Estimation method and system for shield spoil volume based on image processing

Abstract

The application discloses a shield muck quantity estimation method and system based on image processing, and particularly relates to the technical field of tunnel engineering monitoring, which obtains muck RGB and depth images through an image acquisition device, generates high-precision three-dimensional point cloud data in combination with laser scanning and multispectral technology, calculates the muck volume by using a slicing method, and dynamically generates a calibration factor by fusing belt scale mass flow and moisture content data, so that the volume estimation result is corrected in real time. The system comprises image acquisition, point cloud processing, volume calculation and dynamic calibration modules, has the characteristics of automation, high precision and real-time feedback, effectively solves the problems of large error and slow response of traditional manual measurement and single weighing mode, and is suitable for continuous and accurate monitoring of muck quantity in the shield construction process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel engineering monitoring, more particularly, to a shield muck quantity estimation method and system based on image processing. BACKGROUND

[0002] In shield tunnel construction, real-time and accurate monitoring of muck quantity is the core basis for judging the matching of shield tunneling parameters, the stability of the stratum and the control of construction cost, and is directly related to construction safety and engineering efficiency. At present, the mainstream muck quantity estimation method in the industry mainly relies on traditional manual measurement and single mechanical weighing: manual counting of muck trucks combined with experience to estimate the quantity, which is affected by factors such as uneven loading of muck trucks, human counting errors, etc., and the precision is generally low, and the efficiency is low, which cannot meet the real-time monitoring needs of continuous tunneling; although single belt scale weighing can obtain mass data, it needs to rely on fixed muck density to convert the quantity, but the muck density in shield construction is greatly affected by stratum lithology and water content, and the problems of belt tension change and material partial load further reduce the weighing accuracy, which is difficult to directly meet the core needs of quantity measurement in engineering.

[0003] In recent years, some technologies have tried to introduce laser scanning or image processing technology for volume estimation, but there are still significant technical shortcomings. In the point cloud data processing level, the existing scheme mostly uses fixed thickness slicing method to calculate the volume, without considering the influence of belt speed dynamic change on the point cloud density: when the belt speed increases, the number of point clouds in unit width is insufficient, the data representativeness in the slice is poor, leading to a sharp increase in volume calculation error; in terms of environmental adaptability, the laser scanning system is easily disturbed by the construction scene, and the mirror reflection generated by strong light, dust and wet and slippery surface of muck will seriously submerge the laser stripe signal, and the traditional image extraction algorithm is difficult to separate the effective signal from the noise, causing distortion of the three-dimensional point cloud reconstruction; in addition, such technologies have not established a correlation mechanism between material characteristics and estimation results, and lack of response to the dynamic changes of muck loose coefficient, water content and other physical parameters, and the precision decays obviously after long-term operation.

[0004] In summary, the existing shield muck quantity estimation technology has not formed a complete system of "high-precision perception-strong environmental adaptability-dynamic calibration": traditional methods sacrifice precision for convenience, single sensing technology cannot balance environmental robustness and data reliability, and early image / laser schemes lack adaptive processing and closed-loop calibration capabilities, which cannot meet the long-term monitoring needs in complex construction scenes. With the development of shield construction towards long distance and large diameter, higher requirements are put forward for the real-time, precision and stability of muck quantity monitoring, and the defects of existing technologies have become a key bottleneck restricting the intelligent control of construction, and a new estimation technology integrating multi-source perception, anti-interference processing and dynamic calibration is urgently needed to fill the gap. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide a shielded muck volume estimation method and system based on image processing.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] The shielded muck volume estimation method based on image processing comprises the following steps:

[0008] S1, device deployment and parameter calibration: install an image acquisition device at the shielded belt conveyor or the muck discharge port area, regularly calibrate the device to obtain parameters, and demarcate a detection area based on a background image, determine a pixel corresponding real area basis calculation parameter;

[0009] S2, image acquisition and point cloud processing: the device acquires real-time RGB images and depth images of the muck, generates three-dimensional point cloud data through dual-mode feature fusion or multi-view three-dimensional reconstruction technology, removes noise and irrelevant objects through semantic segmentation, and then obtains a muck point cloud model through filtering and point cloud fusion processing;

[0010] S3, volume calculation and dynamic calibration: adopt a slicing method or a triangular subdivision method to calculate the volume of the muck point cloud model, combine the belt speed parameter to accumulate to obtain real-time muck discharge volume; at the same time, fuse the belt scale mass flow and multi-spectral water content data to generate a calibration factor, and dynamically correct the volume estimation result to improve the accuracy.

[0011] Specifically, the process of S1 is as follows:

[0012] System arrangement and synchronization: install a reciprocating swing line laser generator above the muck belt conveyor along the belt transversely; fixedly install an industrial camera on one side of the laser, and the field of view covers the entire area scanned by the laser; the laser and the camera are synchronously triggered by the same controller;

[0013] Dynamic scanning and image acquisition: control the line laser to scan along the belt transversely at a constant angular velocity, and the camera continuously acquires images at a fixed frame rate; each frame of image captures the profile curve of a position of the laser line on the muck and the belt surface; in the scanning period, the laser beam scans the entire belt width, thereby obtaining continuous muck cross-sectional profiles;

[0014] Profile center line extraction and three-dimensional coordinate calculation: for each frame of acquired image, the center line of the laser stripe is extracted using the Steger algorithm based on gradient; according to the pre-calibrated swing angle of the laser, the camera intrinsic parameter, and the extrinsic parameter of the laser plane and the camera, each pixel point in the image is accurately converted into a three-dimensional coordinate (X, Y, Z) in the world coordinate system through optical triangulation method; all frames of three-dimensional point clouds are fused to obtain a complete three-dimensional muck surface data;

[0015] Volume calculation: A three-dimensional planar model of the empty conveyor belt is pre-calibrated as a reference plane; the three-dimensional point cloud of the slag obtained by scanning is compared with the reference plane; the point cloud is preprocessed before using the slicing method;

[0016] The volume is calculated using the slicing method: the point cloud is sliced ​​into multiple thin slices along the conveyor belt's forward direction, the cross-sectional area of ​​the slag in each slice is calculated, multiplied by the slice thickness, and finally the volumes of all slices are summed to obtain the slag volume within that scanning cycle. .

[0017] Specifically, the preprocessing of the point cloud in S1 is as follows:

[0018] Density adaptive slicing: based on the current belt speed Units and total number of scan points per scan Dynamically calculate slice thickness The formula is:

[0019]

[0020] in, The width of the belt transverse scan. The target point cloud density is a preset empirical constant value.

[0021] Statistical outlier filtering: along the belt travel direction, using thickness Divide the point cloud into multiple slices; for the point cloud within each slice, calculate the mean height of all points from the reference plane. and standard deviation ; Remove height values ​​that are not present Outliers within the range These are preset coefficients.

[0022] Specifically, the process of S2 is as follows:

[0023] Deployment of polarization optics system: A linear polarizer, i.e., a polarizer, is installed in front of the online laser generator; another linear polarizer, i.e., a polarizer, is installed in front of the camera lens. The polarization direction of the polarizer is 90° with the polarizer direction, forming an orthogonal polarization system.

[0024] Image acquisition and specular removal: The orthogonal polarization system suppresses ambient light from the diffuse reflective surface of the slag, but allows specular reflection light generated by the laser itself on the smooth surface of the slag to pass through; therefore, in the image acquired by the camera, the laser stripe signal stands out, and the laser stripe signal contains both diffuse and specular reflection components, while the ambient background light is suppressed; subsequently, the image is subjected to thresholding and morphological filtering to separate the complete laser stripe.

[0025] Multispectral background recognition: In the bypass of the laser scanning system, a non-visible light band, namely near-infrared NIR, is added, and a fixed light source and a corresponding filter camera are added, and at the same time, a red waveband Red light source and camera are added, or a multispectral camera with dual-waveband synchronous acquisition capability is used;

[0026] Due to the significant difference in absorption rate of water and slag soil to near-infrared and red waveband, the gray response of wet slag soil and dry slag soil in different waveband images is different; in order to eliminate the influence of environmental light change and realize quantitative monitoring of water content, standard processing is carried out.

[0027] Specifically, the standard processing process in S2 is as follows:

[0028] Synchronous acquisition and regional registration: Synchronously acquire NIR image and Red image of the same slag soil region, and ensure that the two images are pixel-level registered;

[0029] Calculate slag soil moisture index: In the slag soil target region, calculate the average gray value of the near-infrared waveband and the average gray value of the red waveband The ratio of the two values is defined as the slag soil moisture index SMI: ;

[0030] Output quantitative parameters: The strong absorption characteristics of water in the NIR waveband are higher than those in the Red waveband, and the higher the water content ω of the slag soil is, The value decreases by more than , resulting in a decrease in SMI index;

[0031] Through pre-experiment, a calibration curve of SMI and water content ω is established , which converts image information into a quantitative physical parameter; SMI value or calculated water content ω is used as an input parameter, and is output to a subsequent volume calculation calibration unit.

[0032] Specifically, the specific process of S3 is as follows:

[0033] Belt scale integration and data synchronization: A belt scale is installed on the return section of the slag soil belt conveyor to measure the weight of the belt per unit length in real time, and the net weight and speed of the belt are calculated; a position sensor is installed near the belt scale to mark the position of each section of slag soil;

[0034] Mass flow calculation and correlation: According to the belt speed v and the calculated volume , the volume flow based on volume is calculated: is the corresponding time interval for calculating the volume flow; at the same time, the belt scale measures the mass flow of slag soil in the corresponding time period, i.e. the mass of slag soil passing per unit time;

[0035] Real-time calibration factor calculation: Based on the principle of conservation of mass, the following relationship exists: ,in It is the instantaneous density of the construction waste; the real-time calibration factor is calculated. ; The value reflects the apparent density of the current slag and soil, including the effects of the bulk density and moisture content;

[0036] Dynamic feedback calibration: The calculated real-time calibration factor k is fed back to the volume calculation unit; subsequent volume estimates will no longer be directly used. Instead, the calibrated volume is used. ,in It is the initial calibrated baseline density; through closed-loop feedback, it dynamically compensates for volume estimation errors caused by changes in the physical properties of the slag.

[0037] Meanwhile, the multispectral moisture content signal serves as an auxiliary input for the calibration factor. The changing trends are predicted and weighted.

[0038] Specifically, in S3, the calibration factor... The process of predicting and weighting the changing trends is as follows:

[0039] Multi-source calibration factor generation: Generates two calibration factors simultaneously.

[0040] Mass flow rate based on belt scale With volumetric flow The calculated calibration factor, after recursive filtering, has an estimated variance of... ;

[0041] Optical calibration factor based on multispectral moisture index (SMI) prediction; a regression model between SMI and calibration factor is established using historical data, and the variance of the model predictions is determined based on historical error statistics. ;

[0042] Optimal weighted fusion: the final calibration factor Not a choice or Instead, it performs optimal weighted fusion based on the reciprocal of the variances of the two variances, and the calculation formula is as follows: ,in , The optimal weight;

[0043] Volume calibration: Subsequent volume estimates will be calculated using the fused optimal calibration factor. , This is the initial reference density.

[0044] The shield tunneling muck volume estimation system based on image processing includes the following modules:

[0045] The image acquisition module is used to deploy image acquisition equipment in the shield tunnel conveyor or slag outlet area to acquire RGB and depth images of the slag in real time, providing raw visual data for subsequent 3D point cloud generation.

[0046] The point cloud processing module is used to reconstruct three-dimensional point clouds from the acquired images. Through semantic segmentation, filtering, and point cloud fusion preprocessing operations, noise and irrelevant ground features are removed to obtain the slag and soil point cloud model.

[0047] The volume calculation module is used to calculate the volume of the pre-processed slag point cloud using the slicing method, and combined with the belt speed parameters, to obtain the slag volume within the scanning cycle.

[0048] The dynamic calibration module is used to integrate the mass flow rate and multispectral moisture content data measured in real time by the belt scale, generate a real-time calibration factor and feed it back to the volume calculation unit to dynamically correct the volume estimation results and compensate for the errors caused by changes in the physical properties of the slag.

[0049] The technical effects and advantages of this invention are as follows:

[0050] This technology effectively improves the basic accuracy of waste soil volume estimation. Through density-adaptive slicing technology, the slice thickness can be dynamically adjusted based on conveyor belt speed and the total number of scanning points, ensuring sufficient point cloud data within each slice. Combined with statistical outlier filtering to remove noise points, it significantly optimizes the quality of the 3D point cloud. Compared to traditional fixed-slice methods and manual estimation, this reduces errors caused by speed fluctuations and point cloud noise from the data source, laying the foundation for accurate volume calculation.

[0051] Enhance the system's adaptability to complex construction environments. The orthogonal polarization system can effectively suppress ambient light interference and highlight the laser stripe signal, stably extracting effective contours even in strong light or slippery soil conditions; multispectral imaging combined with moisture index (SMI) calculation enables quantitative monitoring of moisture content while further avoiding the impact of light changes on the data, solving the problem of poor environmental robustness of traditional laser / imaging technologies.

[0052] This system enables dynamic closed-loop calibration and accurate output of estimation results. It integrates calibration factors generated from belt scale mass flow rate and multispectral prediction, obtaining a high-confidence final factor through optimal weighted fusion, and dynamically correcting the volume estimate. This mechanism compensates for the influence of changes in physical properties such as soil density and moisture content, making the results closer to the actual volume and meeting the high-precision requirements for cost control and soil monitoring during construction. Attached Figure Description

[0053] Figure 1 This is a flowchart of the method of the present invention;

[0054] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, the method for estimating the volume of tunnel boring machine excavation based on image processing includes the following steps:

[0057] Step 1: Equipment deployment and parameter calibration. Install image acquisition equipment such as cameras (multi-view cameras or depth cameras) in the shield tunnel conveyor or slag outlet area. Regularly calibrate the equipment to obtain key parameters such as internal parameters, and delineate the detection area and determine the basic calculation parameters such as the actual area corresponding to the pixels based on the background image.

[0058] System Layout and Synchronization: A reciprocating linear laser generator is installed above the slag conveyor belt, transverse to the belt (perpendicular to the belt's forward direction). A high-speed industrial camera is fixedly installed to one side of the laser, ensuring its field of view covers the entire area scanned by the laser. The laser and camera are synchronously triggered by the same controller.

[0059] Dynamic scanning and image acquisition: The control line laser scans laterally along the conveyor belt at a constant angular velocity, while the camera continuously acquires images at a fixed frame rate. Each frame captures the profile curve of the laser line at a specific location on the surface of the slag and conveyor belt. During a complete scanning cycle, the laser beam sweeps across the entire width of the conveyor belt, thus obtaining dozens to hundreds of continuous cross-sectional profiles of the slag.

[0060] Centerline extraction and 3D coordinate calculation: For each frame of the acquired image, the centerline of the laser stripes is extracted with sub-pixel precision using the gradient-based Steger algorithm. Based on the pre-calibrated laser oscillation angle (precisely corresponding to the camera frame number), camera intrinsic parameters, and the relative extrinsic parameters between the laser plane and the camera, each pixel in the image is precisely converted into 3D coordinates (X, Y, Z) in the world coordinate system using optical triangulation. The 3D point clouds of all frames are then fused to obtain a complete 3D data frame of the slag surface.

[0061] Volume calculation: A pre-calibrated 3D planar model of the empty conveyor belt is used as a reference plane. The 3D point cloud of the slag obtained from the scan is compared with the reference plane.

[0062] Before using the slicing method, the point cloud is preprocessed as follows;

[0063] Density adaptive slicing: based on the current belt speed unit:( ) and the total number of scan points in a single scan Dynamically calculate slice thickness (unit: The formula is: ,in, Horizontal scan width of the belt (unit: ), The target point cloud density (unit: points / square meter) is a preset empirical constant value. This formula ensures that each slice has a sufficient order of magnitude of point cloud data for calculation, avoiding a decrease in calculation accuracy due to speed variations.

[0064] Statistical outlier filtering: along the belt travel direction, using thickness The point cloud is divided into multiple slices. For the point cloud within each slice, the mean height of all points from the reference plane is calculated. and standard deviation Remove height values ​​that are not present. Outliers within the range ( (This is a preset coefficient, with a value of 2 or 3). These points are mostly noise points or non-representative slag and soil.

[0065] The volume is calculated using the slicing method: the point cloud is sliced ​​into multiple thin slices along the conveyor belt's forward direction, the cross-sectional area of ​​the slag in each slice is calculated, multiplied by the slice thickness, and finally the volumes of all slices are summed to obtain the slag volume within that scanning cycle. .

[0066] Step 2: Image acquisition and point cloud processing. The equipment acquires RGB and depth images of the slag in real time. It generates three-dimensional point cloud data through dual-modal feature fusion or multi-view 3D reconstruction technology. After semantic segmentation to remove noise and irrelevant features, it obtains a high-precision slag point cloud model through filtering, point cloud fusion and other processing.

[0067] Polarization optics system deployment: A linear polarizer (polarizer) is installed in front of the laser generator. Another linear polarizer (analyzer) is installed in front of the camera lens, with its polarization direction at 90° (orthogonal) to the polarizer direction. This arrangement constitutes an orthogonal polarization system.

[0068] Image Acquisition and Highlight Removal: The orthogonal polarization system effectively suppresses ambient light from the diffuse reflective surface of the construction waste, but allows specular reflection light generated by the laser itself on the smooth surface of the waste to pass through. Therefore, in the image acquired by the camera, the laser stripe signal (containing both diffuse and specular reflection components) is extremely prominent, while the ambient background light is greatly suppressed. Subsequently, thresholding and morphological filtering of the image can easily separate the complete laser stripes, even under strong ambient light or when the construction waste is slippery and produces highlights.

[0069] Multispectral background identification: A fixed light source and corresponding filter camera in a non-visible light band (e.g., near-infrared, NIR) are added as a bypass of the laser scanning system. Simultaneously, a red band light source and camera are added, or a multispectral camera with dual-band simultaneous acquisition capability is used. Because water and construction waste exhibit significant differences in their absorption rates to the near-infrared (NIR) and red (Red) bands, wet and dry construction waste show significant differences in grayscale in NIR images, resulting in different grayscale responses in different bands. To eliminate the influence of ambient light variations and achieve quantitative monitoring of moisture content, standard processing is performed as follows:

[0070] Synchronous acquisition and regional registration: Simultaneously acquire NIR and Red images of the same waste soil area, ensuring pixel-level registration between the two images; Calculate the waste soil moisture index (SMI): Calculate the average gray value in the near-infrared band within the target area of ​​the waste soil. Compared with the average gray value of the red band The ratio of the two values ​​is defined as the soil moisture index (SMI). ;

[0071] Output quantitative parameters: Water's strong absorption characteristics are much higher in the NIR band than in the Red band. Therefore, the higher the moisture content (ω) of the slag soil, The value decreased by a much larger margin than This leads to a decrease in the SMI index. A calibration curve of SMI versus moisture content ω was established through preliminary experiments. This allows the image information to be converted into a reliable, quantitative physical parameter. The SMI value or the calculated water content ω serves as a precise input parameter, which is then output to the subsequent volume calculation and calibration unit.

[0072] Step 3: Volume calculation and dynamic calibration. The volume of the slag point cloud model is calculated using the slicing method or triangulation method. The real-time soil output is obtained by accumulating parameters such as belt speed. At the same time, the belt scale mass flow rate and multispectral moisture content data are integrated to generate calibration factors and dynamically correct the volume estimation results to improve accuracy.

[0073] High-precision belt scale integration and data synchronization: A belt scale is installed on the return section of the construction waste belt conveyor (where there is no construction waste underneath) to measure the weight of a unit length of belt in real time, thereby calculating the net weight and speed of the belt. A position sensor (such as a photoelectric encoder) is installed near the belt scale to accurately mark the position of each section of construction waste.

[0074] Mass flow rate calculation and correlation: based on the belt speed v and the volume calculated in step 1.4. Calculate the volumetric flow rate based on volume. Meanwhile, the belt scale measures the "mass flow rate" of the construction waste in real time for the corresponding time period. (The mass of construction waste passing through per unit time).

[0075] Real-time calibration factor calculation: Based on the principle of conservation of mass, a theoretical relationship exists: ,in This is the instantaneous density of the construction waste. Therefore, a real-time "calibration factor" can be calculated. . The value essentially reflects the apparent density of the current slag (including the effects of the bulking coefficient and moisture content).

[0076] Dynamic feedback calibration: The calculated real-time calibration factor k is fed back to the volume calculation unit. Subsequent volume estimates will no longer be directly used. Instead, the calibrated volume is used. ,in This is the initial reference density calibrated by the system. Through this closed-loop feedback, the volume estimation error caused by changes in the physical properties of the slag is dynamically compensated, making the final output value infinitely close to the equivalent volume after conversion from the actual mass. Simultaneously, the multispectral moisture content signal can be used as an auxiliary input to adjust the calibration factor. The changing trends are predicted and weighted to make the calibration process smoother and faster; the specific process is as follows:

[0077] Multi-source calibration factor generation: Generates two calibration factors simultaneously.

[0078] Mass flow rate based on belt scale With volumetric flow The calculated calibration factor, after undergoing recursive filtering (see Optimization Point 1), has an estimated variance of: ; Optical calibration factor based on multispectral moisture index (SMI) prediction. A regression model (e.g., a linear model) between SMI and calibration factor is established using historical data. The variance of the predicted values ​​is determined based on historical error statistics. ;

[0079] Optimal weighted fusion: the final calibration factor It's not a simple choice or Instead, it performs optimal weighted fusion based on the reciprocal of the variances of the two (i.e., precision or confidence level), and the calculation formula is as follows: ,in , Optimal weighting. This is significant because it automatically assigns higher weights to data sources with more accurate estimates (smaller variance); Volume calibration: Subsequent volume estimates will be calculated using the fused optimal calibration factor. , This is the initial reference density.

[0080] The shield tunneling muck volume estimation system based on image processing includes an image acquisition module, a point cloud processing module, a volume calculation module, and a dynamic calibration module.

[0081] The image acquisition module deploys image acquisition equipment (such as multi-view cameras and depth cameras) in the shield tunnel conveyor or slag discharge area to acquire RGB and depth images of the slag in real time, providing raw visual data for subsequent 3D point cloud generation.

[0082] The point cloud processing module performs 3D point cloud reconstruction on the acquired images. Through preprocessing operations such as semantic segmentation, filtering, and point cloud fusion, noise and irrelevant features are removed to obtain a high-precision point cloud model of the slag and soil.

[0083] The volume calculation module uses a slicing method (such as calculating the volume of each slice after density adaptive slicing and then summing them up) to calculate the volume of the pre-processed slag point cloud. Combined with parameters such as belt speed, the volume of slag within the scanning cycle is obtained.

[0084] The dynamic calibration module integrates the mass flow rate and multispectral moisture content data measured in real time by the belt scale, generates a real-time calibration factor and feeds it back to the volume calculation unit, dynamically corrects the volume estimation results, and compensates for errors caused by changes in the physical properties of the slag soil (such as the looseness coefficient and moisture content).

[0085] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0086] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0087] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0088] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0089] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0090] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0091] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0092] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0093] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for estimating the volume of tunnel boring machine excavation based on image processing, characterized in that, Includes the following steps: S1. Equipment deployment and parameter calibration: Install image acquisition equipment in the shield tunnel conveyor or slag outlet area, calibrate the equipment regularly to obtain parameters, and delineate the detection area based on the background image and determine the basic calculation parameters of the actual area corresponding to the pixel. S2. Image Acquisition and Point Cloud Processing: The equipment acquires RGB and depth images of the slag in real time, generates three-dimensional point cloud data through dual-modal feature fusion or multi-view 3D reconstruction technology, removes noise and irrelevant features through semantic segmentation, and then obtains the slag point cloud model through filtering and point cloud fusion processing. S3. Volume Calculation and Dynamic Calibration: The volume of the slag point cloud model is calculated using the slicing method or triangulation method, and the real-time soil discharge is obtained by accumulating the belt speed parameters. At the same time, the belt scale mass flow rate and multispectral moisture content data are integrated to generate calibration factors and dynamically correct the volume estimation results to improve accuracy. The process of S1 is as follows: System layout and synchronization: A reciprocating linear laser generator is installed above the slag conveyor belt, along the transverse direction of the belt; an industrial camera is fixedly installed on one side of the laser, with a field of view covering the entire area scanned by the laser; the laser and camera are synchronously triggered by the same controller. Dynamic scanning and image acquisition: The control line laser scans laterally along the belt at a constant angular velocity, while the camera continuously acquires images at a fixed frame rate; each frame captures the profile curve of the laser line at a position on the surface of the slag and the belt; within the scanning cycle, the laser beam sweeps across the entire width of the belt, thereby obtaining a continuous cross-sectional profile of the slag. Centerline extraction and 3D coordinate calculation: For each frame of the acquired image, the centerline of the laser stripes is extracted using the gradient-based Steger algorithm; based on the pre-calibrated laser swing angle, camera intrinsic parameters, and laser plane and camera extrinsic parameters, each pixel in the image is accurately converted into 3D coordinates (X,Y,Z) in the world coordinate system using optical triangulation; the 3D point clouds of all frames are fused to obtain a complete 3D slag surface data frame; Volume calculation: A three-dimensional planar model of the empty conveyor belt is pre-calibrated as a reference plane; the three-dimensional point cloud of the slag obtained by scanning is compared with the reference plane; the point cloud is preprocessed before using the slicing method; The volume is calculated using the slicing method: the point cloud is sliced ​​into multiple thin slices along the conveyor belt's forward direction, the cross-sectional area of ​​the slag in each slice is calculated, multiplied by the slice thickness, and finally the volumes of all slices are summed to obtain the slag volume within that scanning cycle. ; The specific process of preprocessing the point cloud in S1 is as follows: Density adaptive slicing: based on the current belt speed Units and total number of scan points per scan Dynamically calculate slice thickness The formula is: in, The width of the belt transverse scan. The target point cloud density is a preset empirical constant value. Statistical outlier filtering: along the belt travel direction, using thickness Divide the point cloud into multiple slices; for the point cloud within each slice, calculate the mean height of all points from the reference plane. and standard deviation ; Remove height values ​​that are not present Outliers within the range These are preset coefficients; The specific process of S3 is as follows: Belt scale integration and data synchronization: A belt scale is installed on the return section of the slag conveyor to measure the weight of the belt per unit length in real time and calculate the net weight and speed of the belt; a position sensor is installed near the belt scale to mark the position of each section of slag. Mass flow rate calculation and correlation: based on belt speed v and the calculated volume Calculate the volumetric flow rate based on volume. : , This is the time interval corresponding to the calculation of volumetric flow rate; simultaneously, the belt scale measures the mass flow rate of the slag and soil in real time for the corresponding time period. That is, the mass of construction waste passing through per unit time; Real-time calibration factor calculation: Based on the principle of conservation of mass, the following relationship exists: ,in It is the instantaneous density of the construction waste; the real-time calibration factor is calculated. ; The value reflects the apparent density of the current slag and soil, including the effects of the bulk density and moisture content; Dynamic feedback calibration: The calculated real-time calibration factor k is fed back to the volume calculation unit; subsequent volume estimates will no longer be directly used. Instead, the calibrated volume is used. ,in It is the initial calibrated baseline density; through closed-loop feedback, it dynamically compensates for volume estimation errors caused by changes in the physical properties of the slag. Meanwhile, the multispectral moisture content signal serves as an auxiliary input for the calibration factor. The changing trends are predicted and weighted.

2. The method for estimating the volume of tunnel boring machine excavation based on image processing according to claim 1, characterized in that, The specific process of S2 is as follows: Deployment of polarization optics system: A linear polarizer, i.e., a polarizer, is installed in front of the online laser generator; another linear polarizer, i.e., a polarizer, is installed in front of the camera lens. The polarization direction of the polarizer is 90° with the polarizer direction, forming an orthogonal polarization system. Image acquisition and specular removal: The orthogonal polarization system suppresses ambient light from the diffuse reflective surface of the slag, but allows specular reflection light generated by the laser itself on the smooth surface of the slag to pass through; therefore, in the image acquired by the camera, the laser stripe signal stands out, and the laser stripe signal contains both diffuse and specular reflection components, while the ambient background light is suppressed; subsequently, the image is subjected to thresholding and morphological filtering to separate the complete laser stripe. Multispectral background identification: In the bypass of the laser scanning system, a non-visible light band, namely near-infrared (NIR), is added, with a fixed light source and corresponding filter camera. At the same time, a red band light source and camera are added, or a multispectral camera with dual-band synchronous acquisition capability is used. Because water and slag have significantly different absorption rates in the near-infrared and red bands, wet and dry slag have different grayscale responses in images of different bands. To eliminate the influence of changes in ambient light and to achieve quantitative monitoring of moisture content, standard processing is performed.

3. The method for estimating the volume of tunnel boring machine excavation based on image processing according to claim 2, characterized in that, The standard processing procedure in S2 is as follows: Synchronous acquisition and regional registration: Simultaneously acquire NIR and Red images of the same waste soil area, and ensure that the two images are registered at the pixel level; Calculate the moisture index of construction waste: Within the target area of ​​the construction waste, calculate its average gray value in the near-infrared band. Compared with the average gray value of the red band The ratio of the two values ​​is defined as the soil moisture index (SMI). ; Output quantitative parameters: The strong absorption characteristics of water are higher in the NIR band than in the Red band; the higher the moisture content ω of the slag, the better. The value decreased by more than This leads to a decrease in the SMI index; A calibration curve of SMI versus moisture content ω was established through preliminary experiments. The image information is converted into a quantitative physical parameter; the SMI value or the calculated water content ω is used as the input parameter and output to the subsequent volume calculation and calibration unit.

4. The method for estimating the volume of tunnel boring machine excavation based on image processing according to claim 1, characterized in that, In S3, the calibration factor The process of predicting and weighting the changing trends is as follows: Multi-source calibration factor generation: Generates two calibration factors simultaneously. Mass flow rate based on belt scale With volumetric flow The calculated calibration factor, after recursive filtering, has an estimated variance of... ; Optical calibration factor based on multispectral moisture index (SMI) prediction; a regression model between SMI and calibration factor is established using historical data, and the variance of the model predictions is determined based on historical error statistics. ; Optimal weighted fusion: the final calibration factor Not a choice or Instead, it performs optimal weighted fusion based on the reciprocal of the variances of the two variances, and the calculation formula is as follows: ,in , The optimal weight; Volume calibration: Subsequent volume estimates will be calculated using the fused optimal calibration factor. , This is the initial calibration reference density.

5. A system applied to the image processing-based shield tunneling muck volume estimation method according to any one of claims 1-4, characterized in that, Includes the following modules: The image acquisition module is used to deploy image acquisition equipment in the shield tunnel conveyor or slag outlet area to acquire RGB and depth images of the slag in real time, providing raw visual data for subsequent 3D point cloud generation. The point cloud processing module is used to reconstruct three-dimensional point clouds from the acquired images. Through semantic segmentation, filtering, and point cloud fusion preprocessing operations, noise and irrelevant ground features are removed to obtain the slag and soil point cloud model. The volume calculation module is used to calculate the volume of the pre-processed slag point cloud using the slicing method, and combined with the belt speed parameters, to obtain the slag volume within the scanning cycle. The dynamic calibration module is used to integrate the mass flow rate and multispectral moisture content data measured in real time by the belt scale, generate a real-time calibration factor and feed it back to the volume calculation unit to dynamically correct the volume estimation results and compensate for the errors caused by changes in the physical properties of the slag.

Images