The feature extraction module is used to respectively perform feature extraction on the compensated ultrasonic detection distance data, the preprocessed three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface to obtain the ultrasonic detection distance waveform feature, the three-dimensional point cloud feature of the underground chamber wall surface, and the three-dimensional image feature of the underground chamber wall surface;

[0013] The data fusion module is used to perform data fusion on the ultrasonic detection distance waveform feature, the three-dimensional point cloud feature of the underground chamber wall surface, and the three-dimensional image feature of the underground chamber wall surface to obtain the data fusion feature corresponding to the sampling interval;

[0014] The roughness estimation module based on a neural network is used to perform roughness estimation on the data fusion feature corresponding to the sampling interval, output the roughness estimation value corresponding to the sampling interval, and send it to the cloud analysis platform;

[0015] The cloud analysis platform is used to comprehensively analyze the roughness estimation values corresponding to each sampling interval to generate a three-dimensional roughness model of the underground chamber wall surface.

[0016] Preferably, the multi-sensor fusion device adopts a bracket deployment method, specifically using an adjustable pan-tilt bracket to achieve the adjustment of vertical, pitch, and horizontal angles, so as to realize the precise measurement of each sampling interval by the ultrasonic sensor, the lidar scanning unit, and the structured light camera; the pan-tilt of the adjustable pan-tilt bracket adopts a three-degree-of-freedom electric pan-tilt, and the cloud analysis platform uses a pan-tilt control unit to remotely control and adjust the azimuth angle, pitch angle, and scanning path of the three-degree-of-freedom electric pan-tilt;

[0017] The ultrasonic sensor is arranged in the middle and lower part of the adjustable pan-tilt bracket, vertically pointing to the wall surface of the current sampling interval, and measures the undulation and unevenness of each sampling point in the sampling interval by collecting the time difference between the ultrasonic wave emission and the reception after the ultrasonic wave is reflected and the ultrasonic intensity characteristics;

[0018] The lidar scanning unit is arranged at the top center of the adjustable pan-tilt bracket and adopts a rotary scanning mode to obtain the point cloud contour data of the sampling interval;

[0019] The structured light camera is arranged below the adjustable pan-tilt bracket, and generates a high-precision depth map and texture image of the sampling interval through active projection of fringe light coding and binocular vision principle.

[0020] Preferably, the environmental parameter acquisition sensors include a temperature and humidity sensor, a barometer, and a dust sensor, which are respectively used to collect the temperature and humidity, air pressure, and dust concentration in the underground chamber.

[0021] Preferably, the signal preprocessing module is specifically used for noise suppression and blind source separation of the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface.

[0022] Preferably, the environmental parameter compensation module compensates the ultrasonic wave propagation speed by using the following formula:

[0023] V corrected =V measured ×(1 + αΔT + βΔH)

[0024] Where: V measured is the ultrasonic wave propagation speed under the set temperature and set humidity conditions; α and β are the compensation coefficients of temperature and humidity respectively; ΔT is the change in the current temperature relative to the set temperature; ΔH is the change in the current humidity relative to the set humidity; V corrected is the compensated ultrasonic wave propagation speed;

[0025] Adopt the compensated ultrasonic wave propagation speed V corrected, based on the time difference between the ultrasonic wave emission and the reception after the ultrasonic wave is reflected, the ultrasonic detection distance data is obtained, which is the compensated ultrasonic detection distance data.

[0026] Preferably, the feature extraction module is specifically configured to:

[0027] Extract waveform features from the compensated ultrasonic detection distance data, including signal amplitude, time delay, and peak value, to obtain the ultrasonic detection distance waveform features;

[0028] Extract waveform features from the preprocessed three-dimensional point cloud data of the underground chamber wall surface, including signal amplitude, time delay, and peak value, to obtain the three-dimensional point cloud features of the underground chamber wall surface;

[0029] Perform texture analysis on the preprocessed three-dimensional image data of the underground chamber wall surface, extract the local structure information of the image, including the mean value, variance, and contrast of the image; then extract the object contour through edge detection; the local structure information of the image and the object contour form the three-dimensional image features of the underground chamber wall surface.

[0030] Preferably, the data fusion module is specifically configured to: perform weighted averaging on the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall surface, and the three-dimensional image features of the underground chamber wall surface to obtain the data fusion features corresponding to the sampling interval.

[0031] Preferably, the neural network-based roughness estimation module estimates roughness using the following method:

[0032] For sampling interval j, assume there are N sampling points. For sampling point i, i = 1, 2,..., N, its data fusion feature is represented as F i ';

[0033] Input the data fusion feature F i ' of sampling point i in sampling interval j into the neural network-based roughness estimation module, and the neural network-based roughness estimation module further extracts features from the data fusion feature F i ' to obtain the feature vector F i ; The mean value of the feature vectors of N sampling points is Use the following formula to obtain the roughness estimation value R of sampling interval j j :

[0034]

[0035] Thus, the roughness estimation of the sampling interval is achieved.

[0036] Preferably, the cloud analysis platform is specifically configured to:

[0037] Based on the preprocessed three-dimensional point cloud data and three-dimensional image data of the underground cavern wall surface, and combining the roughness estimation values of each sampling interval, surface fitting and topological reconstruction are performed on the underground cavern wall surface to obtain a three-dimensional reconstruction model of the underground cavern wall surface;

[0038] Analyze the three-dimensional reconstruction model of the underground cavern wall surface. For each sampling interval j, m wall curves are determined at intervals along the height direction. For each wall curve k, k = 1, 2,..., m, the wall curve function f k (x) is read from the three-dimensional reconstruction model of the underground cavern wall surface, where x is the wall depth direction. Using the following formula, the roughness value R of wall curve k is obtained a,k :

[0039]

[0040] where L is the length of the sampling interval; is the roughness estimation value of sampling interval j, equal to R j ;

[0041] Thus, m roughness values of sampling interval j are obtained, and they are combined to form the roughness representation of sampling interval j.

[0042] The present invention also provides a measurement method of the intelligent measurement system for the roughness of the underground cavern wall surface based on multi-sensor fusion, including the following steps:

[0043] Step S1, for each sampling interval of the underground cavern wall surface, the ultrasonic sensor, the lidar scanning unit, and the structured light camera perform synchronous sampling, and respectively and real-time measure the ultrasonic detection distance data, the three-dimensional point cloud data of the underground cavern wall surface, and the three-dimensional image data of the underground cavern wall surface, and upload them to the intelligent signal processing unit in real time;

[0044] At the same time, the environmental parameter acquisition sensor real-time collects the environmental parameters of the underground cavern and uploads them to the intelligent signal processing unit in real time;

[0045] Step S2, the intelligent signal processing unit performs data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground cavern wall surface, and the three-dimensional image data of each sampling interval to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground cavern wall surface, and the three-dimensional image data of the underground cavern wall surface;

[0046] Step S3, the intelligent signal processing unit performs environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data;

[0047] Step S4: The intelligent signal processing unit extracts features from the compensated ultrasonic detection distance data, the preprocessed 3D point cloud data of the underground chamber wall surface, and the 3D image data of the underground chamber wall surface respectively, to obtain ultrasonic detection distance waveform features, 3D point cloud features of the underground chamber wall surface, and 3D image features of the underground chamber wall surface;

[0048] Step S5: The intelligent signal processing unit fuses the ultrasonic detection distance waveform features, the 3D point cloud features of the underground chamber wall surface, and the 3D image features of the underground chamber wall surface to obtain the data fusion features corresponding to the sampling interval;

[0049] Step S6: The intelligent signal processing unit uses a roughness estimation module based on a neural network to estimate the roughness of the data fusion features corresponding to the sampling interval, outputs the roughness estimation value corresponding to the sampling interval, and sends it to the cloud analysis platform;

[0050] Step S7: The cloud analysis platform comprehensively analyzes the roughness estimation values corresponding to each sampling interval to generate a 3D roughness model of the underground chamber wall surface.

[0051] The intelligent measurement system and method for the roughness of the underground chamber wall surface based on multi-sensor fusion provided by the present invention has the following advantages:

[0052] The present invention provides an intelligent measurement system and method for the roughness of the underground chamber wall surface based on multi-sensor fusion, which is a wall surface roughness measurement system integrating a variety of non-contact sensing technologies, with environmental adaptability, intelligent signal processing ability and portable deployment characteristics. It not only improves the measurement accuracy and efficiency, but also can operate stably in complex underground environments, and supports real-time data transmission and 3D modeling, meeting the needs of future intelligent underground engineering management and quality control. Brief Description of the Drawings

[0053] Figure 1 It is the overall structural schematic diagram of the intelligent measurement system for the roughness of the underground chamber wall surface based on multi-sensor fusion provided by the present invention;

[0054] Figure 2 It is the structural schematic diagram of the multi-sensor fusion device provided by the present invention;

[0055] Figure 3 It is the flowchart of the signal processing by the intelligent signal processing unit provided by the present invention;

[0056] Figure 4 It is the overall flowchart of the intelligent measurement method for the roughness of the underground chamber wall surface based on multi-sensor fusion provided by the present invention.

[0057] Wherein: 1 - multi - sensor fusion device; 1.1 - ultrasonic sensor; 1.2 - lidar scanning unit; 1.3 - structured light camera; 1.4 - adjustable pan - tilt bracket. Detailed implementation manners

[0058] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] The present invention provides an intelligent measurement system and method for the roughness of the underground chamber wall based on multi - sensor fusion, which is a wall roughness measurement system integrating a variety of non - contact sensing technologies, having environmental adaptability, intelligent signal processing ability and portable deployment characteristics. It not only improves the measurement accuracy and efficiency, but also can operate stably in a complex underground environment, and supports real - time data transmission and 3D modeling, meeting the needs of future intelligent underground engineering management and quality control.

[0060] Refer to Figure 1 and Figure 2 , the present invention provides an intelligent measurement system for the roughness of the underground chamber wall based on multi - sensor fusion, including a multi - sensor fusion device, an environmental parameter acquisition sensor, an intelligent signal processing unit, a power management system and a cloud analysis platform.

[0061] The following is a detailed introduction to each module:

[0062] (1) Multi - sensor fusion device

[0063] The multi - sensor fusion device is integrally installed with an ultrasonic sensor, a lidar scanning unit and a structured light camera; the ultrasonic sensor, the lidar scanning unit and the structured light camera are used to synchronously sample each sampling interval of the underground chamber wall, and respectively and real - time measure the ultrasonic detection distance data, the three - dimensional point cloud data of the underground chamber wall and the three - dimensional image data of the underground chamber wall in each sampling interval, and upload them to the intelligent signal processing unit in real - time;

[0064] Specifically, at least one ultrasonic sensor is arranged to measure the distance between the underground chamber wall and the multi - sensor fusion device;

[0065] At least one lidar scanning unit is arranged to perform laser scanning on the underground chamber wall and obtain three - dimensional point cloud data;

[0066] At least one structured light camera is arranged to obtain three - dimensional image data of the surface of the underground chamber wall;

[0067] The multi-sensor fusion device adopts a bracket deployment method, specifically using an adjustable pan-tilt bracket to achieve adjustment of vertical, pitch, and horizontal angles, enabling precise measurement of each sampling interval by the ultrasonic sensor, the lidar scanning unit, and the structured light camera; the pan-tilt of the adjustable pan-tilt bracket uses a three-degree-of-freedom electric pan-tilt, and the cloud analysis platform uses a pan-tilt control unit to remotely control and adjust the azimuth angle, pitch angle, and scanning path of the three-degree-of-freedom electric pan-tilt;

[0068] The ultrasonic sensor is arranged in the middle and lower part of the adjustable pan-tilt bracket, vertically pointing to the wall surface of the current sampling interval, and measures the undulation and unevenness of each sampling point in the sampling interval by collecting the time difference between ultrasonic emission and ultrasonic reception after reflection and the ultrasonic intensity characteristics;

[0069] The lidar scanning unit is arranged at the top center of the adjustable pan-tilt bracket and uses a rotational scanning mode to obtain the point cloud contour data of the sampling interval;

[0070] The structured light camera is arranged below the adjustable pan-tilt bracket and generates a high-precision depth map and texture image of the sampling interval through active projection of fringe light coding and binocular vision principle.

[0071] (2) Environmental parameter acquisition sensors

[0072] The environmental parameter acquisition sensors are used to collect the environmental parameters of the underground cavern in real time and upload them to the intelligent signal processing unit in real time; specifically, the environmental parameter acquisition sensors include a temperature and humidity sensor, a barometer, and a dust sensor, which are respectively used to collect the temperature and humidity, air pressure, and dust concentration of the underground cavern, realize real-time monitoring of the change of the measurement environment, and provide environmental data to the intelligent signal processing unit for compensation.

[0073] (3) Intelligent signal processing unit

[0074] The intelligent signal processing unit is used to fuse the signals from the ultrasonic sensor, the lidar scanning unit, and the structured light camera, and calculate the estimated value of the wall roughness;

[0075] Combined with Figure 3 , the intelligent signal processing unit includes a signal preprocessing module, an environmental parameter compensation module, a feature extraction module, a data fusion module, and a roughness estimation module based on a neural network.

[0076] (3.1) Signal preprocessing module

[0077] The signal preprocessing module is used to perform data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface in each sampling interval, so as to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface;

[0078] The signal preprocessing module is specifically used to perform noise suppression and blind source separation on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface.

[0079] (3.2) Environmental parameter compensation module

[0080] The environmental parameter compensation module is used to perform environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data; wherein, the environmental parameters include at least one of the temperature and humidity, air pressure, and dust concentration in the underground chamber.

[0081] The environmental parameter compensation module compensates the ultrasonic wave propagation speed by using the following formula:

[0082] V corrected =V meassured ×(1 + αΔT + βΔH)

[0083] Where: V measured is the ultrasonic wave propagation speed under the set temperature and set humidity conditions; α and β are the compensation coefficients for temperature and humidity respectively; ΔT is the change in the current temperature relative to the set temperature; ΔH is the change in the current humidity relative to the set humidity; V corrected is the compensated ultrasonic wave propagation speed;

[0084] Using the compensated ultrasonic wave propagation speed V corrected , based on the time difference between the ultrasonic wave emission and the reception of the reflected ultrasonic wave, the ultrasonic detection distance data is obtained, which is the compensated ultrasonic detection distance data.

[0085] (3.3) Feature extraction module

[0086] The feature extraction module is used to perform feature extraction on the compensated ultrasonic detection distance data, the preprocessed three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface respectively, so as to obtain the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall surface, and the three-dimensional image features of the underground chamber wall surface;

[0087] The feature extraction module is specifically used for:

[0088] Extract waveform features from the compensated ultrasonic detection distance data, including signal amplitude, time delay, and peak value, to obtain the ultrasonic detection distance waveform features;

[0089] Extract waveform features from the preprocessed 3D point cloud data of the underground chamber wall surface, including signal amplitude, time delay, and peak value, to obtain the 3D point cloud features of the underground chamber wall surface;

[0090] Perform texture analysis on the preprocessed 3D image data of the underground chamber wall surface, extract the local structure information of the image, including the mean, variance, and contrast of the image; then extract the object contour through edge detection; the local structure information of the image and the object contour form the 3D image features of the underground chamber wall surface.

[0091] (3.4) Data fusion module

[0092] The data fusion module is used to perform data fusion on the ultrasonic detection distance waveform features, the 3D point cloud features of the underground chamber wall surface, and the 3D image features of the underground chamber wall surface to obtain the data fusion features corresponding to the sampling interval;

[0093] Specifically, the data fusion module: performs weighted averaging on the ultrasonic detection distance waveform features, the 3D point cloud features of the underground chamber wall surface, and the 3D image features of the underground chamber wall surface to obtain the data fusion features corresponding to the sampling interval.

[0094] (3.5) Roughness estimation module based on neural network

[0095] The roughness estimation module based on neural network is used to estimate the roughness of the data fusion features corresponding to the sampling interval, output the roughness estimation value corresponding to the sampling interval, and send it to the cloud analysis platform;

[0096] The roughness estimation module based on neural network uses the following method for roughness estimation:

[0097] For sampling interval j, assume there are N sampling points. For sampling point i, i = 1, 2,..., N, its data fusion feature is denoted as F i ';

[0098] Input the data fusion feature F i ' of sampling point i in sampling interval j into the roughness estimation module based on neural network, and the roughness estimation module based on neural network performs further feature extraction on the data fusion feature F i ' to obtain the feature vector F i ; The mean value of the feature vectors of N sampling points is Use the following formula to obtain the roughness estimation value R of sampling interval jj :

[0099]

[0100] In this way, the roughness estimation of the sampling interval is realized.

[0101] In the present invention, the roughness estimation module based on neural network adopts a deep learning algorithm, that is, a convolutional neural network (CNN) to estimate the roughness of the data fusion features, so as to improve the roughness estimation accuracy.

[0102] (IV) Cloud analysis platform

[0103] The cloud analysis platform is used to store, process and analyze the measurement data to generate a three-dimensional roughness model. The cloud computing platform generates a three-dimensional reconstruction model of the roughness of the underground cavern wall based on big data analysis and three-dimensional modeling technology for construction management personnel to carry out quality control and analysis.

[0104] Specifically, the cloud analysis platform is used to comprehensively analyze the roughness estimation values corresponding to each sampling interval to generate a three-dimensional roughness model of the underground cavern wall.

[0105] According to the preprocessed three-dimensional point cloud data and three-dimensional image data of the underground cavern wall, combined with the roughness estimation value of each sampling interval, surface fitting and topological reconstruction are carried out on the underground cavern wall to obtain a three-dimensional reconstruction model of the underground cavern wall;

[0106] Analyze the three-dimensional reconstruction model of the underground cavern wall. For each sampling interval j, m wall curves are determined at intervals along the height direction. For each wall curve k, k = 1, 2,..., m, the wall curve function f k (x) is read from the three-dimensional reconstruction model of the underground cavern wall, where x is the wall depth direction. The following formula is used to obtain the roughness value R of the wall curve k a,k :

[0107]

[0108] where L is the length of the sampling interval; is the roughness estimation value of the sampling interval j, equal to R j ;

[0109] Thus, m roughness values of the sampling interval j are obtained, and they are combined to form the roughness representation of the sampling interval j.

[0110] In the above way, for each sampling interval, multiple roughness values are calculated to realize the fine calculation of roughness.

[0111] Refer to Figure 4, the present invention also provides a measurement method for an intelligent measurement system of the roughness of the underground chamber wall based on multi-sensor fusion, including the following steps:

[0112] Step S1, for each sampling interval of the underground chamber wall, the ultrasonic sensor, the lidar scanning unit, and the structured light camera perform synchronous sampling, and respectively measure and obtain the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall, and the three-dimensional image data of the underground chamber wall in real time, and upload them to the intelligent signal processing unit in real time;

[0113] Meanwhile, the environmental parameter acquisition sensor collects the environmental parameters of the underground chamber in real time and uploads them to the intelligent signal processing unit in real time;

[0114] Step S2, the intelligent signal processing unit performs data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall, and the three-dimensional image data of the underground chamber wall for each sampling interval to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall, and the three-dimensional image data of the underground chamber wall;

[0115] Step S3, the intelligent signal processing unit performs environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data;

[0116] Step S4, the intelligent signal processing unit respectively extracts features from the compensated ultrasonic detection distance data, the preprocessed three-dimensional point cloud data of the underground chamber wall, and the three-dimensional image data of the underground chamber wall to obtain the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall, and the three-dimensional image features of the underground chamber wall;

[0117] Step S5, the intelligent signal processing unit performs data fusion on the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall, and the three-dimensional image features of the underground chamber wall to obtain the data fusion features corresponding to the sampling interval;

[0118] Step S6, the intelligent signal processing unit uses a roughness estimation module based on a neural network to estimate the roughness of the data fusion features corresponding to the sampling interval, outputs the roughness estimation value corresponding to the sampling interval, and sends it to the cloud analysis platform;

[0119] Step S7, the cloud analysis platform comprehensively analyzes the roughness estimation values corresponding to each sampling interval to generate a three-dimensional roughness model of the underground chamber wall.

[0120] The present invention provides an intelligent measurement system and method for the roughness of the underground chamber wall based on multi-sensor fusion, having the following characteristics:

[0121] The multi-sensor fusion device of the present invention integrates an ultrasonic transmitting and receiving unit, a lidar scanning unit, and a structured light camera unit, calculates the roughness characteristic parameters of the cave wall surface according to the characteristic changes of the reflected wave (such as reflection time, waveform intensity, etc.); and is equipped with environmental parameter acquisition sensors such as temperature, humidity, and air pressure sensors for real-time compensation of the sound speed and optical measurement errors; the modular design can flexibly configure the type and quantity of sensors according to different underground engineering requirements, and has good scalability and environmental adaptability.

[0122] The intelligent signal processing unit of the present invention adopts a reflected waveform feature extraction algorithm based on a deep neural network to automatically identify the microscopic unevenness of the wall surface; integrates an adaptive frequency adjustment and gain control algorithm, dynamically optimizes the transmission and reception settings in combination with environmental parameters, and improves the measurement accuracy; embeds a blind source separation and noise suppression module to improve the stability and signal-to-noise ratio in a noise interference environment.

[0123] The present invention realizes portable operation and deployment: the system is designed as a portable integrated device, which can be flexibly operated by operators installed on an adjustable bracket. It is suitable for rapid deployment and point-to-point mobile measurement in complex underground environments, adapts to the scanning requirements of different wall heights and angles, and improves the usability and on-site adaptability.

[0124] The present invention realizes low-power and environmental adaptation design: combined with an intelligent power management system (BMS), it extends the on-site battery life; the environmental compensation module collects data such as temperature, humidity, and dust concentration in real time to correct the measurement results.

[0125] The present invention realizes cloud big data and 3D modeling: the wireless communication module adopts a multi-path self-organizing network technology to upload the measurement data to the cloud analysis platform in real time; the cloud analysis platform provides a global roughness trend analysis of the underground chamber based on big data analysis and machine learning models, and generates a 3D roughness model with centimeter-level accuracy.

[0126] The working process of this system is as follows:

[0127] Operators perform point-by-point scanning and sampling on the wall surface in a designated area inside the cave chamber, and each sensor module in the multi-sensor fusion device collects acoustic and optical reflection data; the intelligent signal processing unit fuses multi-source information in real time and calculates the roughness parameters; the measurement data is transmitted to the cloud analysis platform through a wireless network for global analysis and modeling, and finally a 3D roughness model is formed.

[0128] Compared with the prior art, the present invention has the following remarkable advantages and technical effects:

[0129] (1) Significantly improved measurement accuracy: By integrating non-contact sensors such as ultrasonic, lidar, and structured light, and combining deep learning algorithms to extract microscopic features of reflected waves, millimeter-level roughness measurement is achieved, with higher accuracy than traditional single-sensing methods.

[0130] (2) Convenient and flexible operation mode: Adopting a portable operation mode, without the need for complex tracks and mobile platforms, it is suitable for various-shaped caverns and working environments, with fast deployment and easy use.

[0131] (3) Strong environmental adaptability and robustness: Real-time environmental monitoring and parameter compensation ensure high stability and reliability even under harsh underground conditions such as large temperature and humidity fluctuations and a lot of dust.

[0132] (4) Low-power and long-endurance design: Supports long-term field operations, meeting the dual requirements of large cavern measurement tasks for endurance time and equipment portability.

[0133] (5) Big data-driven intelligent analysis ability: Cooperating with the cloud platform for big data analysis and 3D modeling, surface reconstruction can be achieved, significantly improving the level of construction quality control.

[0134] (6) Modular design, wide range of applicable scenarios: Each part of the system is modular and can be combined flexibly. It is applicable to various engineering environments such as hydraulic tunnels, mine ventilation ducts, and underground powerhouses, with good expandability and industrial application prospects.

[0135] The following introduces an embodiment:

[0136] An intelligent measurement system for the roughness of the underground cavern wall based on multi-sensor fusion and autonomous mobile platform proposed in the embodiment of the present invention mainly consists of the following parts:

[0137] Multi-sensor fusion device 1: The multi-sensor fusion device 1 is the core sensing unit of the system, integrating three types of non-contact sensing devices: ultrasonic sensor 1.1, lidar scanning unit 1.2, and structured light camera 1.3. The ultrasonic sensor 1.1 is used to detect the time difference and intensity change of the echo signal, with good low-light adaptability and penetration performance, suitable for measuring irregular walls; the lidar scanning unit 1.2 realizes point cloud data acquisition to accurately obtain the macroscopic geometric contour of the wall; the structured light camera 1.3 provides sub-millimeter-level depth maps and texture information for supplementing high-precision surface features. All sensors achieve data time alignment through a synchronous control module and have the ability to flexibly configure the number of sensors, which can be selectively enabled according to different engineering scenarios.

[0138] Intelligent Signal Processing Unit: This unit integrates a high-performance embedded computing module, runs multi-source data fusion and deep learning algorithms, and completes real-time calculation of roughness parameters. The intelligent signal processing unit has functions of waveform feature extraction and frequency-domain analysis, environment adaptive gain and frequency control, noise suppression and blind source separation, as well as a roughness estimation algorithm based on neural network.

[0139] Environmental Parameter Acquisition Sensor and Compensation Module: The module is equipped with temperature and humidity sensors, barometers, and dust sensors for real-time collection of micro-meteorological data in the cave. The output of this module is used to: correct the measurement error of ultrasonic sensors, judge the intensity of environmental interference, and assist the intelligent signal processing unit to adaptively adjust parameter settings.

[0140] Power Management System: Designed with an intelligent battery management system (BMS), it supports high-energy density lithium battery modules and has functions of adaptive power consumption control, temperature protection, and remaining power prediction. Combining a dynamic switching mechanism between the sleep mode and the working mode, it realizes the endurance guarantee for long-term field operations.

[0141] Wireless Communication and Cloud Analysis Platform: The module supports multi-path ad-hoc network communication protocols and has various data transmission methods such as point-to-point and point-to-cloud. Through built-in encryption algorithms and signal automatic switching mechanisms, it ensures data integrity and real-time performance in underground weak signal environments. The wireless communication module uploads roughness data and original sensing information to the cloud analysis platform. The cloud analysis platform, based on big data analysis and machine learning models, realizes global integration and in-depth analysis of data: fusing test data and trend prediction to generate a wall roughness model with centimeter-level accuracy.

[0142] The above-mentioned modules transmit data and instructions through the system bus. The overall structure adopts a modular design, which is convenient for combined deployment in different engineering scenarios. The system can realize the full-process automatic operation from data acquisition, feature extraction, model reconstruction to analysis and decision-making, and is applicable to various underground environments such as underground powerhouses, hydraulic tunnels, and mine ventilation ducts, with the advantages of high precision, high efficiency, and high adaptability.

[0143] See Figure 2, the multi-sensor fusion device 1 of the present invention adopts a bracket-type fixed deployment method, which is suitable for long-term operations in underground chambers, high-precision wall measurements, and stable data acquisition requirements. The system is mainly installed on the adjustable pan-tilt bracket 1.4, which can achieve flexible adjustment of vertical, pitch, and horizontal angles to ensure optimal alignment and effective coverage of various sensors on the wall. The overall structure uses an aluminum alloy telescopic bracket, which has the characteristics of high strength and light weight; the top of the bracket is equipped with a three-degree-of-freedom electric pan-tilt, which supports remote control to adjust the azimuth angle, pitch angle, and scanning path; the pan-tilt control unit is connected to the central processing unit to realize automatic positioning and scanning area planning; the bottom of the bracket is equipped with stable anchor feet and a level to ensure stability on terrains such as slopes and uneven bedrock surfaces.

[0144] Various sensors are arranged collaboratively according to their functions, following the principles of overlapping sampling, non-interference, and field-of-view complementarity. The ultrasonic sensor 1.1 is arranged in the middle and lower part of the platform, pointing vertically at the target wall, mainly collecting the reflection time difference and intensity characteristics for estimating surface undulations and irregularities; the lidar scanning unit 1.2 is installed at the center of the top of the bracket, adopting a 360° rotation scanning mode to obtain large-range point cloud contour data; the structured light camera 1.3 is fixed below the lidar, generating high-precision depth maps and texture images through active projection of stripe light encoding and binocular vision principles to supplement detailed information. All sensors are hardware-triggered and data timestamp-aligned through a unified time synchronization control module to ensure the synchrony and consistency of multi-source data.

[0145] The system supports fine adjustment of the parameters of different sensors to adapt to the changing underground environmental conditions. Specifically, the operating frequency of the ultrasonic sensor is 40–200 kHz, the detection distance range is 0.2–6 m, and the transmission power can be automatically optimized according to the echo intensity; the scanning resolution of the lidar scanner is 0.1°–0.5°, the horizontal scanning range is 360°, the vertical range is -15°– +15°, the effective measurement radius is ≤15 m, and the typical accuracy is ±10 mm; the depth resolution of the structured light camera is ≤0.3 mm, the field of view angle is 70°×55°, the effective working distance is 0.5-3 m, the light adaptation range is 5-1000 lux, and it has an automatic exposure compensation algorithm built in. All sensors have a temperature drift correction function and can perform dynamic error compensation according to the micro-meteorological data collected on-site.

[0146] Through the arrangement of a highly stable bracket and the collaborative acquisition of various sensors, the system can perform continuous and high-precision roughness measurements on underground walls of different sizes, shapes, and materials, ensuring the spatio-temporal consistency and quality reliability of the data.

[0147] Combined with Figure 3, The signal processing and roughness calculation process mainly includes five steps: noise suppression and blind source separation, real-time compensation mechanism for environmental parameters, preprocessing (filtering, enhancement) of echo and optical data, feature extraction and data fusion, and roughness parameter estimation using a deep neural network.

[0148] 1. Noise suppression and blind source separation

[0149] The echo signal and optical image after data acquisition usually contain noise and interference signals, so preprocessing is required to improve data quality and accuracy. The filter used is Wiener filtering; subsequently, independent component analysis (ICA) is used to separate the mixed signals and extract the original signals.

[0150] 2. Real-time compensation mechanism for environmental parameters

[0151] The current environmental parameters, such as temperature, humidity, and air pressure, are obtained in real time through sensors. According to the temperature and humidity data, the propagation speed of the signal collected by the ultrasonic sensor is corrected. The compensation formula for the propagation speed is:

[0152] V corrected =V measured ×(1 + αΔT + βΔH)

[0153] Where: V measured is the ultrasonic propagation speed under the set temperature and set humidity conditions; α and β are the compensation coefficients for temperature and humidity respectively; ΔT is the change in the current temperature relative to the set temperature; ΔH is the change in the current humidity relative to the set humidity; V corrected is the compensated ultrasonic propagation speed.

[0154] 3. Feature extraction and data fusion algorithm

[0155] The preprocessed echo signal and optical image need to extract features for subsequent roughness calculation and estimation. The ways of feature extraction include waveform feature extraction, texture feature extraction, and contour feature extraction.

[0156] Waveform feature extraction: Extract waveform features from the echo signal, mainly including signal amplitude, time delay, and peak value.

[0157] Texture feature extraction: By performing texture analysis on the optical image, extract the local structure information of the image, including the mean, variance, and contrast of the image. Contour feature extraction: Extract the object contour through Canny edge detection to obtain accurate boundary information from the image.

[0158] Data fusion: Fusion of the echo signal, optical image, and the extracted feature data through weighted average.

[0159] 4. Roughness parameter estimation process using a deep neural network

[0160] A deep neural network (DNN) is used for the estimation of roughness parameters. The main process is as follows:

[0161] Data input: The data after feature extraction is input into the deep neural network. The network structure includes an input layer, a hidden layer, and an output layer. The input layer receives the feature data after the fusion of the echo signal and the optical image.

[0162] Network training: The network is trained with known roughness sample data. The backpropagation algorithm is used to minimize the error between the roughness prediction value and the true value.

[0163] Roughness estimation: After training, the network is used to predict the roughness of new feature data and output the predicted roughness parameters.

[0164] Specifically, the roughness estimation module based on the neural network estimates roughness using the following method:

[0165] For the sampling interval j, assume there are N sampling points. For the sampling point i, where i = 1, 2,..., N, its data fusion feature is denoted as F i ';

[0166] The data fusion feature F i ' of the sampling point i in the sampling interval j is input into the roughness estimation module based on the neural network. The roughness estimation module based on the neural network further extracts features from the data fusion feature F i ' to obtain the feature vector F i ; The mean of the feature vectors of the N sampling points is The roughness estimation value R of the sampling interval j is obtained using the following formula j :

[0167]

[0168] Thus, the roughness estimation of the sampling interval is realized.

[0169] Such as Figure 4 , in underground engineering, the measurement of the wall roughness is carried out by a mobile platform for sampling, combined with non-contact measurement technologies such as acoustic wave reflection, lidar, and structured light. The specific process is as follows:

[0170] The system conducts automated sampling in underground caverns through an autonomous mobile platform. The mobile platform is equipped with various sensors (such as ultrasonic sensors, lidar scanning units, and structured light cameras) to continuously measure the wall roughness. The sensors collect data in real-time and store it in the local cache. The collected data is preliminarily processed by the built-in intelligent signal processing unit, including data filtering, feature extraction, and preliminary roughness estimation, and the measurement accuracy is improved using sensor fusion algorithms.

[0171] After the calculation is completed, the data is uploaded to the cloud analysis platform through the wireless communication module. The data upload process is carried out at preset time intervals or sampling periods. The data on the mobile platform is compressed and encrypted and then uploaded through a secure communication protocol. All uploaded data is encrypted to ensure that it is not tampered with or leaked during transmission. The encryption method usually uses the AES encryption algorithm. The data upload uses the common communication protocol MQTT protocol to ensure low-latency and high-performance message delivery. The MQTT protocol uses the publish / subscribe mode to ensure that data can be efficiently transmitted between various system nodes. According to the preset upload period or trigger condition, the mobile platform uploads the data to the cloud analysis platform, and the cloud analysis platform receives and stores the data through the API interface.

[0172] After receiving the uploaded data, the cloud analysis platform stores the uploaded raw data through a distributed storage system. After the data storage is completed, the cloud analysis platform performs preliminary formatting processing, extracts useful feature information, and eliminates invalid or abnormal data. Subsequently, a three-dimensional model of the underground cavern wall is generated based on the measurement data, and the wall roughness is calculated based on this model. The three-dimensional point cloud data is jointly constructed based on the depth information of structured light and lidar, and the Poisson reconstruction algorithm is used for wall surface fitting and topological reconstruction, specifically expressed as:

[0173] S(x,y,z)=f(P laser ,D structured )

[0174] where P laser is the lidar point cloud data, D structured is the depth image obtained by structured light, and S(x, y, z) is the wall model function after three-dimensional fitting.

[0175] Based on the three-dimensional reconstruction model, the system calculates the standard roughness parameters, using the average profile deviation (Ra) as the characterization index. The calculation method is as follows: Analyze the three-dimensional reconstruction model of the underground cavern wall. For each sampling interval j, m wall curves are determined at intervals along the height direction. For each wall curve k, k = 1, 2,..., m, the wall curve function f is read from the three-dimensional reconstruction model of the underground cavern wall k(x), where x is the wall depth direction. Using the following formula, the roughness value R of the wall curve k is obtained a,k :

[0176]

[0177] where L is the length of the sampling interval; is the roughness estimate of the sampling interval j, equal to R j ;

[0178] Thus, m roughness values of the sampling interval j are obtained and combined to form the roughness representation of the sampling interval j.

[0179] In addition, the system introduces a time series trend prediction algorithm, fuses historical measurement data, and uses a sliding window convolution and long short-term memory network (LSTM) model to predict and warn the roughness change trend of the upcoming construction area, and finally generates a three-dimensional roughness model with centimeter-level accuracy. After being exported in a standard format, this model can be used for construction planning optimization, quality assessment, and comparison in later re-inspection.

[0180] The present invention provides an intelligent measurement system and method for the roughness of the wall surface of an underground cavern based on multi-sensor fusion, having the following characteristics:

[0181] (1) The signal processing unit denoises the data of ultrasonic, lidar, and structured light sensors through a sensor fusion algorithm and generates high-precision roughness information of the wall surface of the underground cavern. (2) The measuring device is of a portable design, can be flexibly deployed in the underground engineering scenario, and has high measurement accuracy. (3) The system can be applied in a variety of underground engineering environments, including underground factories, tunnels, mines, etc., to meet different measurement requirements. (4) The measurement results can be uploaded to the cloud platform in real time for remote experts to analyze and make decisions, supporting the real-time monitoring and management of construction quality. (5) The system can perform high-precision roughness measurement in complex environments, reduce the dependence of traditional measurement methods on the environment, and improve the measurement efficiency and accuracy. (6) The measurement system has an adaptive ability and can adjust measurement parameters / measurement modules according to different underground environmental conditions to ensure effective application in various complex scenarios.

[0182] The present invention provides an intelligent measurement system and method for the roughness of the wall surface of an underground cavern based on multi-sensor fusion, which is a system for intelligent measurement of wall characteristic parameters in an underground space environment, especially a multi-sensor fusion roughness measurement system based on acoustic wave reflection, fusing lidar and structured light technologies, suitable for non-contact and high-precision measurement of the roughness of the cavern wall surface. This system is widely used in underground engineering fields such as underground factories, pumped storage power stations, tunnels, mines, etc., aiming to improve the accuracy of construction quality management and ventilation fluid simulation, and has important engineering practical value.

[0183] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent measurement system for the roughness of the wall surface of underground chambers based on multi-sensor fusion, characterized in that, It includes a multi-sensor fusion device, an environmental parameter acquisition sensor, an intelligent signal processing unit, a power management system, and a cloud analysis platform; The multi-sensor fusion device integrally installs an ultrasonic sensor, a lidar scanning unit, and a structured light camera; the ultrasonic sensor, the lidar scanning unit, and the structured light camera are used to synchronously sample each sampling interval of the underground chamber wall surface, and respectively and real-time measure the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface of each sampling interval, and upload them to the intelligent signal processing unit in real time; The environmental parameter acquisition sensor is used to acquire the environmental parameters of the underground chamber in real time and upload them to the intelligent signal processing unit in real time; The intelligent signal processing unit includes a signal preprocessing module, an environmental parameter compensation module, a feature extraction module, a data fusion module, and a roughness estimation module based on a neural network; The signal preprocessing module is used to respectively perform data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface of each sampling interval to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface; The environmental parameter compensation module is used to perform environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data; The feature extraction module is used to respectively perform feature extraction on the compensated ultrasonic detection distance data, the preprocessed three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface to obtain the ultrasonic detection distance waveform feature, the three-dimensional point cloud feature of the underground chamber wall surface, and the three-dimensional image feature of the underground chamber wall surface; The data fusion module is used to perform data fusion on the ultrasonic detection distance waveform feature, the three-dimensional point cloud feature of the underground chamber wall surface, and the three-dimensional image feature of the underground chamber wall surface to obtain the data fusion feature corresponding to the sampling interval; The roughness estimation module based on a neural network is used to perform roughness estimation on the data fusion feature corresponding to the sampling interval, output the roughness estimation value corresponding to the sampling interval, and send it to the cloud analysis platform; The cloud analysis platform is used to comprehensively analyze the roughness estimation values corresponding to each sampling interval to generate a three-dimensional roughness model of the underground chamber wall surface.

2. The intelligent measurement system for the roughness of the underground chamber wall surface based on multi-sensor fusion according to claim 1, wherein The multi-sensor fusion device adopts a bracket deployment method, specifically using an adjustable pan-tilt bracket to realize the adjustment of vertical, pitch, and horizontal angles, and realize the accurate measurement of each sampling interval by the ultrasonic sensor, the lidar scanning unit, and the structured light camera; the pan-tilt of the adjustable pan-tilt bracket adopts a three-degree-of-freedom electric pan-tilt, and the cloud analysis platform uses a pan-tilt control unit to remotely control and adjust the azimuth angle, pitch angle, and scanning path of the three-degree-of-freedom electric pan-tilt; The ultrasonic sensor is arranged at the middle and lower part of the adjustable pan-tilt bracket, vertically pointing to the wall surface of the current sampling interval, and measures the undulation and unevenness of each sampling point in the sampling interval by collecting the time difference between ultrasonic emission and ultrasonic reception after reflection and the ultrasonic intensity characteristics. The lidar scanning unit is arranged at the top center of the adjustable pan-tilt bracket and acquires the point cloud contour data of the sampling interval by using a rotating scanning mode. The structured light camera is arranged below the adjustable pan-tilt bracket and generates a high-precision depth map and texture image of the sampling interval through active projection of fringe light coding and binocular vision principle.

3. An intelligent measurement system for the wall roughness of underground chambers based on multi-sensor fusion according to claim 1, characterized in that, The environmental parameter acquisition sensors include a temperature and humidity sensor, a barometer, and a dust sensor, which are respectively used to acquire the temperature and humidity, air pressure, and dust concentration in the underground cavern.

4. An intelligent measurement system for the wall roughness of an underground chamber based on multi-sensor fusion according to claim 1, characterized in that, The signal preprocessing module is specifically used to perform noise suppression and blind source separation on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground cavern wall surface, and the three-dimensional image data of the underground cavern wall surface.

5. An intelligent measurement system for the wall roughness of underground chambers based on multi-sensor fusion according to claim 1, characterized in that, The environmental parameter compensation module compensates the ultrasonic propagation speed by using the following formula: V corrected = V measured ×(1 + αΔT + β△H) Where: V measured is the ultrasonic wave propagation speed under the set temperature and humidity conditions; α and β are the compensation coefficients of temperature and humidity respectively; ΔT is the change in the current temperature relative to the set temperature; ΔH is the change in the current humidity relative to the set humidity; V corrected is the compensated ultrasonic wave propagation speed; The compensated ultrasonic wave propagation velocity V corrected , based on the time difference between the ultrasonic wave emission and the reception of the reflected ultrasonic wave, the ultrasonic wave detection distance data is obtained, which is the compensated ultrasonic wave detection distance data.

6. The intelligent measurement system for the wall roughness of an underground chamber based on multi-sensor fusion according to claim 1, characterized in that, The feature extraction module is specifically used for: extracting waveform features from the compensated ultrasonic detection distance data, including signal amplitude, time delay, and peak value, to obtain the ultrasonic detection distance waveform features; extracting waveform features from the preprocessed three-dimensional point cloud data of the underground cavern wall surface, including signal amplitude, time delay, and peak value, to obtain the three-dimensional point cloud features of the underground cavern wall surface; performing texture analysis on the preprocessed three-dimensional image data of the underground cavern wall surface, extracting the local structure information of the image, including the mean value, variance, and contrast of the image; then extracting the object contour through edge detection; the local structure information of the image and the object contour form the three-dimensional image features of the underground cavern wall surface.

7. An intelligent measurement system for the roughness of the underground chamber wall surface based on multi-sensor fusion according to claim 1, characterized in that, The data fusion module is specifically used for: performing weighted averaging on the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground cavern wall surface, and the three-dimensional image features of the underground cavern wall surface to obtain the data fusion features corresponding to the sampling interval.

8. An intelligent measurement system for the roughness of the wall surface of an underground chamber based on multi-sensor fusion according to claim 1, wherein, The neural network-based roughness estimation module estimates roughness by using the following method: For the sampling interval j, assume there are N sampling points. For the sampling point i, where i = 1, 2,..., N, its data fusion feature is represented as F i '; The data fusion feature F of the sampling point i in the sampling interval j i is input into the neural network-based roughness estimation module, and the neural network-based roughness estimation module further extracts features from the data fusion feature F i to obtain a feature vector F i ; the mean value of the feature vectors of N sampling points is Using the following formula, the roughness estimation value R of the sampling interval j is obtained j : Thus, the roughness estimation of the sampling interval is realized.

9. An intelligent measurement system for the wall roughness of underground chambers based on multi-sensor fusion according to claim 1, characterized in that, The cloud analysis platform is specifically used for: According to the preprocessed three-dimensional point cloud data of the underground cavern wall surface and the three-dimensional image data of the underground cavern wall surface, combined with the roughness estimation value of each sampling interval, performing surface fitting and topological reconstruction on the underground cavern wall surface to obtain a three-dimensional reconstruction model of the underground cavern wall surface; Analyze the three-dimensional reconstruction model of the underground chamber wall surface. For each sampling interval j, m wall curves are determined at intervals along the height direction. For each wall curve k, where k = 1, 2,..., m, the wall curve function f k (x) is read from the three-dimensional reconstruction model of the underground chamber wall surface, where x is the depth direction of the wall surface. Using the following formula, the roughness value R of wall curve k is obtained a,k : where L is the length of the sampling interval; is the roughness estimate of the sampling interval j and is equal to R j ; Thus, m roughness values of the sampling interval j are obtained, and they are combined to form the roughness representation of the sampling interval j.

10. A measurement method of an intelligent measurement system for the wall roughness of an underground chamber based on multi-sensor fusion according to any one of claims 1 to 9, characterized in that, Including the following steps: Step S1, for each sampling interval of the underground cavern wall surface, the ultrasonic sensor, the lidar scanning unit, and the structured light camera perform synchronous sampling, and respectively and real-time measure the ultrasonic detection distance data, the three-dimensional point cloud data of the underground cavern wall surface, and the three-dimensional image data of the underground cavern wall surface of each sampling interval, and upload them to the intelligent signal processing unit in real time. Meanwhile, the environmental parameter acquisition sensor collects the environmental parameters of the underground chamber in real time and uploads them to the intelligent signal processing unit in real time; Step S2, the intelligent signal processing unit performs data preprocessing on the ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface in each sampling interval to obtain the preprocessed ultrasonic detection distance data, the three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface; Step S3, the intelligent signal processing unit performs environmental parameter compensation on the preprocessed ultrasonic detection distance data to obtain the compensated ultrasonic detection distance data; Step S4, the intelligent signal processing unit respectively extracts features from the compensated ultrasonic detection distance data, the preprocessed three-dimensional point cloud data of the underground chamber wall surface, and the three-dimensional image data of the underground chamber wall surface to obtain the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall surface, and the three-dimensional image features of the underground chamber wall surface; Step S5, the intelligent signal processing unit performs data fusion on the ultrasonic detection distance waveform features, the three-dimensional point cloud features of the underground chamber wall surface, and the three-dimensional image features of the underground chamber wall surface to obtain the data fusion features corresponding to the sampling interval; Step S6, the intelligent signal processing unit uses a roughness estimation module based on a neural network to estimate the roughness of the data fusion features corresponding to the sampling interval, outputs the roughness estimation value corresponding to the sampling interval, and sends it to the cloud analysis platform; Step S7, the cloud analysis platform comprehensively analyzes the roughness estimation values corresponding to each sampling interval to generate a three-dimensional roughness model of the underground chamber wall surface.

Images