CN117036998A - Random forest-based open-air rock mass fracture surface identification method - Google Patents
Random forest-based open-air rock mass fracture surface identification method Download PDFInfo
- Publication number
- CN117036998A CN117036998A CN202310997791.9A CN202310997791A CN117036998A CN 117036998 A CN117036998 A CN 117036998A CN 202310997791 A CN202310997791 A CN 202310997791A CN 117036998 A CN117036998 A CN 117036998A
- Authority
- CN
- China
- Prior art keywords
- point
- points
- point cloud
- fracture surface
- rock mass
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000011435 rock Substances 0.000 title claims abstract description 54
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000013598 vector Substances 0.000 claims abstract description 70
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 11
- 239000002689 soil Substances 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 239000012634 fragment Substances 0.000 claims description 2
- 238000002372 labelling Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 7
- 238000009826 distribution Methods 0.000 abstract description 6
- 238000011835 investigation Methods 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 238000007667 floating Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/17—Terrestrial scenes taken from planes or by drones
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Remote Sensing (AREA)
- Image Analysis (AREA)
Abstract
The application relates to the technical field of open-air rock mass fracture surface identification, and discloses an open-air rock mass fracture surface identification method based on random forests, which comprises the following steps of; step 1, three-dimensional point cloud data of an open-air rock mass with RGB information are obtained by using an unmanned aerial vehicle-mounted three-dimensional laser scanning accessory; and step 2, preprocessing the three-dimensional point cloud data of the open-air rock mass with the RGB information, eliminating noise, and calculating the characteristic vector and the normal vector of each point of the three-dimensional point cloud. According to the extracted crack surface point cloud set, the main distribution direction of the crack surface is further determined, and automatic statistics of the main distribution direction of the crack surface is realized. And further extracting a point cloud set of a single fracture surface based on the point cloud set of the main direction, and realizing the yield calculation of the single fracture surface under the main distribution of the fracture surface direction. The automatic statistics of fracture surface parameters of the open-air rock mass is directly realized, and the working efficiency of geological investigation is improved.
Description
Technical Field
The application relates to the technical field of open-air rock mass fracture surface identification, in particular to an open-air rock mass fracture surface identification method based on random forests.
Background
The investigation, measurement and statistics of fracture surfaces are digital bases for classifying rock mass, evaluating rock mass quality and extracting physical and mechanical parameters of the rock mass. Likewise, understanding the pattern of fracture surfaces-the composition of the fracture surfaces and their overall geometric characteristics is important in many sub-disciplines of earth science. Quantification of the observed fracture surface patterns is a necessary prerequisite for understanding the physical basis of fracture surface formation and for reliable predictions of its subsurface extent and size, which ultimately determines the transmission characteristics (i.e., permeability) of the network.
Traditional fracture surface investigation is also in a stage of inefficiency of manual operations. In order to improve the working efficiency and save the engineering cost, it is very necessary to find an efficient, simple and practical field measurement method to solve the above problems. And the three-dimensional laser scanning technology provides powerful technical support for achieving the aim. Secondly, for the interference of other ground features existing in the real ground rock mass point cloud obtained from the three-dimensional laser scanning, how to accurately identify the rock mass fracture surface from the real ground rock mass three-dimensional point cloud data and obtain fracture surface parameters is a main problem to be solved at present.
Disclosure of Invention
In order to solve the technical problems in the background art, the application provides an open-air rock mass fracture surface identification method based on random forests.
The application is realized by adopting the following technical scheme: a random forest-based open-air rock mass fracture surface identification method comprises the following steps of;
step 1, three-dimensional point cloud data of an open-air rock mass with RGB information are obtained by using an unmanned aerial vehicle-mounted three-dimensional laser scanning accessory;
step 2, preprocessing the three-dimensional point cloud data of the open-air rock mass with RGB information, eliminating noise, and calculating the feature vector and normal vector of each point of the three-dimensional point cloud;
step 3, making samples based on the feature vectors of each point in the point cloud, dividing the samples into a training set and a testing set, and respectively training and testing the random forest classifier;
step 4, classifying the open-air rock mass point cloud based on the random forest classifier trained by the training set, and dividing the open-air rock mass three-dimensional point cloud into crack surface points and non-crack surface points (vegetation, soil and rock debris points);
and 5, determining the main direction of the crack surface points, clustering the crack surface points by taking the main direction as a center, further clustering the point cloud sets of all the main directions to obtain a single crack surface point cloud set, and calculating related parameters of the crack surface.
Further, the step 2 includes the following sub-steps:
step 201, calculating the mean mu and variance sigma of the distances between each point and the nearest N points of the three-dimensional point cloud data of the open-air rock mass with RGB information, which are obtained in the step A, and removing noise points with the distances larger than mu+3sigma to obtain filtered point cloud data;
step 202, inquiring K nearest neighbor points closest to an inquiry point based on Euclidean distance for single points in the point cloud data based on the filtered three-dimensional point cloud data;
step 203, calculating a covariance matrix of the adjacent point based on the K nearest neighbor point coordinates closest to the query point, and calculating a feature value and a feature vector of the covariance matrix according to the calculated covariance matrix, wherein the feature vector corresponding to the minimum feature value is used as a normal vector of the point;
step 204, calculating the characteristic based on the characteristic value by using the calculated characteristic value, wherein the characteristic value is respectively: curvature, two-dimensional characteristics, three-dimensional characteristics;
step 205, calculating a direction characteristic based on a normal vector by using the normal vector of the query point: verticality characteristics;
step 206, calculating a radius-based surface descriptor by using the normal vectors of all K adjacent points in the query point neighborhood;
step 207, extracting RGB information of the three-dimensional point cloud data based on the filtered three-dimensional point cloud data in step 201, and calculating relative color characteristics;
step 208, combining curvature, two-dimensional features, three-dimensional features, verticality, radius-based surface descriptors and relative color features of the query point into feature vectors.
Further, the radius-based surface descriptor in step 206 is defined as follows:
according to the distance between the point and the neighborhood point and the normal vector direction, an isosceles triangle with a vertex angle as a normal vector included angle and a bottom edge as the distance between the neighborhood points can be determined, the isosceles triangle determines a unique sphere, and the waist length of the triangle is the radius r of the sphere. K neighborhood points may determine K spheres, reserving the largest and smallest of all K sphere radii as radius-based surface descriptors.
Further, the step 3 further includes the following sub-steps:
step 301, based on the filtered three-dimensional point cloud data, cutting by using cloudcomputer software, selecting four sample points of fracture surface points, vegetation, rock fragments and soil, and obtaining indexes of the sample points;
step 302, extracting a feature vector of a sample point based on the sample point index number, and giving a label according to the category of the sample point;
step 303, randomly sequencing the feature vectors of the sample points, dividing 70% of samples into training sets, and taking 30% of samples as test sets;
step 304, inputting the training set into a random forest classifier model for training to obtain a trained random forest classifier model;
and 305, performing parameter optimization and accuracy verification on the trained random forest classifier model by using the test set to obtain a classification model with higher classification accuracy.
Further, the method based on the step 4 comprises the following sub-steps:
step 401, inputting the feature vectors in step 208 into a trained random forest classifier for prediction, and labeling each feature vector with a class label;
and 402, dividing the three-dimensional point cloud data after filtering into four categories of fracture surface points, vegetation, cuttings and soil based on the labels and indexes of the feature vectors.
Further, the step 5 includes the following sub-steps:
step 501, converting normal vectors of the fracture surface points into trends and dip angles, and then projecting the trends and dip angles into a Cartesian coordinate system to obtain coordinates of the trends and dip angles in the Cartesian coordinate system;
step 502, performing density estimation on each coordinate point by using a two-dimensional Gaussian kernel function on the coordinates of the tendency and the inclination angle in a Cartesian coordinate system to obtain a density value of each point;
step 503, searching points with density values larger than those of surrounding points based on the coordinates and the density values of the coordinate points to obtain maximum points, traversing the maximum points, inquiring all the maximum points with density values larger than the current maximum point, calculating the included angle between the current maximum point and the maximum point with density larger than the current maximum point, selecting the minimum included angle in all the included angles as the minimum included angle of the current maximum point, and aiming at the maximum points, sharing two attributes: (1) a density value; (2) minimum included angle. The average of the two attributes of all maxima points is calculated separately. Taking a normal vector corresponding to a point with a density value larger than the average value and a minimum included angle value larger than the average included angle as a main direction;
step 504, calculating the included angle between each point normal vector and the main direction vector based on the main direction, and dividing each point into main direction categories with the minimum included angle to obtain a point set belonging to each main direction;
step 505, based on the point sets belonging to each main direction, respectively executing a clustering algorithm based on density on the point set of each main direction to obtain a point set of a single fracture surface;
step 506, calculating parameters of each fracture surface based on the point set of the single fracture surface, including: the position of the fracture surface, the fracture surface inclination angle and the main direction of the fracture surface.
Further, in said step 505:
the minimum distance is set to the average value of the fourth neighbor point of all points plus two times the standard deviation, and the allowable minimum number is set to the number K of the neighbor points.
Compared with the prior art, the application has the beneficial effects that:
according to the method for identifying the open-air rock mass fracture surface based on the random forest, which is provided by the application, the three-dimensional point cloud data of the open-air rock mass in the area which is difficult to reach by researchers is acquired based on the three-dimensional laser scanning accessory carried by the unmanned aerial vehicle, so that the researchers can safely and rapidly acquire the relevant parameters of the fracture surface. The application provides a method for extracting a fracture surface point cloud set from a complex geographic environment. Thus, not only saving the time of the researchers working indoors, but also ensuring the manual error caused by the negligence of the researchers. According to the extracted crack surface point cloud set, the main distribution direction of the crack surface is further determined, and automatic statistics of the main distribution direction of the crack surface is realized. And further extracting a point cloud set of a single fracture surface based on the point cloud set of the main direction, and realizing the yield calculation of the single fracture surface under the main distribution of the fracture surface direction. The automatic statistics of fracture surface parameters of the open-air rock mass is directly realized, and the working efficiency of geological investigation is improved.
Drawings
FIG. 1 is a flow chart of a random forest-based open-air rock fracture surface identification method provided by an embodiment of the application;
fig. 2 is a schematic diagram of an extraction result of a three-dimensional point cloud fracture surface of an open rock mass according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a result of clustering a main direction of a fracture surface point cloud set according to an embodiment of the present application;
fig. 4 is a schematic diagram of a main direction point cloud set further clustered into a single fracture surface set result provided by an embodiment of the present application.
Detailed Description
The present application will be further described with reference to the accompanying drawings and detailed description, wherein it is to be understood that, on the premise of no conflict, the following embodiments or technical features may be arbitrarily combined to form new embodiments.
Example 1:
the present example takes the quarry pit area of the floating joint mountain in the Fu Dong county of Anhui as the main study object.
Referring to fig. 1, fig. 1 is a flowchart of a method for identifying open-air rock fracture surfaces based on random forests according to an embodiment of the present application, including
Step 1: acquiring three-dimensional point cloud data of an open-air rock mass with RGB information by using an unmanned aerial vehicle-mounted three-dimensional laser scanning accessory;
for the step (1), before the field data acquisition, the data of the data acquisition site should be collected in detail, including: knowing about relevant laws, regulations, whether there is a limited flight zone.
The method carries out field investigation in advance, and specifically comprises the following steps: and evaluating field conditions, wherein whether potential factors influencing the flight of the unmanned aerial vehicle exist or not, whether potential factors influencing the life safety of a driver of the unmanned aerial vehicle exist or not, and whether potential factors influencing the three-dimensional laser scanning exist or not.
The method for preparing the project implementation scheme specifically comprises the following steps: and integrating the collected data and the field investigation data to prepare the optimal unmanned aerial vehicle flight route, flight height, flight date, key scanning area and relevant parameter determination of unmanned aerial vehicle three-dimensional laser scanning.
The national authorities are applied for flight permissions.
And acquiring image data and three-dimensional point cloud data of the floating joint rock-mining pit by using a three-dimensional laser scanning accessory carried by the unmanned aerial vehicle, and carrying out color assignment on the point cloud data according to RGB information of the image data.
Aiming at the original data acquired by the unmanned aerial vehicle, professional software is required to process the original data, and the method specifically comprises the following steps: precision optimization, coordinate conversion, format conversion, and the like.
It should be noted that, the three-dimensional laser scanning accessory carried by the unmanned aerial vehicle used in the embodiment can simultaneously acquire image data and point cloud data, directly color the point cloud data, and can obtain the three-dimensional point cloud data of the open-air rock mass with RGB information without post-processing. The prior art provides various technical schemes for how to acquire three-dimensional point cloud data of the open-air rock mass with RGB information, and the application is not particularly limited.
Step 2: and preprocessing the three-dimensional point cloud data of the open-air rock mass with RGB information, eliminating noise, and calculating the eigenvectors and normal vectors of each point of the three-dimensional point cloud.
For the above steps, specifically:
step 201, calculating the mean mu and the variance sigma of the distances between each point and the nearest 200 points of the three-dimensional point cloud data of the open-air rock mass with RGB information, and removing noise points with the distances greater than mu+3sigma to obtain filtered point cloud data.
Step 202, based on the filtered three-dimensional point cloud data; for a single point in the point cloud data, based on Euclidean distance, the 200 nearest neighbors to the query point are queried.
It should be noted that, in the unmanned aerial vehicle and the three-dimensional laser scanning accessory thereof used in the embodiment, the point cloud data acquired under the influence of the weather factor of the day when the data acquisition is carried out and various influence factors such as the unmanned aerial vehicle flying hand are adopted, and the local area characteristics can be accurately calculated by testing to obtain 200 adjacent points. Under the influence of different sensors, weather, acquisition personnel, repeated scanning times and other conditions, the acquired point cloud data precision, density and the like are different, and the number of specific optimal adjacent points is required to be obtained by testing according to the finally acquired data.
Step 203, calculating covariance matrix of the adjacent point according to the 200 nearest neighbor point coordinates, and calculating eigenvalue and eigenvector of the covariance matrix according to the calculated covariance matrix, wherein the eigenvector corresponding to the minimum eigenvalue is used as the normal vector of the point.
Step 204, calculating the characteristic based on the characteristic value by using the calculated characteristic value, wherein the characteristic value is respectively: curvature, two-dimensional characteristics, three-dimensional characteristics.
Step 205, calculating a direction characteristic based on a normal vector according to the normal vector of the query point: verticality characteristics.
Step 206, calculating a surface descriptor based on the radius for the normal vector of all 200 adjacent points in the query point neighborhood.
Step 207, extracting RGB information of the three-dimensional point cloud data after filtering, and calculating relative color characteristics.
Step 208, combining curvature, two-dimensional features, three-dimensional features, verticality, radius-based surface descriptors and relative color features of the query point into feature vectors.
The covariance matrix, eigenvalues and eigenvectors of the covariance matrix, curvature, two-dimensional characteristics, and three-dimensional characteristics of the adjacent points are described in detail in the prior art, and are not repeated here.
It should be noted that, the calculation method of the perpendicularity feature mentioned in the present application is described as formula (1)
V＝1-N z (1)
Wherein N is z Is the third component of the point normal vector N.
The radius-based surface descriptor is described as [ r ] max ,r min ]The calculation method is as follows formula (2):
wherein d is i The distance between the query point and the ith adjacent point is the included angle between the normal vector of the query point and the normal vector of the ith adjacent point. For all r calculated i Only the maximum value r is saved max And a minimum value r min As radius-based surface descriptors.
The relative color is characterized by the following formula (3) (4) (5):
wherein R, G, B are three component values of red, green and blue of the RGB color of the point cloud.
Step 3: and (3) manufacturing samples based on the feature vectors of each point in the point cloud, dividing the samples into a training set and a testing set, and respectively training and testing the random forest classifier.
For the above step (3), specifically:
step 301, clipping the filtered three-dimensional point cloud data by using cloudCompare software, respectively selecting four types of sample points of fracture surface points, vegetation, rock debris and soil, and obtaining indexes of the sample points.
It should be noted that, the types of the samples need to be determined according to the on-site situation, and as described above, the natural environment around the surface exposed fracture surface is complex and changeable, and for the embodiment, the ground features existing around the fracture surface are mainly three types of vegetation, rock debris and soil. The present application is not particularly limited as to the type of sample.
And 302, extracting feature vectors corresponding to indexes from the feature vector set calculated in the step 2 aiming at the sample point indexes obtained in the step, and endowing feature vectors of crack surface points with No. 1 labels, wherein the feature vectors of vegetation points are No. 2 labels, the feature vectors of rock debris points are No. 3 labels, and the feature vectors of soil are No. 4 labels.
Step 303, randomly sequencing all the feature vector samples, dividing the samples of which the first 70% are training sets and the samples of which the last 30% are test sets.
And 304, inputting the obtained training set into a random forest classifier model for training to obtain a trained random forest classifier model.
And 305, performing parameter optimization and accuracy verification on the trained random forest classifier model by using the test set to obtain a classification model with higher classification accuracy.
It should be noted that, in this embodiment, the random forest classifier uses the TreeBagger function provided by Statistics and Machine Learning Toolbox of MATLAB, and the parameters are set: the number of decision trees is 100, and the minimum leaf number is 1. The application is not particularly limited with respect to the random forest classifier written in what language. For the division of the ratio of training set to test set, the present embodiment uses the ratio of 7:3 suggested by the existing institute for division. If the sample size is larger, the calculation requirement is higher, the proportion occupied by the training set can be properly reduced, or the proportion can be adjusted according to the actual situation, and the application is not particularly limited.
Step 4: classifying the open-air rock mass point cloud based on a random forest classifier trained by using a training set, and dividing the open-air rock mass three-dimensional point cloud into fracture surface points and non-fracture surface points (vegetation, soil and rock debris points).
For the above step (4)
And 401, inputting the feature vector set calculated in the step 2 into a trained random forest classifier for prediction, and marking a class number with the highest possibility for each feature vector.
Step 402, extracting index numbers of feature vectors with all types being 1 according to the label number of the fracture surface point in step (3), and extracting all fracture surface points according to the obtained index numbers of the fracture surface point. Referring to fig. 2, fig. 2 is a schematic diagram of a three-dimensional point cloud fracture surface extraction result of an open-air rock according to an embodiment of the present application.
In the process of calculating the features of the present embodiment,the index of the point and the index of the feature vector are one A corresponding pair of the two pairs of the two,i.e. points with index 1 correspond to feature vectors with index 1. Representing the data with an index saves system storage space so that more data can be processed in a limited memory space.
Step 5: and determining the main direction of the fracture surface points, clustering the fracture surface points by taking the main direction as a center, further clustering the point cloud sets of all the main directions to obtain a single fracture surface point cloud set, and calculating related parameters of the fracture surface.
For the above step 5, specifically:
step 501, converting normal vectors of fracture surface points into trends and dip angles, and then projecting the trends and dip angles into a Cartesian coordinate system to obtain coordinates of the trends and dip angles in the Cartesian coordinate system.
It should be noted that, the principle of projecting the inclination and the inclination angle into the cartesian coordinate system may refer to the conversion relationship between the polar coordinate system and the cartesian coordinate system, which is an operation convenient for computer operation and display, and the conversion formula of projecting the inclination and the inclination angle into the cartesian coordinate system is described in detail in the prior art, and will not be repeated herein.
And 502, carrying out density estimation on each coordinate point by using a two-dimensional Gaussian kernel function according to the coordinates of the tendency and the inclination angle in the Cartesian coordinate system, and obtaining the density value of each point.
In the present embodiment, the point density estimation is an approximate point density estimation, and since the data size of the three-dimensional point cloud data is large, the calculation cost and the storage cost of the individual density estimation for each point are not estimated for a personal computer. The approximate principle is that a two-dimensional space is uniformly divided into a plurality of square lattices, the number of points inside each square lattice is used as the value of the square lattice, and the density value of each lattice is obtained by carrying out Gaussian kernel density estimation on each lattice. For better-performing computers, it may also be attempted to directly perform gaussian kernel density estimation on the points to obtain more accurate calculation results.
And 503, searching for points with density values larger than those of surrounding points according to the coordinates and the density values of the coordinate points in the step, and obtaining maximum points. Traversing the maximum points, inquiring all maximum points with density values larger than the current maximum point, calculating the included angles between the current maximum point and the maximum points with density larger than the current maximum point, and selecting the minimum included angle in all included angles as the minimum included angle of the current maximum point. For the maximum point, there are two attributes: (1) a density value; (2) minimum included angle. The average of the two attributes of all maxima points is calculated separately. And taking the normal vector corresponding to the point with the density value larger than the average value and the minimum included angle value larger than the average included angle as the main direction. Referring to fig. 3, fig. 3 is a schematic diagram of a result of clustering the main directions of the point cloud set of the fracture surface according to the embodiment of the present application.
It should be noted that the above-mentioned basic principle of the steps refers to the principle based on density peak clustering, and an improvement is made in the present application, so that the method is more suitable for use in three-dimensional point cloud data processing with huge data volume.
And 504, calculating the included angles between the normal vectors of each point and the main direction vector according to the main directions obtained in the steps, and dividing all the fracture surface points into point sets of each main direction under the condition that each point is divided into main direction categories with minimum included angles. And based on the point sets belonging to the main directions, respectively executing a clustering algorithm based on density on the point sets of each main direction to obtain the point set of the single fracture surface. Calculating parameters of each fracture surface based on the point set of the single fracture surface, including: the position of the fracture surface, the fracture surface inclination angle and the main direction of the fracture surface. Referring to fig. 4, fig. 4 is a schematic diagram of a main direction point cloud set further clustered into a single fracture surface set according to an embodiment of the present application.
Step 505, setting the minimum distance as the average value of the fourth adjacent points of all the points plus two times of standard deviation for the density-based clustering algorithm parameters in the step, and setting the allowable minimum number as the adjacent point number 200.
The minimum allowable number means the minimum number of points that can form a single fracture surface under the conditions of the accuracy and density of the current point cloud. The present application is not particularly limited with respect to the minimum number setting.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be noted that: like reference numerals and letters in the following figures denote like items, and thus once an item is defined in one figure, no further definition or explanation of it is required in the following figures, and furthermore, the terms "first," "second," "third," etc. are used merely to distinguish one description from another and are not to be construed as indicating or implying relative importance.
Finally, it should be noted that: the above examples are only specific embodiments of the present application, and are not intended to limit the scope of the present application, but it should be understood by those skilled in the art that the present application is not limited thereto, and that the present application is described in detail with reference to the foregoing examples: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application, and are intended to be included in the scope of the present application. Therefore, the protection scope of the application is subject to the protection scope of the claims.
Claims (6)
1. The method for identifying the open-air rock mass fracture surface based on the random forest is characterized by comprising the following steps of;
step 1, three-dimensional point cloud data of an open-air rock mass with RGB information are obtained by using an unmanned aerial vehicle-mounted three-dimensional laser scanning accessory;
step 2, preprocessing the three-dimensional point cloud data of the open-air rock mass with RGB information, eliminating noise, and calculating the feature vector and normal vector of each point of the three-dimensional point cloud;
step 3, making samples based on the feature vectors of each point in the point cloud, dividing the samples into a training set and a testing set, and respectively training and testing the random forest classifier;
step 4, classifying the open-air rock mass point cloud based on the random forest classifier trained by the training set, and dividing the open-air rock mass three-dimensional point cloud into fracture surface points and non-fracture surface points;
and 5, determining the main direction of the crack surface points, clustering the crack surface points by taking the main direction as a center, further clustering the point cloud sets of all the main directions to obtain a single crack surface point cloud set, and calculating related parameters of the crack surface.
2. The method for identifying open-air rock mass fracture surfaces based on random forests as claimed in claim 1, wherein the step 2 comprises the following sub-steps:
step 201, calculating the mean mu and variance sigma of the distances between each point and the nearest N points of the three-dimensional point cloud data of the open-air rock mass with RGB information obtained in the step 2, and removing noise points with the distances larger than mu+3sigma to obtain filtered point cloud data;
step 202, inquiring K nearest neighbor points closest to an inquiry point based on Euclidean distance for single points in the point cloud data based on the filtered three-dimensional point cloud data;
step 203, calculating a covariance matrix of the adjacent point based on the K nearest neighbor point coordinates closest to the query point, and calculating a feature value and a feature vector of the covariance matrix according to the calculated covariance matrix, wherein the feature vector corresponding to the minimum feature value is used as a normal vector of the point;
step 204, calculating the characteristic based on the characteristic value by using the calculated characteristic value, wherein the characteristic value is respectively: curvature, two-dimensional characteristics, three-dimensional characteristics;
step 205, calculating a direction characteristic based on a normal vector by using the normal vector of the query point: verticality characteristics;
step 206, calculating a radius-based surface descriptor by using the normal vectors of all K adjacent points in the query point neighborhood;
step 207, extracting RGB information of the three-dimensional point cloud data based on the filtered three-dimensional point cloud data in step 201, and calculating relative color characteristics;
step 208, combining the curvature, two-dimensional characteristics, three-dimensional characteristics, verticality, radius-based surface descriptors and relative color characteristics of the query points into a characteristic vector;
the radius-based surface descriptor in step 206 is defined as follows:
according to the distance between the point and the neighborhood point and the normal vector direction, an isosceles triangle with a vertex angle as a normal vector included angle and a bottom edge as a neighborhood point distance can be determined, the isosceles triangle determines a unique sphere, the triangle waist length is a sphere radius r, K neighborhood points can determine K spheres, and the maximum radius and the minimum radius in all K sphere radii are reserved as a radius-based surface descriptor.
3. The method for identifying open-air rock mass fracture surfaces based on random forests as claimed in claim 1, wherein the step 3 further comprises the following sub-steps:
step 301, based on the filtered three-dimensional point cloud data, cutting by using cloudcomputer software, selecting four sample points of fracture surface points, vegetation, rock fragments and soil, and obtaining indexes of the sample points;
step 302, extracting a feature vector of a sample point based on the sample point index number, and giving a label according to the category of the sample point;
step 303, randomly sequencing the feature vectors of the sample points, dividing 70% of samples into training sets, and taking 30% of samples as test sets;
step 304, inputting the training set into a random forest classifier model for training to obtain a trained random forest classifier model;
and 305, performing parameter optimization and accuracy verification on the trained random forest classifier model by using the test set to obtain a classification model with higher classification accuracy.
4. The method for identifying open-air rock mass fracture surfaces based on random forests as claimed in claim 1, wherein the method based on step 4 comprises the following sub-steps:
step 401, inputting the feature vectors in step 208 into a trained random forest classifier for prediction, and labeling each feature vector with a class label;
and 402, dividing the three-dimensional point cloud data after filtering into four categories of fracture surface points, vegetation, cuttings and soil based on the labels and indexes of the feature vectors.
5. The method for identifying open-air rock mass fracture surfaces based on random forests as claimed in claim 1, wherein the step 5 comprises the following sub-steps:
step 501, converting normal vectors of the fracture surface points into trends and dip angles, and then projecting the trends and dip angles into a Cartesian coordinate system to obtain coordinates of the trends and dip angles in the Cartesian coordinate system;
step 502, performing density estimation on each coordinate point by using a two-dimensional Gaussian kernel function on the coordinates of the tendency and the inclination angle in a Cartesian coordinate system to obtain a density value of each point;
step 503, searching points with density values larger than those of surrounding points based on the coordinates and the density values of the coordinate points to obtain maximum points, traversing the maximum points, inquiring all the maximum points with density values larger than the current maximum point, calculating the included angle between the current maximum point and the maximum point with density larger than the current maximum point, selecting the minimum included angle in all the included angles as the minimum included angle of the current maximum point, and aiming at the maximum points, sharing two attributes: (1) a density value; (2) minimum included angle. The average of the two attributes of all maxima points is calculated separately. Taking a normal vector corresponding to a point with a density value larger than the average value and a minimum included angle value larger than the average included angle as a main direction;
step 504, calculating the included angle between each point normal vector and the main direction vector based on the main direction, and dividing each point into main direction categories with the minimum included angle to obtain a point set belonging to each main direction;
step 505, based on the point sets belonging to each main direction, respectively executing a clustering algorithm based on density on the point set of each main direction to obtain a point set of a single fracture surface;
step 506, calculating parameters of each fracture surface based on the point set of the single fracture surface, including: the position of the fracture surface, the fracture surface inclination angle and the main direction of the fracture surface.
6. The method for identifying open-air rock mass fracture surfaces based on random forests as claimed in claim 1, wherein in said step 505:
the minimum distance is set to the average value of the fourth neighbor point of all points plus two times the standard deviation, and the allowable minimum number is set to the number K of the neighbor points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310997791.9A CN117036998A (en) | 2023-08-09 | 2023-08-09 | Random forest-based open-air rock mass fracture surface identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310997791.9A CN117036998A (en) | 2023-08-09 | 2023-08-09 | Random forest-based open-air rock mass fracture surface identification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117036998A true CN117036998A (en) | 2023-11-10 |
Family
ID=88601689
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310997791.9A Pending CN117036998A (en) | 2023-08-09 | 2023-08-09 | Random forest-based open-air rock mass fracture surface identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117036998A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117854061A (en) * | 2024-03-07 | 2024-04-09 | 山东大学 | Tunnel face trace crack extraction method and system based on three-dimensional point cloud |
| CN118053152A (en) * | 2024-04-16 | 2024-05-17 | 中国地质大学（武汉） | Rock mass structural plane rapid measurement method and equipment based on mass point cloud data |
-
2023
- 2023-08-09 CN CN202310997791.9A patent/CN117036998A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117854061A (en) * | 2024-03-07 | 2024-04-09 | 山东大学 | Tunnel face trace crack extraction method and system based on three-dimensional point cloud |
| CN117854061B (en) * | 2024-03-07 | 2024-05-10 | 山东大学 | Tunnel face trace crack extraction method and system based on three-dimensional point cloud |
| CN118053152A (en) * | 2024-04-16 | 2024-05-17 | 中国地质大学（武汉） | Rock mass structural plane rapid measurement method and equipment based on mass point cloud data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117036998A (en) | Random forest-based open-air rock mass fracture surface identification method | |
| Polewski et al. | Marker-free coregistration of UAV and backpack LiDAR point clouds in forested areas | |
| US9947109B2 (en) | Method and apparatus for direct detection, location, analysis, identification, and reporting of vegetation clearance violations | |
| Ge et al. | Enhanced subpixel mapping with spatial distribution patterns of geographical objects | |
| CN110992341A (en) | Segmentation-based airborne LiDAR point cloud building extraction method | |
| CN103268358B (en) | Multi-source control point image database builds and update method | |
| JP6621445B2 (en) | Feature extraction device, object detection device, method, and program | |
| Chen et al. | Rapid urban roadside tree inventory using a mobile laser scanning system | |
| CN106529029B (en) | The point cloud data extracting method and device of electric power line pole tower | |
| CN112347894B (en) | Single plant vegetation extraction method based on transfer learning and Gaussian mixture model separation | |
| CN111462134A (en) | Single-tree segmentation method and system for fusing high-resolution remote sensing image and laser radar point cloud | |
| CN110798805B (en) | Data processing method and device based on GPS track and storage medium | |
| Yancho et al. | Fine-scale spatial and spectral clustering of UAV-acquired digital aerial photogrammetric (DAP) point clouds for individual tree crown detection and segmentation | |
| CN110910407B (en) | Street tree trunk extraction method based on mobile laser scanning point cloud data | |
| CN112529044B (en) | Method for extracting and classifying railway contact network based on vehicle-mounted LiDAR | |
| CN112419505A (en) | Vehicle-mounted point cloud road rod-shaped object automatic extraction method combining semantic rule and model matching | |
| CN113570005A (en) | Long-distance ship type identification method based on airborne photon radar | |
| CN110348478B (en) | Method for extracting trees in outdoor point cloud scene based on shape classification and combination | |
| CN109359680B (en) | Explosion sillar automatic identification and lumpiness feature extracting method and device | |
| CN113033386B (en) | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system | |
| CN105701856B (en) | A kind of vegetation extracting method and system | |
| CN117765006A (en) | Multi-level dense crown segmentation method based on unmanned aerial vehicle image and laser point cloud | |
| Xiaojuan et al. | Top-Down Approach to the Automatic Extraction of Individual Trees from Scanned Scene Point Cloud Data. | |
| Lalak et al. | Methodology of detection and classification of selected aviation obstacles based on UAV dense image matching | |
| CN115063698A (en) | Automatic identification and information extraction method and system for slope surface deformation crack |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |