CN113901968A - Railway point cloud feature extraction method based on machine learning - Google Patents
Railway point cloud feature extraction method based on machine learning Download PDFInfo
- Publication number
- CN113901968A CN113901968A CN202111487560.0A CN202111487560A CN113901968A CN 113901968 A CN113901968 A CN 113901968A CN 202111487560 A CN202111487560 A CN 202111487560A CN 113901968 A CN113901968 A CN 113901968A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- data
- extracting
- machine learning
- railway
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a machine learning-based railway point cloud feature extraction method, which comprises the following steps of: point cloud feature establishment: aiming at the point cloud image distribution, determining a feature extraction object; point cloud clipping: cutting the number of original point clouds in ten million levels, and processing partial point clouds; target segmentation: dividing the cut point cloud data by using DBSCAN clustering in machine learning, and extracting class I characteristics; feature extraction: processing the ground point cloud data by using a plane fitting algorithm based on RANSAC, and extracting class II features; and (3) filtering treatment: filtering the preliminarily extracted sleeper and rail data; fine extraction of the sleepers: and extracting key points, constructing a rectangular neighborhood, and finally extracting the sleeper points. The invention provides a sleeper extraction algorithm based on a rectangular neighborhood, which is used for searching and automatically generating key points, constructing a four-point rectangular neighborhood, and accurately extracting sleeper features by combining cross multiplication properties to achieve higher accuracy.
Description
Technical Field
The invention relates to the field of railway traffic management, in particular to a machine learning-based point cloud feature extraction method for ensuring railway traffic safety.
Background
The railway is an important national traffic facility and a major artery of national economic development, and the acceleration of railway development has become a consensus in all aspects of society. With the increasing number of railway miles in China, the nation has higher requirements for the safety and comfort of trains, and the attributes of railway engineering systems highly depend on the refined data requirements on the technical level, so that the three-dimensional scanning technology is developed vigorously with the powerful real-scene depicting function, and the three-dimensional point cloud data also becomes the hot spot of the current research. However, more advanced data information acquisition devices are also required to be matched with more powerful and diversified computational analysis technologies, and in recent years, the continuous development of computer technologies provides powerful technical support for analyzing and processing point cloud data, and can provide more accurate data support for railway transportation industry.
Machine learning is an important computer technology in the context of big data and is widely used in various fields. Before the machine learning era, experienced workers are often required to perform boring work of capturing characteristics by hand, and the efficiency is extremely low. The vigorous development of machine learning makes the machine learning become a popular technology for processing and analyzing three-dimensional point cloud data at present, and more meaningful information can be mined from disordered three-dimensional point cloud data through the machine learning technology. The extraction of the point cloud features is one of the important steps in the processing of the point cloud data, the application effect of the point cloud data is seriously influenced by the result of the extraction, and the application of a machine learning algorithm to the extraction of the point cloud features becomes popular research content at present. Therefore, a machine learning algorithm is used for carrying out a series of processing such as target segmentation, feature extraction and three-dimensional reconstruction on the railway point cloud data so as to improve the accuracy of point cloud data feature identification and improve the working efficiency of point cloud data three-dimensional reconstruction.
The point cloud feature extraction is an important part in the reverse reconstruction of the model, and becomes a research hotspot of numerous scholars at home and abroad in recent years.
Based on the traditional point cloud feature extraction method, many scholars continuously provide various feature extraction algorithms by combining with multi-field knowledge, and the method is widely applied to respective fields. Sterculia et al propose a characteristic line extraction algorithm of normal vector region clustering, which adopts PCA of a self-adaptive neighborhood to calculate normal vector, utilizes FCM algorithm of FA optimization to cluster the normal vector and realize effective segmentation, selects candidate characteristic points and extracts the characteristic points in the direction of a local neighborhood main shaft; baying Zhou et al propose a three-dimensional point cloud deformation monitoring method based on image information, convert the point cloud into two-dimensional gray scale image, construct the topological relation, then extract the gray scale image characteristic based on the maximum value technology of wavelet modulus, calculate the variable quantity according to the point cloud characteristic finally; yue Ma et al estimate by referencing the point cloud normal, extract the expected plane features of the point cloud data by using the RANSAC algorithm in PCL, and discuss the unmanned aerial vehicle positioning effect in combination with the plane features; jiajie Tan et al propose an enhanced feature extraction algorithm based on feature entropy, can realize that the best neighborhood is effectively extracted in a self-adaptive manner, and extract features by calculating and analyzing the feature entropy among a plurality of three-dimensional points, and the method is effective and has higher accuracy; zhang Jing et al propose a method for extracting scattered point cloud features based on Markov random field, quote the knowledge extraction feature points of curvature estimation, problem transformation, Bayes estimation, MAP-MRF framework, graph segmentation method and point labels, etc.; the Wangxiaohui et al provides a non-structural point cloud feature line extraction algorithm based on region clustering segmentation, can distinguish significant features in point cloud, and can well identify the feature of a feature intensity change curve.
Disclosure of Invention
The purpose of the invention is as follows: in order to construct a more accurate railway point cloud feature extraction algorithm model and improve the efficiency of feature recognition, a point cloud target segmentation is realized by using a DBSCAN clustering algorithm, feature extraction is realized based on a RANSAC algorithm, and finally, statistical filtering and rectangular neighborhood are adopted to improve the extraction precision.
In order to realize the method described above, the invention provides a railway point cloud feature extraction method based on machine learning, which comprises the following steps:
step1: point cloud feature establishment: aiming at the point cloud image distribution, determining a feature extraction object;
step 2: point cloud clipping: cutting the number of original point clouds in ten million levels, and processing partial point clouds;
step 3: target segmentation: dividing the cut point cloud data by using a DBSCAN clustering algorithm in machine learning, and extracting I-type features including vegetation, power lines, power line struts and the ground;
step 4: feature extraction: processing the ground point cloud data by using a plane fitting algorithm based on RANSAC, and extracting II types of characteristics, namely crossties and rails;
step 5: and (3) filtering treatment: filtering the sleeper and rail data extracted preliminarily, and filtering error points;
step 6: fine extraction of the sleepers: and extracting key points, constructing a rectangular neighborhood, and finally extracting the sleeper points.
Further, point cloud cutting in Step2 is to cut a railway power line strut as a node, and cut the point cloud into node data and non-node data;
further, in Step3, node data and non-node data are input in the DBSCAN clustering algorithm, data dimensions include XYZ three-dimensional coordinates and RGB color values, and radius parameters ϵ and neighborhood density threshold MinPts of the non-node data are 1.89 times and 1.8 times of the node data;
further, extracting local ground data selected by the rail based on a plane fitting algorithm of RANSAC in Step4, and extracting the rail characteristics at the highest layer; selecting data left after extracting the rails for extracting the sleepers, preliminarily extracting the characteristics of the middle-layer sleepers, wherein the data dimensions are XYZ three-dimensional coordinates and RGB color values, and the input parameters comprise a distance threshold value d, the number n of matching points and the iteration number k;
further, the filtering process in Step5 is a statistical filter process;
further, the Step6 is to extract key points by calculating the mean value in the sample data area of a single sleeper and generating the key points according to the fixity of the distance between sleepers; the construction of the rectangular neighborhood is obtained by continuously correcting the selection of key points in the process of constructing the rectangular neighborhood.
The invention has the beneficial effects that: and performing feature extraction on the railway point cloud data by using a machine learning algorithm, performing target segmentation processing on the original point cloud data by using a density clustering-based DBSCAN clustering algorithm, and extracting ground data containing class II features. Experimental images show that the DBSCAN clustering algorithm can effectively divide data of each part, can effectively filter noise points in the clustering process, and is good in effect. According to the method, plane fitting is carried out by combining the hierarchical characteristics of ground point cloud data through the RANSAC algorithm, data below a plane model are filtered through a proper threshold value, high-level feature data are extracted, outliers in a preliminary extraction result are filtered and filtered through statistical filtering, a rectangular neighborhood-based sleeper extraction algorithm is provided for ensuring sleeper feature extraction accuracy, and experiments show that the feature extraction accuracy is high.
Drawings
FIG. 1 is a diagram of the overall method steps;
FIG. 2 is a railway point cloud original image (after cropping);
FIG. 3 node data;
FIG. 4 non-node data;
FIG. 5 a railway point cloud DBSCAN target segmentation framework;
FIG. 6 color labeling image;
FIG. 7 is a cross-section of a rail;
FIG. 8 a ground data image;
FIG. 9 middle track area data image;
fig. 10 RANSAC extraction result (left), statistical filtering result (right);
FIG. 11 is a rectangular neighborhood based crosstie extraction flow;
FIG. 12 crosstie extraction based on rectangular neighborhoods;
FIG. 13 Final extraction results (overview);
FIG. 14 Final extraction results (ground details);
fig. 15 shows the final extraction result (sectional view).
Detailed Description
The method adopts a density clustering-based DBSCAN clustering algorithm to carry out target segmentation on initial point cloud, and extracts I-type characteristics such as vegetation, the ground, power lines, pillars and the like. Extracting class I features extracted by target segmentation, including class II features, and further extracting, aiming at the level characteristics of ground point cloud data, performing primary feature extraction on the point cloud of the track area by adopting an improved RANSAC algorithm, and separating out class II features (rails and sleepers); in order to improve the feature extraction precision, most outliers in the preliminary extraction result are filtered by using statistical filtering, a rectangular neighborhood-based sleeper extraction algorithm is provided, key points are searched and automatically generated, a four-point rectangular neighborhood is constructed, and sleeper features are accurately extracted by combining cross multiplication properties.
In general, this process can be used as shown in FIG. 1:
step1, point cloud characteristic establishment: scanning along the way by adopting an unmanned aerial vehicle carrying a LiDAR three-dimensional laser scanner by taking a rail as a route to obtain point cloud data, and determining objects of feature extraction aiming at the point cloud image distribution, wherein the objects comprise the rail, a power line pillar, the ground and vegetation on two sides of the rail;
step 2: point cloud clipping: the number of original point clouds is in the tens of millions level, and the method is used for verifying that the method is obviously low in efficiency, so that the point clouds are cut, partial point clouds are processed, and different parameter combinations of a DBSCAN clustering algorithm achieve different clustering effects, so that point cloud data are divided into node data and non-node data according to nodes in order to reduce the coupling between targets;
step 3: target segmentation: dividing the cut point cloud data by using a DBSCAN clustering algorithm in machine learning, and extracting I-type features including vegetation, power lines, power line struts and the ground;
step 4: feature extraction: rotating an original ground point cloud XOY plane by adopting a rotation matrix, then using slicing operation on a track data interval to obtain track area data, processing the ground point cloud data by adopting a plane fitting algorithm based on RANSAC, and adjusting parameters to effectively extract II-type features (rails and sleepers) belonging to different levels;
step 5: and (3) filtering treatment: carrying out statistical filtering processing on the sleeper and rail data extracted preliminarily, and filtering most error points;
step 6: fine extraction of the sleepers: and cutting the point cloud of the sleepers after statistical filtering, selecting a single sleeper with regular shape and uniform density as a reference sample, taking the average value point of all points in the sample area as a reference key point, roughly calculating the key point of any sleeper according to the parameter standard of the national railway construction track specification to obtain a key point set, automatically forming an accurate rectangular neighborhood according to the key point according to the national standard of the sleeper, and extracting the sleeper point according to the key point and the neighborhood by cross multiplication. And finally obtaining the sleeper characteristic points with improved precision.
The method includes the steps that large-scale point cloud data are collected by an unmanned aerial vehicle carrying a LiDAR three-dimensional laser scanner, the obtained point cloud data contain XYZ three-dimensional coordinates and RGB color values in data dimensions, an original image can be constructed, as shown in fig. 2, the original image is analyzed, and data characteristics can be divided as shown in a table 1. Since the method proposed in this patent is directed primarily to rails and crossties of the type ii, and other type ii features in table 1 are not studied herein, the type ii features in this specification refer to rails and crossties.
TABLE 1 railway Point cloud feature Classification
The point cloud cutting is to cut the railway power line support as a node and cut the node data and the non-node data. The point Cloud is clipped by using the Cloud company, and the point Cloud data is divided into node data and non-node data by using the power line struts as nodes, as shown in fig. 3 and 4. The point cloud data is distinguished according to nodes, the purpose is to reduce the coupling between targets, and because DBSCAN is density clustering, the clustering effect of the point cloud data with nodes under the conventional condition of utilizing the same group of global parameters is poor, even I-type characteristics can not be segmented, so that the clustering effect is optimized by adopting different parameter combinations through partitioning. The parameter settings are shown in table 2:
TABLE 2 DBSCAN parameters
| Data set | ϵ | MinPts |
| Non-node data | 1100 | 18 |
| Node data | 580 | 10 |
Inputting node data and non-node data in a DBSCAN clustering algorithm, wherein data dimensions comprise XYZ three-dimensional coordinates and RGB color values, and the radius parameter ϵ and the neighborhood density threshold MinPts of the non-node data are 1.89 times and 1.8 times of the node data.
In the processing process of point cloud data feature extraction, point cloud target segmentation is a process of independently extracting and processing objects which are obviously distinguished from point clouds, so that the aim of separating and treating is achieved. The method comprises the steps of taking node data and non-node data as input, conducting DBSCAN clustering to obtain vegetation points, power line points, ground points, support points and noise points, and in order to obtain a point cloud image which is easy to distinguish by human eyes, adding RGB data to data of the same type to conduct color labeling to obtain a final target segmentation result, wherein the specific flow is shown in figure 5, and the segmentation result is shown in figure 6.
According to the level characteristics of the track data image, as shown in fig. 7, namely, the track is higher than the sleeper and the sleeper is higher than the ground on the level, high-level data extraction is performed by adopting a plane fitting algorithm-RANSAC, and parameters are adjusted to effectively extract the track and the sleeper belonging to different levels primarily.
As shown in fig. 8, the railway ground includes three sections of tracks, i.e. left, middle and right, and does not belong to the same layer, i.e. there are multiple models, and RANSAC can only estimate a single model from a specific data set, so it is necessary to extract track region data. The ground data are subjected to coordinate transformation when accurate extraction of the track area is achieved, the rotation matrix is adopted to rotate the XOY plane of the original ground point cloud, the X axis is perpendicular to the track straight line of the area, the Y axis is parallel to the track straight line of the area, then the data are sequenced according to the X value, slicing operation is carried out on the track data interval, and the track area data are obtained, as shown in figure 9.
The RANSAC algorithm is applied to the orbit point cloud, and appropriate improvements are made in combination with the extraction requirements. Taking the data of the middle orbit area as an example, defining the data as a point set P, randomly selecting three points from P according to the basic principle of RANSAC algorithm, and extracting the three points from the point cloud data. Let the planar model equation be as follows.
The method comprises the following specific steps:
step1 randomly selecting three point great circles from point set P
}。
Step2 according to
The position of the plane model S is determined and the parameters A, B, C are determined, as follows.
。
Step3, counting the number of points on the plane S, and setting the plane thickness as a radius parameter set by the DBSCAN clustering algorithm
Calculating any point in P
Distance to plane S
，
The calculation is as follows.
Statistics of
The number of points (c) is expressed as a fraction score of the plane model S.
Step4, repeating the steps from Step1 to Step3 for N times, and selecting the plane model with the highest checking score
。
Step5, any point in the point set P
With its model in plane
Projected point on
Performing subtraction to obtain the result
As follows:
filtration
The upper layer data of the plane is preserved. The reason for doing this on the basis of the RANSAC algorithm is that the sleepers and rails extracted each time are highThe data of the layer can not be filtered when the RANSAC algorithm is used for fitting the plane layer data, and the bottom layer data which influences the extraction precision can be filtered through the operation, so that the accurate extraction is realized. The RANSAC algorithm specific parameter settings include a distance threshold d, the number n of matching points, and the number k of iterations, as shown in Table 3.
TABLE 3 RANSAC parameters
| Data set | d | n | k |
| Non-node data | 50 | 100 | 10000 |
| Node data | 20 | 100 | 10000 |
The features extracted by RANSAC often have more error points, so that the estimation of the local point cloud features becomes very complicated, which often results in erroneous estimation results, and further results in poor performance of the point cloud feature extraction accuracy. Therefore, the error point is further filtered by using the statistical filtering method, as shown in fig. 10.
The method has the advantages that the effect of extracting rails is good by filtering error points through statistical filtering, but more error points still exist in the extraction of sleepers, so that a sleeper extraction algorithm based on a rectangular neighborhood is provided, which mainly comprises the steps of extracting key points and constructing the rectangular neighborhood, wherein the key points are generated by calculating the mean value in a sample data area of a single sleeper and according to the distance fixity between sleepers; the construction of the rectangular neighborhood is obtained by continuously correcting the selection of key points in the process of constructing the rectangular neighborhood.
The extraction of key points is a core step for improving the extraction precision of the sleepers. The selection of the key points can be four vertexes or a central point of a rectangle, the central point is selected as a reference key point, only XY two-dimensional coordinates of the data points are analyzed, the extraction of the key points is performed by adopting a point cloud cutting and automatic generation mode, and the specific steps are as follows:
step1 point cloud cutting of the sleeper: and (3) utilizing the Cloud match point Cloud cutting function to cut the sleeper data after statistical filtering, and selecting a single sleeper with regular shape and uniform density as a reference sample.
Step2, calculating key points of the sample sleeper: the mean of all points within the sample area is taken as the first reference keypoint, i.e.
；
Step3 key point generation: because the state has strict parameter standards for the track specifications in railway construction, the key point of any sleeper can be roughly calculated according to the standards. Assume that the distance between the two rails isDThe distance between adjacent sleepers on a single rail beingMIf the reference sample is located on the No. 1 crosstie of the left rail (from top to bottom), the positions of the rest key points are expressed as follows, and the key point set is obtained through the stepsK。
In the national railway construction project, the distance between two rails of a rail, the specifications of sleepers and the distance between adjacent sleepers are all strictly regulated. Therefore, in the process of constructing the rectangular neighborhood of the sleeper, the accurate rectangular neighborhood can be automatically formed according to the key points by referring to the national standard of the sleeper, so that the accurate extraction of the sleeper points is realized. Suppose that the specifications of the crosstie are found by consulting: and if the length of a single sleeper is L and the width of the sleeper is W, the specific construction method comprises the following steps:
according to the key point setKKey point of (1)
Generating rectangular neighborhoods
Wherein:
according to key points and neighborhoods
) And (3) extracting the crosstie points, and judging whether a certain point is positioned in a rectangular neighborhood by using the property of cross multiplication, wherein the specific algorithm flow is shown in fig. 11.
And circularly traversing the initial sleeper point set, judging whether the sleeper points belong to the rectangular neighborhood, if so, setting the RGB of the sleeper points as other values, otherwise, continuously judging the next sleeper point until all sleeper points are traversed. Finally, the crosstie characteristic points with improved precision are obtained, as shown in fig. 12. The final extraction result is shown in fig. 13, and the detail is shown in fig. 14 and 15, and the results show that the extraction effect is good, and the features can be effectively distinguished.
The foregoing is only a specific embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (6)
1. A railway point cloud feature extraction method based on machine learning comprises the following steps:
step1: point cloud feature establishment: aiming at the point cloud image distribution, determining a feature extraction object;
step 2: point cloud clipping: cutting the number of original point clouds in ten million levels, and processing partial point clouds;
step 3: target segmentation: dividing the cut point cloud data by using a DBSCAN clustering algorithm in machine learning, and extracting I-type features including vegetation, power lines, power line struts and the ground;
step 4: feature extraction: processing the ground point cloud data by using a plane fitting algorithm based on RANSAC, and extracting II types of characteristics, namely crossties and rails;
step 5: and (3) filtering treatment: filtering the sleeper and rail data extracted preliminarily, and filtering error points;
step 6: fine extraction of the sleepers: and extracting key points, constructing a rectangular neighborhood, and finally extracting the sleeper points.
2. The method for extracting features of a railway point cloud based on machine learning of claim 1, wherein the point cloud clipping in Step2 is performed by using railway power line struts as nodes, and is clipped into node data and non-node data.
3. The method for extracting railway point cloud features based on machine learning as claimed in claim 1, wherein Step3 inputs node data and non-node data in DBSCAN clustering algorithm, data dimensions include XYZ three-dimensional coordinates and RGB color values, and radius parameters of non-node dataϵAnd neighborhood density thresholdMinPts1.89 times and 1.8 times of the node data.
4. The method for extracting point cloud features of railway based on machine learning as claimed in claim 1, wherein the RANSAC-based plane fitting algorithm in Step4 is used for extracting local ground data of railway track selection and extracting characteristics of railway track at the highest layer; the data used for extracting the sleepers is data left after the rails are extracted, the characteristics of the middle-layer sleepers are extracted preliminarily, the data dimensions are XYZ three-dimensional coordinates and RGB color values, and the input parameters comprise a distance threshold value d, the number n of matching points and the iteration number k.
5. The method for extracting features of a railway point cloud based on machine learning of claim 1, wherein the filtering process in Step5 is a statistical filter process.
6. The method for extracting point cloud features of railway based on machine learning as claimed in claim 1, wherein the Step6 is to generate key points by calculating the mean value in the sample data area of a single sleeper and according to the distance fixity between sleepers; the construction of the rectangular neighborhood is obtained by continuously correcting the selection of key points in the process of constructing the rectangular neighborhood.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111487560.0A CN113901968B (en) | 2021-12-08 | 2021-12-08 | Railway point cloud feature extraction method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111487560.0A CN113901968B (en) | 2021-12-08 | 2021-12-08 | Railway point cloud feature extraction method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113901968A true CN113901968A (en) | 2022-01-07 |
| CN113901968B CN113901968B (en) | 2022-02-25 |
Family
ID=79025639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111487560.0A Active CN113901968B (en) | 2021-12-08 | 2021-12-08 | Railway point cloud feature extraction method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113901968B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117994778A (en) * | 2024-04-07 | 2024-05-07 | 吉林大学 | Stamping part automatic counting and classifying method based on point cloud data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105701862A (en) * | 2014-11-28 | 2016-06-22 | 星际空间（天津）科技发展有限公司 | Ground object key point extraction method based on point cloud |
| CN110110802A (en) * | 2019-05-14 | 2019-08-09 | 南京林业大学 | Airborne laser point cloud classification method based on high-order condition random field |
| CN110189419A (en) * | 2019-05-27 | 2019-08-30 | 西南交通大学 | Vehicle-mounted Lidar rail data reduction method based on broad sense neighborhood height difference |
| CN110490918A (en) * | 2019-07-12 | 2019-11-22 | 中铁第六勘察设计院集团有限公司 | Existing railway track switch frog extracting method based on three-dimensional laser scanning technique |
| CN110647798A (en) * | 2019-08-05 | 2020-01-03 | 中国铁路设计集团有限公司 | Automatic track center line detection method based on vehicle-mounted mobile laser point cloud |
| CN110728689A (en) * | 2019-11-29 | 2020-01-24 | 中铁第六勘察设计院集团有限公司 | Existing railway center line extraction method based on point cloud data |
| CN112070769A (en) * | 2020-09-18 | 2020-12-11 | 福州大学 | Layered point cloud segmentation method based on DBSCAN |
| CN112962373A (en) * | 2021-02-20 | 2021-06-15 | 中国铁建重工集团股份有限公司 | Spike identification method based on line structured light and spike identification operation vehicle |
-
2021
- 2021-12-08 CN CN202111487560.0A patent/CN113901968B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105701862A (en) * | 2014-11-28 | 2016-06-22 | 星际空间（天津）科技发展有限公司 | Ground object key point extraction method based on point cloud |
| CN110110802A (en) * | 2019-05-14 | 2019-08-09 | 南京林业大学 | Airborne laser point cloud classification method based on high-order condition random field |
| CN110189419A (en) * | 2019-05-27 | 2019-08-30 | 西南交通大学 | Vehicle-mounted Lidar rail data reduction method based on broad sense neighborhood height difference |
| CN110490918A (en) * | 2019-07-12 | 2019-11-22 | 中铁第六勘察设计院集团有限公司 | Existing railway track switch frog extracting method based on three-dimensional laser scanning technique |
| CN110647798A (en) * | 2019-08-05 | 2020-01-03 | 中国铁路设计集团有限公司 | Automatic track center line detection method based on vehicle-mounted mobile laser point cloud |
| CN110728689A (en) * | 2019-11-29 | 2020-01-24 | 中铁第六勘察设计院集团有限公司 | Existing railway center line extraction method based on point cloud data |
| CN112070769A (en) * | 2020-09-18 | 2020-12-11 | 福州大学 | Layered point cloud segmentation method based on DBSCAN |
| CN112962373A (en) * | 2021-02-20 | 2021-06-15 | 中国铁建重工集团股份有限公司 | Spike identification method based on line structured light and spike identification operation vehicle |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117994778A (en) * | 2024-04-07 | 2024-05-07 | 吉林大学 | Stamping part automatic counting and classifying method based on point cloud data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113901968B (en) | 2022-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292276B (en) | Vehicle-mounted point cloud clustering method and system | |
| CN111192284B (en) | Vehicle-mounted laser point cloud segmentation method and system | |
| CN105957076B (en) | A kind of point cloud segmentation method and system based on cluster | |
| CN108510516A (en) | A kind of the three-dimensional line segment extracting method and system of dispersion point cloud | |
| CN111299815A (en) | Visual detection and laser cutting trajectory planning method for low-gray rubber pad | |
| CN106599915B (en) | A kind of vehicle-mounted laser point cloud classifications method | |
| Wang et al. | Speed and accuracy tradeoff for LiDAR data based road boundary detection | |
| CN102750449B (en) | Point cloud linear feature extraction method based on substep three-dimensional space and feature dimension mapping | |
| CN107610223A (en) | Power tower three-dimensional rebuilding method based on LiDAR point cloud | |
| CN102938066A (en) | Method for reconstructing outer outline polygon of building based on multivariate data | |
| CN104240251A (en) | Multi-scale point cloud noise detection method based on density analysis | |
| CN113744311A (en) | Twin neural network moving target tracking method based on full-connection attention module | |
| CN103336801A (en) | Multi-feature locality sensitive hashing (LSH) indexing combination-based remote sensing image retrieval method | |
| CN113901968B (en) | Railway point cloud feature extraction method based on machine learning | |
| CN111462017A (en) | Denoising method for tunnel laser point cloud data | |
| CN112070787B (en) | Aviation three-dimensional point cloud plane segmentation method based on opponent reasoning theory | |
| CN104765852A (en) | Data mining method based on fuzzy algorithm under big data background | |
| Xing et al. | An improved automatic pointwise semantic segmentation of a 3D urban scene from mobile terrestrial and airborne LiDAR point clouds: A machine learning approach | |
| Sun et al. | Automated segmentation of LiDAR point clouds for building rooftop extraction | |
| CN116452604A (en) | Complex substation scene segmentation method, device and storage medium | |
| Han et al. | Automated Extraction of Rail Point Clouds by Multi-Scale Dimensional Features From MLS Data | |
| CN101866483A (en) | Texture image segmentation method based on Lamarck multi-target immune algorithm | |
| Wu et al. | A Systematic Point Cloud Edge Detection Framework for Automatic Aircraft Skin Milling | |
| Kim et al. | Optimal Pre-processing of Laser Scanning Data for Indoor Scene Analysis and 3D Reconstruction of Building Models | |
| CN103530905A (en) | Elimination method for flat triangle area of contour line |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |