Laser radar-based working face end straightness detection method and device

Abstract

This invention relates to the field of automated control technology for fully mechanized mining faces, and discloses a method and device for detecting the straightness of the working face end based on lidar. The method establishes geometric and radiation benchmarks by collecting data from a fixed reference target. In real-time detection, the displacement vector and atmospheric attenuation coefficient are calculated using the instantaneous state of the target, and a dual-domain dynamic correction is performed on the original point cloud using a rigid inverse coordinate transformation and adaptive intensity threshold adjustment. Subsequently, the corrected point cloud is clustered and segmented, and the centroids of the slices are extracted to obtain an ordered set of feature points. Multi-factor weighted curve fitting is performed on the feature point set based on geometric stability and environmental reliability to construct a morphological fitting curve. A quantitative index of straightness deviation is output based on the comparison between this curve and the ideal reference straight line. This invention eliminates the interference of equipment vibration and dust attenuation on the measurement, improving the robustness and accuracy of the straightness detection of scraper conveyors.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for fully mechanized mining faces, specifically to a method and device for detecting the straightness of the working face end based on lidar. Background Technology

[0002] In fully mechanized mining operations, scraper conveyors are not only the main channel for coal transport but also the guide rails for the coal mining machine. The straightness of the scraper conveyor directly determines the stability and operational safety of the coal mining machine, and is a core parameter for ensuring automated, intelligent, and efficient mining at the working face.

[0003] Currently, the industry mainly uses inertial navigation devices integrated into the coal mining machine to infer the straightness of the scraper conveyor. However, this indirect measurement method has an insurmountable problem of a blind zone at the ends. When the coal mining machine moves to both ends of the working face (head or tail) for cutting operations, due to the limited length of the coal mining machine and the installation position of the inertial navigation sensor (usually located in the middle of the machine), the inertial navigation system cannot cover the areas at the very ends of the scraper conveyor, resulting in a detection blind zone of about 10 meters at both ends of the working face. This means that the straightness data of the end area is completely missing, which can easily cause deviations in the end-end adjustment, leading to safety hazards such as equipment collisions and roof leaks during cutting.

[0004] To fill this detection gap, introducing lidar for direct scanning of the end-face area is currently the most effective technical approach. However, to successfully apply lidar at the end of a fully mechanized mining face, the unique and harsh environmental obstacles of this area must be overcome, which has become a key challenge in the implementation of this technology.

[0005] First, the end area is close to the cutting point, and high-frequency strong vibration will cause the measurement coordinate system of the lidar to drift dynamically. If it is not corrected, the straightness data scanned will have serious geometric distortion.

[0006] Secondly, the high concentration of dust and sprayed water vapor in the terminal area causes real-time fluctuations in the transmittance of the laser transmission medium, resulting in attenuation of the echo signal. If only conventional algorithms are used, it is very easy to lose the target due to the weak signal, or introduce a lot of noise because of the inability to distinguish suspended particles.

[0007] Therefore, there is an urgent need for a straightness detection method that can effectively cover the blind zone at the end and resist vibration and dust interference, so as to achieve precise control of the entire longwall mining face. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a method and apparatus for detecting the straightness of the working face end based on lidar. It mainly solves the problem that existing inertial navigation detection technology has blind spots at the end of the fully mechanized mining face and cannot obtain effective straightness data. At the same time, it solves the technical problem that when lidar is introduced for blind spot detection at the end, the measurement data is distorted and the straightness detection fails due to the influence of high-frequency vibration and high concentration of dust and water mist.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] The first aspect of this invention provides a method for detecting the straightness of the working face end based on lidar. This method constructs a dynamic correction benchmark by setting up a fixed reference target, and combines dual-domain signal processing and multi-factor weighted fitting technology to quantify the straightness of the scraper conveyor.

[0011] Before the measurement task begins, the method performs initial benchmark calibration, collects data from a fixed reference target module, calculates and stores its benchmark geometric position and benchmark echo intensity as a reference zero point for subsequent elimination of environmental interference.

[0012] After entering the real-time detection process, raw point cloud data containing three-dimensional coordinates and echo intensity information is acquired, and dual-domain dynamic correction is performed. The measured value of the fixed reference target module at the current moment is compared with the baseline parameters: the instantaneous geometric displacement vector is calculated, and a rigid inverse coordinate transformation is performed on the raw point cloud to eliminate geometric distortion caused by equipment vibration; at the same time, the dynamic atmospheric attenuation coefficient is calculated, and the intensity filtering threshold is dynamically adjusted accordingly to filter out low signal-to-noise ratio noise points generated by absorption or scattering by the environmental medium, generating a corrected point cloud after position compensation and denoising processing.

[0013] After obtaining the corrected point cloud, feature extraction is performed. By setting a 3D spatial bounding box and clustering analysis algorithm, the target point cloud cluster representing the scraper conveyor is separated from the corrected point cloud. The target point cloud cluster is sliced ​​and discretized along the direction of the measurement coordinate system, and the geometric centroid of each slice subset is calculated to extract an ordered set of feature points characterizing the macroscopic morphology of the equipment.

[0014] For the extracted ordered feature point set, a multi-factor weighted curve fitting process is employed. This process comprehensively considers geometric stability, environmental credibility, point cloud density, and reflectivity, calculating a comprehensive weight for each feature point. Specifically, a geometric stability weight model is used to reduce the weight of data in high-vibration areas, while an environmental credibility weight model is used to reduce the weight of data in high-attenuation areas. Based on this comprehensive weight, a weighted least squares objective function is constructed and solved to obtain a morphology fitting curve reflecting the morphology of the scraper conveyor.

[0015] Deviation calculation is performed based on the morphological fitting curve. An ideal reference line is constructed according to the effective domain of the morphological fitting curve. The difference between the morphological fitting curve and the ideal reference line over the entire length is calculated. The term with the largest absolute value of the difference is selected as the quantitative index characterizing the straightness deviation, which is used for subsequent feedback control.

[0016] In the above method, the geometric position compensation of dual-domain dynamic correction is achieved by calculating the difference vector between the instantaneous centroid of the fixed reference target module and the reference geometric position at the current moment, and then applying this vector inversely to the full-field point cloud. Environmental filtering is achieved by calculating the ratio of the average intensity of the fixed reference target module at the current moment to the reference intensity to determine the degree of atmospheric attenuation, and then adaptively setting the minimum effective echo intensity threshold to retain effective target points and remove suspended particle noise when visibility decreases.

[0017] In the above method, the system compares the calculated quantitative index with the allowable straightness threshold set according to the coal mining process requirements. When the quantitative index exceeds the limit, a correction command is generated and sent to the electro-hydraulic control system to adjust the stroke of the hydraulic support pushing jack.

[0018] A second aspect of the present invention provides a working surface straightness detection device based on lidar, comprising a lidar module, a fixed reference target module, and a data processing and control module.

[0019] The lidar module is used to scan the end area of ​​the working face and output raw point cloud data; the fixed reference target module is installed on a long-term stable structure within the lidar's field of view to provide physical geometric reference and radiation reference.

[0020] The data processing and control module is equipped with multiple functional units to execute the above detection methods:

[0021] The signal preprocessing and environmental correction unit is used to calculate the geometric displacement and atmospheric attenuation coefficient based on the real-time state of the fixed reference target module, and to perform rigid correction of point cloud coordinates and adaptive adjustment of intensity threshold.

[0022] The target feature extraction and reconstruction unit is used to extract an ordered set of feature points representing the shape of the scraper conveyor from the corrected point cloud through spatial filtering, clustering and slice centroid algorithms.

[0023] The multi-factor adaptive fitting unit is used to calculate the comprehensive weight of each feature point based on the vibration amplitude and environmental attenuation characteristics, and obtain the morphological fitting curve by solving the weighted least squares objective function.

[0024] The straightness deviation quantization unit is used to construct an ideal reference straight line and calculate the maximum straightness deviation index.

[0025] The communication and interaction interface unit is used to feed back quantitative indicators to the external control system and issue a sensor cleaning alarm when the environmental attenuation coefficient is too low.

[0026] This invention provides a method and apparatus for detecting the straightness of the working surface end based on lidar. It has the following beneficial effects:

[0027] 1. This invention enables effective straightness detection of the blind zone of the inertial navigation system at the end of a fully mechanized mining face. By deploying fixed reference targets and establishing a geometric position compensation mechanism, the real-time centroid displacement of the targets is used to correct the coordinates of the entire field point cloud. This method eliminates the interference of high-frequency vibrations generated by the coal cutting machine on the lidar measurement coordinate system, ensuring the geometric position accuracy of the three-dimensional spatial data of the scraper conveyor.

[0028] 2. This invention overcomes the measurement obstacles of strong vibration and high dust levels in the end-point region by establishing a dual-domain dynamic correction mechanism. This method utilizes the real-time state of a fixed reference target to prioritize geometric position compensation for the high-frequency vibrations generated by the coal mining machine, eliminating coordinate drift. Simultaneously, it employs adaptive environmental filtering combined with a dynamic atmospheric attenuation coefficient, automatically adjusting the signal threshold when visibility is reduced due to dust or water mist. This dual correction method ensures that geometrically accurate point cloud data with a high signal-to-noise ratio can still be acquired even under harsh end-point conditions.

[0029] 3. This invention employs a multi-factor weighted curve fitting method, which improves the robustness of the detection results. During the fitting process, the vibration amplitude and environmental interference at each feature point are comprehensively considered and assigned differentiated weights, automatically reducing the data weights of high-vibration or high-attenuation regions. This method effectively suppresses the impact of sudden changes in the local environment on the overall straightness calculation, making the generated shape fitting curve closer to the actual physical state of the scraper conveyor and meeting the requirements of automated control for data stability. Attached Figure Description

[0030] Figure 1 This is a functional module architecture diagram of a straightness detection system according to an embodiment of the present invention;

[0031] Figure 2 This is a flowchart illustrating a method for detecting the straightness of the working face end according to an embodiment of the present invention.

[0032] Figure 3 The logical structure block diagram of the working face end straightness detection device provided in the embodiment of the present invention.

[0033] Among them, 10 is the lidar module; 20 is the fixed reference target module; and 30 is the data processing and control module. Detailed Implementation

[0034] The technical solutions in 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.

[0035] See attached document Figure 1 The present invention provides a working surface end straightness detection system based on lidar, the system comprising: lidar module 10, fixed reference target module 20 and data processing and control module 30.

[0036] The lidar module 10 is a laser measurement device with multi-echo information acquisition and echo intensity value output functions. This module is fixedly installed on the hydraulic support at the end of the working face, and its scanning field of view covers the area of ​​the scraper conveyor to be measured and the location of the fixed reference target module 20.

[0037] It is important to emphasize here the core intention behind using the lidar module 10 in this embodiment:

[0038] For the straightness detection of fully mechanized mining faces, the current mainstream method in the industry is to rely on the inertial navigation of the coal mining machine. However, this method has an unavoidable physical limitation - when the coal mining machine runs to the end, the inertial navigation cannot detect the shape of its front / rear (i.e., the head / tail of the scraper conveyor), resulting in the two ends of the working face always being in an "unmeasurable" blind zone.

[0039] This invention addresses this specific pain point by mounting the lidar module 10 on the end-mounted hydraulic support. Utilizing its non-contact, wide-field-of-view characteristics, it specifically fills the data gap in the final 10-meter area that the inertial navigation system cannot cover. This system directly scans to acquire the true spatial shape of the end-mounted device, filling the data gap in this area. Subsequent data processing (such as atmospheric correction in S2 and weighted fitting in S4) are all supporting technical means to ensure successful completion of the aforementioned blind-detection objective in this specific area (high vibration, high dust).

[0040] The fixed reference target module 20 is a passive physical target. This module has stable and uniform surface reflectivity characteristics and is deployed on a long-term stable structure within the field of view of the lidar module 10, such as a tunnel sidewall or a fixed piece of equipment. The spatial position of the fixed reference target module 20 remains unchanged during the measurement period.

[0041] The data processing and control module 30 is an industrial computer or embedded computing device. This module connects to the lidar module 10 via a data interface to receive raw point cloud data containing three-dimensional coordinates and echo intensity information output by the lidar module 10. The data processing and control module 30 internally contains program instructions implementing the method of this invention, used to perform data processing, analysis, and calculation, and output the final straightness detection result. This system may also include auxiliary modules for power supply and result interaction.

[0042] See attached document Figure 2 This invention provides a method for detecting the straightness of the working face end, which may include the following steps:

[0043] S1. Perform initial reference calibration. Before the measurement task begins, collect data from the fixed reference target module under relatively clean environmental conditions and calculate the reference parameters of the fixed reference target module, including its reference geometric position and reference echo intensity.

[0044] S2: Real-time acquisition of raw point cloud data containing three-dimensional coordinates and echo intensity information, and based on the reference parameters obtained in step S1, performing dual-domain dynamic correction processing on the raw point cloud data to generate a corrected point cloud.

[0045] S3. Perform target segmentation and feature extraction processing on the corrected point cloud generated in step S2. Identify and separate the target point cloud cluster representing the scraper conveyor from the corrected point cloud, and extract a set of ordered feature points from the target point cloud cluster.

[0046] S4. Perform multi-factor weighted curve fitting on the ordered feature point set extracted in step S3. Calculate a comprehensive weight for each feature point in the ordered feature point set, and perform curve fitting on the ordered feature point set based on the comprehensive weight to obtain a shape fitting curve that represents the true shape of the scraper conveyor.

[0047] S5. Compare and analyze the shape fitting curve obtained in step S4 with the preset ideal reference line, calculate and output the quantitative index characterizing the straightness deviation, and use it as the final straightness detection result.

[0048] The technical details and specific implementation methods involved in each of the above steps will be explained in detail below.

[0049] In one specific embodiment of the present invention, for the hardware deployment and configuration of the system, an industrial-grade lidar with multi-echo acquisition capability and echo intensity information output function is selected. The lidar module 10 is fixedly installed on the hydraulic support beam or its auxiliary structure at the end of the working face. Its installation position must ensure that the scanning field of view can cover the head or tail area of ​​the scraper conveyor, and simultaneously cover the preset fixed reference target area. The fixed reference target module 20 is a physical entity with high and uniform reflectivity, such as a metal plate with an industrial-grade reflective film or a standard diffuse reflector. The fixed reference target module 20 is rigidly connected to a long-term stable, non-moving structure within the field of view of the lidar module 10, such as the anchor bolt support on the sidewall of the tunnel or the stationary end of equipment that does not need to move with the advancement of the working face.

[0050] After the system is physically installed, a right-handed Cartesian coordinate system is established. The origin of this coordinate system is located at the optical center of the lidar module 10. After initial horizontal calibration is completed by the inertial measurement unit (IMU) built into the lidar module 10, the X-axis is defined along the theoretical direction of the working face scraper conveyor, the Y-axis is perpendicular to the X-axis in the horizontal plane and points towards the coal wall, and the Z-axis is perpendicular to the XY plane and points upward.

[0051] Before performing the straightness test at the working face end, the data processing and control module 30 needs to execute an initial datum calibration procedure. This procedure aims to establish a geometric reference zero point and a radiation reference datum for subsequent dynamic correction. Specifically, the initial datum calibration procedure may include the following steps:

[0052] S110, Environmental Verification and Data Acquisition. During periods of machine downtime for maintenance, shift handover, or good ventilation at the work site, operators confirm or use dust sensors to detect that the current dust concentration in the environment is at a low level. The data processing and control module 30 controls the lidar module 10 to continuously acquire NN frames of raw point cloud data. Frame of raw point cloud data. Among them, The preset number of sampling frames is usually selected as a positive integer that satisfies statistical significance, for example... or .

[0053] S120, Reference Target Point Cloud Segmentation. For each frame of point cloud data acquired, the point cloud clusters belonging to the fixed reference target module 20 are segmented from the background point cloud using the spatial location range (Region of Interest, ROI) surrounding the fixed reference target module 20 pre-defined during the system configuration phase and its high echo intensity characteristics. For the specific algorithm of point cloud segmentation, those skilled in the art can use a combination of pass-through filtering and Euclidean clustering algorithm, or a density-based DBSCAN clustering algorithm, which are well-known techniques in the field and will not be elaborated upon here.

[0054] S130, Geometric reference parameter calculation. For each point cloud cluster of the fixed reference target module 20 segmented in step S120, its geometric centroid coordinates are calculated. Subsequently, the centroid coordinates of the NN frames of data are statistically averaged to eliminate random measurement noise, obtaining the reference geometric position of the fixed reference target module 20. This calculation process is performed using the FRT geometric reference calibration formula, which is:

[0055] ;

[0056] in: The reference geometric position obtained by the solution is a three-dimensional coordinate vector with the unit being meters (m). This represents the total number of frames collected during the calibration process and is a dimensionless integer. This represents the frame sequence index during the calibration process, with values ​​ranging from 1 to... ; Represents a sequence arrive The terms are summed up. Indicates the first In the frame data, the geometric centroid three-dimensional coordinate vector of the point cloud cluster of the segmented fixed reference target module 20 is expressed in meters (m).

[0057] S140, Radiation reference parameter calculation. This is the same as step S130. In the frame data, the echo intensity values ​​of all points in the point cloud cluster of the fixed reference target module 20 in each frame are extracted, and the average intensity of that frame is calculated. Subsequently, for... The average intensity of the frame data is statistically averaged to obtain the reference echo intensity of the fixed reference target module 20. This calculation process is performed using the FRT radiation reference calibration formula, which is:

[0058] ;

[0059] in, The calculated reference echo intensity is a dimensionless value that characterizes the laser echo energy. This represents the total number of frames collected during the calibration process and is a dimensionless integer. This represents the frame sequence index during the calibration process, with values ​​ranging from 1 to... ; Represents a sequence arrive The terms are summed up. Indicates the first In the frame data, the average echo intensity of the point cloud cluster of the segmented fixed reference target module 20 is dimensionless.

[0060] Through the above steps S110 to S140, the system completed the process of... and The calibration is performed and stored in the memory of the data processing and control module 30 as a reference constant for subsequent real-time dual-domain dynamic correction. When the geological conditions of the working face change significantly, causing roadway deformation, or when the installation position of the lidar module 10 is manually adjusted, the above steps S110 to S140 need to be repeated to update the reference parameters.

[0061] After initial benchmark calibration, the system enters the real-time detection phase. In this phase, the data processing and control module 30 performs dual-domain dynamic correction on each frame of raw data transmitted by the lidar module 10. This process aims to eliminate the impact of high-frequency vibrations caused by the coal mining machine's cutting on geometric accuracy, and to eliminate the attenuation interference of dynamically changing coal dust on the laser echo intensity. Specifically, the real-time dual-domain dynamic correction process may include the following steps:

[0062] S210, Real-time FRT state parameter calculation. For the single-frame raw point cloud data acquired by the lidar module 10 at timestamp t, the data processing and control module 30 uses the same region of interest (ROI) parameters and segmentation algorithm as in the initial calibration stage to quickly locate and extract the point cloud clusters belonging to the fixed reference target module 20 at the current moment. Subsequently, the instantaneous centroid coordinates of the point cloud clusters are calculated. and instantaneous average echo intensity .

[0063] S220, Instantaneous geometric displacement vector calculation. Based on the instantaneous centroid coordinates obtained in step S210. Reference geometric positions stored during the initial calibration phase The three-dimensional translational offset of the radar sensor relative to a reference state is calculated. This calculation process is performed using the instantaneous geometric displacement solution formula, which is:

[0064] ;

[0065] in, Indicates at time The instantaneous geometric displacement vector is a three-dimensional vector with units of meters (m). This vector represents the instantaneous components of the equipment vibration in three spatial axes. Indicates at time The extracted three-dimensional coordinate vectors of the centroids of the 20 cloud clusters in the fixed reference target module are in meters (m). The reference geometric position of the fixed reference target module 20 obtained from the initial calibration is represented by a three-dimensional constant vector in meters (m).

[0066] S230, Calculation of dynamic atmospheric attenuation coefficient. Based on the instantaneous average echo intensity obtained in step S210. Reference echo intensity stored during the initial calibration phase The transmittance characteristics of the current optical path medium are calculated. This calculation process uses the dynamic atmospheric attenuation coefficient formula, which is:

[0067] ;

[0068] in, Indicates time The dynamic atmospheric attenuation coefficient is a dimensionless scalar whose value is usually in the range of (0,1]. Indicates at time The average echo intensity of the extracted 20 cloud clusters from the fixed reference target module is dimensionless. This represents the reference echo intensity of the fixed reference target module 20 obtained from the initial calibration, and is dimensionless.

[0069] S240, Point cloud geometric position compensation. The instantaneous geometric displacement vector calculated in step S220 is used... This process is applied to all original scene point clouds in the current frame, excluding the fixed reference target module 20, transforming the point cloud data from the vibrating sensor coordinate system to a stable reference measurement coordinate system. This calculation is performed using a point cloud geometric position compensation formula, which is:

[0070] ;

[0071] in, Indicates time The The accurate 3D coordinate vector of each scene point after compensation, in meters (m). Indicates time The The three-dimensional coordinate vectors of the original scene points are in meters (m). Indicates time The instantaneous geometric displacement vector, in meters (m); This indicates the index identifier of the location in the current frame.

[0072] S250, Point Cloud Adaptive Environment Filtering. The point cloud data after geometric compensation in step S240 is filtered to remove noise points generated by suspended coal dust and water mist. This step specifically includes a multi-echo selection sub-step and an intensity denoising sub-step. In the multi-echo selection sub-step, if the lidar module 10 returns multiple echo signals for the same emission angle, the data processing and control module 30 performs a logical judgment, extracting only the point data corresponding to the last echo as the valid point and discarding the preceding echo data, thereby utilizing the laser's penetration characteristics to filter out suspended interference at the path's leading edge.

[0073] In the intensity denoising sub-step, the dynamic atmospheric attenuation coefficient calculated in step S230 is used. The intensity filtering threshold is adaptively adjusted. This calculation process uses an adaptive intensity denoising threshold formula to determine the current filtering threshold. The adaptive intensity denoising threshold formula is as follows:

[0074] ;

[0075] in, Indicates time The dynamic intensity denoising threshold is dimensionless. This represents the preset basic intensity threshold, which is dimensionless and is set based on the noise floor characteristics of the lidar module 10 and the lower limit of the reflectivity of the target object in a clean environment. Indicates time The dynamic atmospheric attenuation coefficient is dimensionless.

[0076] Calculated Then, the data processing and control module 30 traverses the current frame point cloud and executes a comparison and judgment instruction: if the echo intensity value of a certain point is less than... If the dust concentration increases, the point is marked as a noise point and removed from the point cloud dataset; otherwise, the point is retained as valid measurement data. This mechanism allows for the detection of noise points when increased dust concentration leads to a decrease in overall echo intensity (i.e.,...). When the threshold is reduced, the filtering threshold is adjusted accordingly. Automatic synchronization and reduction prevent the accidental filtering out of valid target points that have darkened due to attenuation, thus ensuring the integrity of target data under harsh operating conditions.

[0077] After obtaining high-quality point cloud data with real-time dual-domain dynamic correction, the system enters the target segmentation and feature extraction stage. In this stage, the data processing and control module 30 performs the operation of identifying the scraper conveyor entity from the massive 3D point cloud and transforming its discrete surface morphology into a set of ordered, representative mathematical feature points. This process is a crucial link connecting the original measurement data with the final straightness evaluation. Specifically, the target segmentation and feature extraction process may include the following steps:

[0078] S310, Coarse positioning of the target area. The data processing and control module 30 first sets a three-dimensional spatial bounding box containing the expected head or tail of the scraper conveyor based on the prior geometric layout information of the working face end. This bounding box covers the effective measurement length of the scraper conveyor in the X-axis direction and sets a generous tolerance range in the Y and Z-axis directions. The data processing and control module 30 uses this bounding box to perform pass-through filtering on the corrected point cloud output from step S250, eliminating obvious non-target interference points such as those on the tunnel roof, sides, and distant background, retaining only the point cloud data within the bounding box as a candidate set.

[0079] S320, Refined Target Segmentation. Connectivity component clustering analysis is performed on the candidate point cloud obtained in step S310 to separate the scraper conveyor from other nearby objects (such as hydraulic support columns, personnel, or ground-level coal piles). In this step, the data processing and control module 30 employs the Euclidean clustering algorithm. The system sets a clustering tolerance distance threshold. (e.g., 0.1 meters to 0.2 meters) and minimum cluster point threshold The algorithm iterates through the candidate set, selecting those with a spatial distance less than [a certain value]. The points are grouped into a single independent point cloud cluster. After clustering, the data processing and control module 30 calculates the number of points contained in all generated point cloud clusters, and based on the physical characteristics of the scraper conveyor as the main large-scale equipment in the field of view, determines the point cloud cluster containing the most points as the target point cloud cluster representing the scraper conveyor. The specific mathematical implementation of the Euclidean clustering algorithm is well-known in this field and will not be elaborated upon here.

[0080] S330, Discretization along the direction. To analyze the bending morphology of the scraper conveyor, it needs to be discretized along the direction (i.e., the X-axis direction of the measurement coordinate system). The data processing and control module 30 first acquires the target point cloud cluster. Maximum value in the X-axis direction and minimum value And set a slice sampling interval. (For example, 0.5 meters or 1.0 meter, this value is set according to the length of the central slot of the coal mining machine). Then, the target point cluster is mapped along the X-axis. Divided into A spatially spaced slice, in which The total number of slices is calculated as follows: Each slice contains a subset of point clouds with X coordinates within a specific interval.

[0081] S340, Ordered Feature Point Calculation. For each subset of point clouds within a spatial slice, a three-dimensional feature point representing the center position of that section of the scraper conveyor is calculated. To reduce the impact of local surface unevenness or residual loose coal on positional accuracy, the data processing and control module 30 uses a centroid algorithm to extract feature points. This calculation process is performed using the feature point centroid extraction formula, which is:

[0082] ;

[0083] in, Represents the sequence index of feature points, with values ​​ranging from arrive This corresponds to the order along the X-axis from one end to the other. Represents the calculated first... The three-dimensional coordinate vector of each feature point contains Quantity, in meters (m); Indicates the first The total number of point clouds contained within a spatial slice is a dimensionless integer. Indicates the first The traversal index of a point within a spatial slice, with values ​​ranging from... arrive ; Indicates the first Within the slice from the first From point 1 to point 2 Perform a summation operation on each point; Indicates the first Within the spatial slice, the first The three-dimensional coordinate vector of the original point is expressed in meters (m).

[0084] After processing in steps S310 to S340, the data processing and control module 30 transforms the massive unordered point cloud into a set containing... An ordered set of feature points This set of points is monotonically arranged according to the X-axis coordinate, accurately reflecting the macroscopic geometry of the scraper conveyor at the current moment, and is transmitted as input data to the subsequent multi-factor weighted curve fitting module.

[0085] After obtaining an ordered set of feature points representing the discrete morphology of the scraper conveyor through the aforementioned target segmentation and feature extraction modules, the data processing and control module 30 performs multi-factor weighted curve fitting. This processing aims to overcome the deficiency of single-definition fitting being sensitive to outliers. By fusing multi-dimensional information such as the equipment's vibration state, environmental visibility, point cloud density, and reflection characteristics, a confidence weight is assigned to each feature point, thereby reconstructing the true nonlinear bending morphology of the scraper conveyor. Specifically, this process may include the following steps:

[0086] S410, Multi-factor adaptive weight calculation. The data processing and control module 30 iterates through each feature point in the ordered feature point set. (in (For feature point indexing), and tracing back to the original point cloud slice data that constitutes the feature point, the geometric stability weight, environmental confidence weight, density weight, and reflectivity weight are calculated respectively, and finally a comprehensive weight is synthesized. .

[0087] First, calculate the geometric stability weights. The data processing and control module 30 retrieves the... The instantaneous geometric displacement vectors of all original points in each slice at the time of acquisition. (Obtained from step S220), calculate the average value of its modulus, and record it as the average vibration amplitude of the slice. Subsequently, geometric stability weights were calculated based on the negative exponential decay model to reduce the impact of data collected during periods of high vibration on the fitting. This calculation process employed the geometric stability weight model formula, which is as follows:

[0088] ;

[0089] in, Indicates the first The geometric stability weights of each feature point take values ​​in the range (0,1] and are dimensionless. Represented by natural constant An exponential function with base 0; The vibration sensitivity coefficient is a preset constant greater than 0, and its unit is 1 / 2 Ω·cm. This is used to adjust how quickly the weight decays with the vibration amplitude, and is usually set according to the inherent accuracy of the lidar module 10. Indicates the first The average three-dimensional vibration displacement modulus corresponding to each slice is expressed in meters (m). This represents scalar multiplication and negation operations, with the result being a dimensionless numerical value.

[0090] Secondly, calculate the environmental credibility weight. Data processing and control module 30 retrieves the... Dynamic atmospheric attenuation coefficients corresponding to all original points in each slice (Obtained from step S230), and its average value is calculated. This value directly reflects the air cleanliness level at the time the data was collected. The calculation process uses the environmental credibility weight model formula, which is:

[0091] ;

[0092] in, Indicates the first The environmental credibility weight of each feature point takes a value range of (0,1] and is dimensionless. Indicates the first The average atmospheric attenuation coefficient of each slice is dimensionless. The environmental impact moderating factor is a positive real number (e.g., or This is used to strengthen the weighted penalty in high-dust environments.

[0093] In addition, the data processing and control module 30 also calculates density weights. and reflectivity weight For density weights, the formula is used. ,in For the first The number of points in each slice. This represents the maximum number of points across all slices. For the reflectance weight, this calculation process uses the reflectance weight formula, which is:

[0094] ;

[0095] In this formula and its operands:

[0096] Indicates the first The reflectance weights of each feature point are dimensionless. Indicates the first The average value of the echo intensity of all point clouds within a slice, dimensionless; This represents the reference echo intensity obtained from the initial calibration, which is dimensionless. Finally, the weighting factors of the above four dimensions are multiplied and fused to obtain the... The combined weight of each feature point .

[0097] S420, weighted least squares curve fitting. This is used to obtain each feature point. coordinates and their corresponding comprehensive weights Subsequently, the data processing and control module 30 uses the weighted least squares (WLS) method to fit the horizontal orientation of the scraper conveyor. The objective function for fitting is set as follows: ,in For one order polynomial (usually taken) To adapt to the common "S"-shaped bends in scraper conveyors. The data processing and control module 30 solves for the polynomial coefficients by minimizing the weighted sum of squared residuals. This solution process is achieved by constructing and minimizing a weighted least squares objective function, which is:

[0098] ;

[0099] in, This represents the objective function value to be minimized, i.e., the weighted sum of squared residuals; Represents the vector of polynomial coefficients to be solved. ; This represents the total number of feature points involved in the fitting process; it is dimensionless. The sequence index represents the feature point, with values ​​ranging from 1 to... ; This indicates that the first step calculated in step S410 is... The comprehensive weight of each feature point is dimensionless. Indicates the first The measured Y-axis coordinates (i.e., coordinates deviating from the coal face) of each feature point are in meters (m). represents the order of the fitted polynomial, a dimensionless integer; Represents the term index of a polynomial, from arrive ; Indicates the first Coefficients of a polynomial of order 1; Indicates the first The X-axis coordinates of the feature points Power of; This overall term represents the corresponding value on the fitted curve. Predicted Y-coordinate of the location.

[0100] The data processing and control module 30 processes the above objective function. Regarding each coefficient Find the partial derivatives and set them to zero to construct a system of linear equations. Solve the system using matrix decomposition methods (such as QR decomposition or SVD decomposition) to obtain the optimal coefficient vector. The polynomial curve determined from this. This is the shape fitting curve of the scraper conveyor at the current moment. This curve takes into account the effects of equipment vibration and environmental interference, and has extremely high robustness.

[0101] In the previous embodiment, a fitting polynomial curve that characterizes the actual bending shape of the scraper conveyor at the current moment was obtained by the multi-factor weighted least squares method. Based on the fitting results, the data processing and control module 30 performs a quantitative evaluation of straightness, aiming to calculate the degree of deviation of the actual shape of the scraper conveyor from the ideal straight state, and generate control commands to guide the hydraulic support's pushing operation. Specifically, the straightness quantification and output process may include the following steps:

[0102] S510, Construction of the Ideal Reference Line. To quantitatively evaluate straightness, a comparison benchmark must first be established. The data processing and control module 30 then uses the fitted curve... Valid domain (This domain is determined by the slice range in Example 3), extract the coordinates of the two endpoints of the curve. Let the coordinates of the starting point be... The endpoint coordinates are The system defines the straight line connecting these two points as the "ideal reference line," which represents the theoretical trajectory of the scraper conveyor when it is in an absolutely straight state, given that the current positions of the head and tail are fixed. The data processing and control module 30 constructs a mathematical model of this line, and its calculation process uses the ideal reference line equation, which is:

[0103] ;

[0104] in, Indicates the position along the direction. The Y-axis coordinate of the ideal reference line at that location, in meters (m); This represents the coordinate variable of any position along the scraper conveyor, with a value range of [value range missing]. ; The X-coordinate represents the starting point of the effective domain of the fitted curve, which usually corresponds to the head of the scraper conveyor or the starting point of the measurement section, and is expressed in meters (m). Indicates the fitted curve at The function value at that point, i.e. The unit is meters (m); The X-coordinate represents the termination of the effective domain of the fitted curve, which usually corresponds to the tail of the scraper conveyor or the end of the measurement section, and is expressed in meters (m). Indicates the fitted curve at The function value at that point, i.e. The unit is meters (m); This overall term represents the slope of the ideal reference line, characterizing the degree of inclination of the entire scraper conveyor relative to the X-axis of the measurement coordinate system, and is dimensionless; Indicates the current position The distance offset relative to the starting point, in meters (m).

[0105] S520, Straightness deviation global calculation. After establishing the datum, the data processing and control module 30 calculates the straightness deviation within the defined domain. Within, at a preset resolution step size Generate a series of discrete sampling points (e.g., 0.1 meters). For each sampling point, calculate the fitted curve. With the ideal reference line The difference between these values ​​is used to obtain the deviation distribution sequence along the entire length of the scraper conveyor. To obtain a single evaluation index, the term with the largest absolute value in this sequence is selected as the current straightness deviation index. This calculation process is performed using the maximum straightness deviation formula, which is:

[0106] ;

[0107] in, This represents the calculated maximum straightness deviation of the scraper conveyor, expressed in meters (m). The smaller the value, the closer the equipment is to a straight state. This represents the sequence index of discrete sampling points, with values ​​ranging from 1 to... ; This represents the total number of sampling points, and its value is... This rounding result; Indicates the first The x-axis coordinates of each sampling point are given in meters (m). Indicates the location At that point, based on the polynomial curve obtained in step S420 The calculated actual Y-axis coordinates are in meters (m). Indicates the location The ideal reference Y-axis coordinates, calculated according to the equation in step S510, are given in meters (m). This represents the absolute value of the difference between the actual coordinates and the reference coordinates, i.e., the magnitude of the local deviation at that point; This indicates that all sampling points are traversed. to Select the term with the largest absolute difference as the final result.

[0108] S530, Deviation Assessment and Data Release. The data processing and control module 30 will calculate the... Compared with the preset straightness tolerance threshold (For example, 0.05 meters) for comparison. If If the straightness of the scraper conveyor exceeds the standard, it is determined that the current straightness is out of tolerance. At this time, the data processing and control module 30 not only outputs... The numerical values ​​will also include the deviation distribution sequence. The corresponding values ​​are mapped to the serial number of each hydraulic support, packaged into a push-off correction command, and sent to the electro-hydraulic control system of the longwall face via a communication interface. Based on this command, the electro-hydraulic control system dynamically adjusts the stroke of the push-off jacks of each support, thereby achieving closed-loop automatic correction of the straightness of the working face in the next cutting cycle. The specific data transmission protocol and the action execution mechanism of the electro-hydraulic control system are well-known technologies in this field and will not be elaborated upon here.

[0109] See attached document Figure 3 This device is designed to achieve real-time, high-precision monitoring of the straightness of scraper conveyors in harsh underground mining environments, and to provide data support for automated conveyor pushing. The device's physical architecture mainly includes a lidar module 10, a fixed reference target module 20, and a data processing and control module 30. The data processing and control module 30, as the core computing unit, is connected to the lidar module 10 via a bus interface and stores computer-readable instructions in its internal memory. When these instructions are executed by the processor, the data processing and control module 30 is logically configured into the following functional subunits:

[0110] Signal preprocessing and environmental correction unit. This unit receives the raw 3D point cloud data acquired by the lidar module 10 and performs dual-domain dynamic correction based on pre-calibrated parameters. Specifically, this unit is configured to identify the fixed reference target module 20 within the field of view and calculate the instantaneous geometric displacement vector of the current frame in real time. and dynamic atmospheric attenuation coefficient This unit utilizes A rigid inverse transformation is performed on the coordinates of the entire point cloud to eliminate geometric distortions caused by fuselage vibration, while utilizing... Dynamically adjust intensity filtering threshold When performing intensity denoising, this unit is configured to compare the echo intensity at each point with the calculated... By comparing the data, only valid points with intensity higher than the threshold and belonging to the last return are retained, thereby filtering out suspended coal dust interference.

[0111] The target feature extraction and reconstruction unit is used to separate the scraper conveyor entity from the corrected point cloud background and transform it into mathematically tractable discrete feature points. This unit is configured to perform spatial pass-through filtering based on preset 3D bounding box parameters and utilize Euclidean clustering to identify the core point cloud cluster with the most points as the scraper conveyor target. To achieve morphological dimensionality reduction, the unit is further configured to move along the X-axis with a fixed step size. The target point cloud cluster is sliced, and its geometric centroid is calculated for each subset of slices, thereby generating a... An ordered set of feature points composed of coordinate points .

[0112] Multi-factor adaptive fitting unit. This unit is the core component for achieving highly robust morphological reconstruction. This unit is configured to adapt to each feature point... Calculate the overall weight This weight is the product of four factors: geometric stability, environmental confidence, point cloud density, and reflectivity. The calculation logic for the geometric stability weight follows the geometric stability weight model formula, i.e. To reduce the confidence level of data in high-vibration areas; the calculation logic of environmental confidence weights is based on the environmental confidence weight model formula, i.e. After obtaining the weighted parameters, the unit constructs and solves the weighted least squares objective function. The polynomial coefficient vector that minimizes the weighted sum of squared residuals is calculated using a matrix factorization algorithm. The weighted least squares objective function is:

[0113] ;

[0114] In this formula, Represents the term index of a polynomial. The symbols represent the corresponding polynomial coefficients, and the meanings of the remaining symbols are consistent with those in the aforementioned embodiments. By solving this function, a fitting curve characterizing the current state of the scraper conveyor is obtained. .

[0115] Straightness deviation quantization unit. This unit is used to convert the fitted curve into specific deviation values ​​required for industrial control. This unit is configured to construct an ideal reference straight line equation based on the endpoint coordinates of the fitted curve. The theoretical trajectory under absolutely flat conditions is established. Subsequently, the unit performs discretization sampling over the entire length, calculates the absolute value of the difference between the actual fitted coordinates and the ideal reference coordinates at each sampling location, and extracts the maximum value as the maximum straightness deviation at the current moment. The calculation logic strictly follows the maximum straightness deviation formula in step S520 above:

[0116] ;

[0117] Communication and interaction interface unit. This unit is equipped with industrial Ethernet or CAN bus driver circuitry for transmitting calculated data. The deviation sequence data distributed along the scraper conveyor is packaged and sent to the external electro-hydraulic control system of the hydraulic support. This unit also has a self-testing function; when the calculated deviation sequence data is sent to the external electro-hydraulic control system... When the temperature remains below the preset safety threshold (meaning the sensor surface may be completely covered by mud), a cleaning alarm signal is proactively sent to the central control center.

[0118] In this embodiment, the data processing and control module 30 can be an embedded industrial computer (IPC) or a field-programmable gate array (FPGA) hardware platform based on the ARM architecture. Its internal memory includes, but is not limited to, high-speed random access memory (RAM) and non-volatile memory (such as Flash or hard disk), used to store the aforementioned computer program code and initial calibration parameters. This includes point cloud data caching during runtime. The selection of specific hardware models and circuit connection methods can be conventionally designed by those skilled in the art based on actual operating conditions; these are well-known technologies in the field and will not be elaborated upon here.

Claims

1. A method for detecting the straightness of the working face end based on lidar, characterized in that, Includes the following steps: S1, Perform initial reference calibration: Before the measurement task begins, collect data from the fixed reference target module and calculate the reference parameters of the fixed reference target module, including the reference geometric position and the reference echo intensity; S2, Dual-domain dynamic correction: Real-time acquisition of raw point cloud data containing three-dimensional coordinates and echo intensity information, and based on the reference parameters obtained in step S1, performing dual-domain dynamic correction processing on the raw point cloud data to generate a corrected point cloud; The dual-domain dynamic correction process specifically includes: Extract the point cloud clusters that belong to the fixed reference target module at the current moment, and calculate the instantaneous centroid coordinates and instantaneous average echo intensity of the point cloud clusters; The difference between the instantaneous centroid coordinates and the reference geometric position in the reference parameters is calculated to obtain the instantaneous geometric displacement vector. The instantaneous geometric displacement vector is then applied to all original scene point clouds in the current frame except for the fixed reference target module, and the point cloud data is converted to the reference measurement coordinate system. The ratio of the instantaneous average echo intensity to the reference echo intensity in the reference parameters is calculated to obtain the dynamic atmospheric attenuation coefficient; the intensity filtering threshold is adjusted based on the dynamic atmospheric attenuation coefficient; the point cloud data after transformation to the reference measurement coordinate system is filtered using the adjusted intensity filtering threshold; points with echo intensity values ​​less than the intensity filtering threshold are marked as noise points and removed to generate the corrected point cloud. S3, Feature Extraction: Perform target segmentation and feature extraction processing on the corrected point cloud generated in step S2, identify and separate the target point cloud clusters representing the scraper conveyor, and extract an ordered set of feature points from the target point cloud clusters; S4, Curve Fitting: Perform multi-factor weighted curve fitting on the ordered feature point set extracted in step S3, calculate the comprehensive weight for each feature point in the ordered feature point set, and perform curve fitting on the ordered feature point set based on the comprehensive weight to obtain the morphological fitting curve. S5, Deviation Calculation: Compare and analyze the shape fitting curve obtained in step S4 with the preset ideal reference straight line, calculate and output the quantitative index characterizing the straightness deviation, as the final straightness detection result.

2. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, The specific parameters of the fixed reference target module calculated in step S1 include: The lidar is controlled to continuously acquire multiple frames of raw point cloud data, and point cloud clusters belonging to the fixed reference target module are segmented from each frame of raw point cloud data. The geometric centroid coordinates of each frame are calculated based on the point cloud cluster using the FRT geometric reference calibration formula and then statistically averaged to obtain the reference geometric position in the reference parameters. The average echo intensity of each frame is calculated based on the point cloud cluster using the FRT radiation reference calibration formula and then statistically averaged to obtain the reference echo intensity in the reference parameters.

3. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, The step S3, which identifies and separates the target point cloud cluster representing the scraper conveyor and extracts an ordered set of feature points from the target point cloud cluster, specifically includes: A three-dimensional bounding box containing the expected scraper conveyor is set, and the corrected point cloud is subjected to pass-through filtering to obtain a candidate set; Cluster analysis is performed on the candidate set to determine the point cloud cluster containing the most points as the target point cloud cluster representing the scraper conveyor; The target point cloud cluster is divided into multiple equally spaced slices along the X-axis of the measurement coordinate system; The geometric centroid of each subset of point clouds within the spatial slice is calculated using the feature point centroid extraction formula, thus obtaining the ordered feature point set.

4. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, The calculation of the comprehensive weight for each feature point in the ordered feature point set in step S4 specifically includes: The geometric stability weight is calculated using the geometric stability weight model formula, based on the average vibration amplitude of the slice corresponding to the feature points in the ordered feature point set, and through the negative exponential decay model. The environmental credibility weight is calculated based on the average atmospheric attenuation coefficient of the slice corresponding to the feature points in the ordered feature point set using the environmental credibility weight model formula. The geometric stability weight is multiplied and fused with the environmental credibility weight, density weight, and reflectivity weight to obtain the comprehensive weight.

5. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, In step S4, curve fitting is performed on the ordered feature point set based on the comprehensive weights to obtain the morphological fitting curve. Specifically, this includes: Based on the polynomial function and incorporating the comprehensive weights, a weighted least squares objective function is constructed. Minimize the weighted least squares objective function to solve for the polynomial coefficients and determine the morphological fitting curve.

6. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, The calculation and output of the quantitative indicators characterizing the straightness deviation in step S5 specifically includes: Based on the effective domain of the morphology fitting curve, extract the coordinates of the two endpoints of the morphology fitting curve; An ideal reference line connecting the two endpoints is constructed based on the coordinates of the two endpoints using the equation of an ideal reference line. Within the effective domain, discrete sampling points are generated, and the difference between the shape fitting curve and the ideal reference line at each sampling point is calculated. The maximum absolute value of the difference is selected as the quantitative index of straightness deviation using the maximum straightness deviation formula, thus obtaining the quantitative index.

7. The method for detecting the straightness of the working face end based on lidar according to claim 1, characterized in that, The process following step S5 also includes: The quantitative index of the straightness deviation is compared with the preset straightness tolerance threshold. If the quantitative index is greater than the straightness tolerance threshold, a displacement correction command is generated and sent to the electro-hydraulic control system of the longwall mining face to adjust the stroke of the hydraulic support displacement jack. The straightness tolerance threshold is the maximum permissible absolute value of deviation set according to the coal mining process requirements of the fully mechanized mining face.

8. A device for detecting the straightness of the working face end based on lidar, characterized in that, The apparatus implements the method as described in claim 1, the apparatus comprising: The lidar module is used to scan the end area of ​​the working face and output raw point cloud data containing three-dimensional coordinates and echo intensity information to the data processing and control module. A fixed reference target module is deployed on a long-term stable structure within the field of view of the lidar module to provide a physical reference for generating reference parameters, including reference geometric position and reference echo intensity. A data processing and control module, connected to the lidar module, is used to receive the raw point cloud data. The data processing and control module includes: The signal preprocessing and environmental correction unit is configured to calculate the instantaneous geometric displacement vector and the dynamic atmospheric attenuation coefficient based on the original point cloud data and the reference parameters, perform a rigid inverse coordinate transformation using the instantaneous geometric displacement vector, and dynamically adjust the intensity filtering threshold using the dynamic atmospheric attenuation coefficient to output the corrected point cloud. The target feature extraction and reconstruction unit is configured to receive the corrected point cloud, separate the scraper conveyor target through pass-through filtering and cluster analysis, and perform slice segmentation and centroid calculation on the scraper conveyor target along the coordinate axis, extracting and outputting an ordered set of feature points; The multi-factor adaptive fitting unit is configured to receive the ordered feature point set, calculate the comprehensive weight based on geometric stability and environmental credibility, construct a weighted least squares objective function, and obtain the morphological fitting curve by solving the polynomial coefficients. The straightness deviation quantization unit is configured to receive the shape fitting curve, construct an ideal reference straight line based on the curve endpoints, calculate the maximum difference between the shape fitting curve and the ideal reference straight line, and output a quantization index characterizing the straightness deviation. The communication and interaction interface unit is configured to receive the quantitative index and send it to the external electro-hydraulic control system, and to send an alarm signal when the dynamic atmospheric attenuation coefficient is lower than a preset safety threshold. The safety threshold is the critical value at which the echo intensity fails to meet the measurement requirements due to a severe decrease in the cleanliness of the sensor surface.

Images