CN118376290A - Gaussian prediction coal pile volume calculation method based on point cloud registration of push harrow machine - Google Patents
Gaussian prediction coal pile volume calculation method based on point cloud registration of push harrow machine Download PDFInfo
- Publication number
- CN118376290A CN118376290A CN202410806274.3A CN202410806274A CN118376290A CN 118376290 A CN118376290 A CN 118376290A CN 202410806274 A CN202410806274 A CN 202410806274A CN 118376290 A CN118376290 A CN 118376290A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- cloud data
- radar
- coal pile
- data
- 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
- 239000003245 coal Substances 0.000 title claims abstract description 101
- 238000004364 calculation method Methods 0.000 title claims abstract description 44
- 238000012216 screening Methods 0.000 claims abstract description 10
- 230000002194 synthesizing effect Effects 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 33
- 230000008569 process Effects 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 description 13
- 238000001514 detection method Methods 0.000 description 8
- 238000001914 filtration Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000009616 inductively coupled plasma Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a Gaussian prediction coal pile volume calculation method based on a rake point cloud registration. In order to solve the problem that the coal pile volume calculation method in the prior art is limited by data volume; the invention comprises the following steps: a radar is installed on a rake pushing machine, and Lei Dadian cloud data are obtained; registering the radar point cloud data acquired each time with the map point cloud data, and outputting positioning information; fusing radar point cloud data with positioning information, and incrementally synthesizing a cabin map; projecting radar point cloud data to a Z axis, screening cabin boundary point cloud data through variance, and calculating the inclination angle of the cabin boundary point cloud data to distinguish cabin walls and coal piles; and carrying out Gaussian region prediction on the obtained radar point cloud data, dividing grids for the coal pile point cloud data, and obtaining the coal pile volume in the cabin by calculating the coal pile volume in each grid. And registering the point cloud to obtain positioning information, then constructing a grid by utilizing Gaussian prediction, calculating the volume, and improving the accuracy of volume calculation.
Description
Technical Field
The invention relates to the field of port automation, in particular to a Gaussian prediction coal pile volume calculation method based on point cloud registration of a push rake machine.
Background
In a cabin environment, accurate calculation of the coal pile volume is a critical task. This calculation process mainly relies on high precision laser sensors mounted on the pushing and raking machine. Through rapid scanning, the sensors can capture three-dimensional space information in the cabin in detail, and then a three-dimensional model of the coal pile is built. Based on this model, the volume of the coal pile inside the hold can be accurately calculated.
The calculation of the volume of the coal pile in the cabin has non-negligible guiding significance for key equipment in port operation, such as grab buckets and screw ship unloaders. Accurate volume data helps the devices to more accurately know the distribution condition and morphological characteristics of the coal pile in the cabin, thereby optimizing the operation flow and reducing ineffective operation.
Meanwhile, accurate volume data can also provide decision support for port management departments, and help the port management departments to more scientifically plan and manage coal transportation work. Through optimizing loading, unloading and transferring processes of coal, the port working efficiency can be improved, energy consumption and environmental pollution can be reduced, and green and efficient port operation is realized.
However, in the volume calculation under the cabin environment, the measurement method is often limited by various factors, such as low measurement precision, complex operation and the like, and the requirements of high efficiency and accuracy of modern ports are difficult to meet. Therefore, it is very urgent to develop a more advanced and efficient method for calculating the volume of the coal pile.
For example, in the prior art, a method and a system for calculating a volume of a stockyard stockpile disclosed in the chinese patent literature, the publication number CN115018903a collects stockpile data by using a camera, and uses a deep learning model to classify and identify the stockpile and calculate the volume, but the differences of colors of coal piles in a cabin environment are smaller and are black or gray, and a strategy using a camera cannot cope with the calculation of the volume of the coal piles in the cabin environment.
Other bulletin numbers disclosed in the chinese patent literature are: method and system for calculating volume of materials in bin based on point cloud data in CN117152239A, with the bulletin number: the three-dimensional object volume calculation method based on the point cloud slice of CN110889885A calculates the volume by adopting a triangular mesh method and a plane slice method respectively, but the influence of the condition of point cloud vacancy on the volume calculation under the condition of point cloud sparseness is not considered.
Disclosure of Invention
The invention mainly solves the problem that the coal pile volume calculation method in the prior art is limited by data volume; the utility model provides a calculation method of Gaussian prediction coal pile volume based on the point cloud registration of a push harrow machine, positioning information is obtained by registering the point cloud, and then grid construction and volume calculation are carried out by utilizing Gaussian prediction.
The technical problems of the invention are mainly solved by the following technical proposal:
A Gaussian prediction coal pile volume calculation method based on a push harrow machine point cloud registration comprises the following steps:
S1: a radar is installed on a rake pushing machine, and Lei Dadian cloud data are obtained;
S2: resolving the radar point cloud data to obtain map point cloud data, registering the radar point cloud data obtained each time with the map point cloud data, and outputting positioning information;
S3: fusing radar point cloud data with positioning information, and incrementally synthesizing a cabin map;
S4: projecting radar point cloud data to a Z axis, screening cabin boundary point cloud data through variance calculation and comparison, and calculating the inclination angle of the cabin boundary point cloud data to distinguish cabin wall point cloud and coal pile point cloud;
S5: and carrying out Gaussian region prediction on the obtained radar point cloud data, dividing network grids for the coal pile point cloud data, and obtaining the coal pile volume in the cabin by calculating the coal pile volume in each network grid.
According to the scheme, other sensors are not required to be installed on the cabin and the rake pushing machine, only point cloud data of the laser radar are required, and data fusion difficulty caused by multiple devices is avoided. Registration is carried out on point clouds to obtain positioning information, then grid construction is carried out by utilizing Gaussian prediction, and the volume is calculated, so that point cloud data at the sparse position of the point clouds can be predicted by utilizing Gaussian prediction, and the accuracy of volume calculation is improved; and projecting the point cloud data to a Z axis, and judging the boundary between the bulkhead and the coal pile by calculating and comparing the variance and the inclination angle of the Z axis, so as to remove the interference of the bulkhead on the calculation of the volume of the coal pile.
Preferably, in step S4, the specific process of distinguishing the bulkhead and the coal pile point cloud data includes:
S4.1: projecting all the radar point cloud data to a Z axis, and screening the radar point cloud data of the bulkhead of the ship by calculating variance;
s4.2: taking the point cloud data of the screened bulkhead and the point cloud data in the rated distance of the periphery of the bulkhead as boundary point cloud data, and adopting polar coordinate conversion to perform dimension reduction treatment on the boundary point cloud data;
s4.3: and calculating the inclination angle of the boundary point cloud data, and distinguishing the boundary line between the bulkhead and the coal pile by comparing the inclination angle.
Preferably, dividing a plurality of grids, projecting all the radar point cloud data in the grids to a Z axis respectively, and calculating the Z value variance of all the radar point cloud data in each grid;
setting a boundary point cloud variance threshold, and screening a plurality of grids with Z value variance larger than the boundary point cloud variance threshold as bulkhead point cloud grids;
And calculating the continuity of the screened cabin wall point cloud grids, and taking the corresponding point cloud data in the cabin wall point cloud grids as radar point cloud data of the cabin wall when the cabin wall point cloud grids are in the continuous condition.
The vertical length of the cabin wall is shortest 4 meters, and the boundary of the cabin is a boundary difference point of the coal pile under the condition that other areas are non-vertical, so that the boundary point cloud variance threshold is set to be 4; the point cloud grid with variance exceeding 4 is taken as the bulkhead point cloud grid.
Preferably, three-dimensional coordinates of the boundary point cloud data are converted into polar coordinates;
Calculating the inclination angle of the boundary point cloud data according to the polar coordinates of the boundary point cloud data;
if the inclination angle of the boundary point cloud data is larger than the maximum repose angle of the coal pile, the boundary point cloud data is cabin wall point cloud, otherwise, the boundary point cloud data is coal pile point cloud.
The angle θ of the bulkhead is approximately 90 degrees, and the angle θ at the boundary is set to 50 degrees according to the repose angle of the coal pile at the maximum of 50 degrees.
Preferably, the expression of the gaussian region prediction is:
Wherein, Is a sample point; Is a predicted value.
Representing sparse point cloud data,And (5) the point cloud data after the filling.
Is point cloud noise;
Is a predicted value;
Is the variance of the predicted value;
Is a kernel function.
The point cloud sparsity is a common defect of all radars, is particularly obvious on mechanical radars, and cannot achieve the whole restoration like a camera. And carrying out Gaussian three-dimensional reconstruction on the acquired laser radar point cloud data, selecting the existing radar point cloud data as a sample set and a test set, and generating predicted data distribution.
Preferably, in step S5, network raster division is performed on the radar point cloud data after gaussian region prediction reconstruction;
Traversing all network grids in the map data, and calculating the volume of the coal pile in the single columnar network grid according to the value of the highest point and the lowest value of the coal pile point cloud in each network grid;
And accumulating the volumes of the coal piles in all columnar network grids to obtain the total volume of the coal piles in the cabin.
Preferably, the expression of the Lei Dadian cloud data and map point cloud data registration process is as follows:
Wherein, Positioning information for registering ith radar point cloud data and map point cloud data;
Covariance of map point cloud data;
Covariance of ith radar point cloud data;
the error between the ith radar point cloud data and the target point cloud data;
Error of The expression of (2) is:
Wherein, The radar point cloud data obtained for the ith time;
is map point cloud data.
Converting the radar point cloud information acquired once into PCL communication information through the ROS topic information; registering radar point cloud data in the radar point cloud information with map point cloud data by adopting PCL.
Preferably, the first radar point cloud data is stored as map point cloud data, and the second radar point cloud data is registered with the first radar data;
registering the positioning information after the second time for registering the radar point cloud data and the map point cloud data acquired each time.
Because the radar is arranged on the top of the push harrow machine, the radar point cloud information obtained by scanning the radar comprises radar point cloud data of a detection target and map point cloud data around the push harrow machine.
Preferably, in the radar point cloud data acquired by the first scanning of the radar, the radar point cloud data of the detection target cannot be distinguished from the map point cloud data, that is, when the first point cloud information is aligned, the map point cloud data has no data, and positioning registration cannot be performed; therefore, the first positioning information is set as an identity matrix, the rotation heading is 0, the roll is 0, and the pitch angle is 0.
Preferably, multiplying the radar point cloud data acquired each time with the corresponding registered positioning information to acquire registered point cloud data of corresponding acquisition times; and accumulating all the registered point cloud data to form cabin map data.
Because the point cloud information obtained by the radar has sparsity, single radar point cloud data cannot be subjected to volume calculation, map registration is performed by converting single Lei Dadian cloud data into point cloud of a world coordinate system, and the registered point cloud data is inserted into a map to perform increment generation of cabin map data.
The beneficial effects of the invention are as follows:
1. other sensors are not required to be installed on the cabin and the rake pushing machine, only point cloud data of the laser radar are required, and data fusion difficulty caused by multiple devices is avoided.
2. Point cloud data at the sparse position of the point cloud can be predicted by Gaussian prediction, and accuracy of volume calculation is improved.
3. And projecting the point cloud data to a Z axis, and judging the boundary between the bulkhead and the coal pile by calculating and comparing the variance and the inclination angle of the Z axis, so as to remove the interference of the bulkhead on the calculation of the volume of the coal pile.
Drawings
FIG. 1 is a flow chart of a method for calculating the volume of a coal pile according to the present invention.
Fig. 2 is an original point cloud data map of the present invention.
Fig. 3 is a graph of point cloud data of the present invention with the bulkhead removed.
Fig. 4 is a schematic diagram of the volumetric calculation of the present invention.
Fig. 5 is a graph of the measurement results of the present invention.
Detailed Description
The technical scheme of the invention is further specifically described below through examples and with reference to the accompanying drawings.
Examples:
The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine in the embodiment is shown in fig. 1, and comprises the following steps:
S1: and (5) mounting a radar on the push harrow machine to obtain radar point cloud data.
The radar is arranged at the top of the rake pushing machine, radar point cloud data are collected along with the operation of the rake pushing machine, map point cloud data around the rake pushing machine are collected, and a control system of the rake pushing machine obtains radar point cloud information of the radar through topic communication in ROS communication.
In the embodiment, the specifically adopted radar is MID-360 radar, and the bulk coal cargo ship berthed at the wharf is tested.
When the push harrow machine starts to operate, the radar at the top of the push harrow machine is synchronously started, and the acquisition of radar point cloud data around the push harrow machine is started. These radar point cloud data are formed by radar waves emitted by radar reflecting back after hitting surrounding objects, and they contain rich spatial information such as the shape, size, position, etc. of the objects. By collecting the data in real time, three-dimensional point location information of the operating area of the scraper can be obtained, and an important basis is provided for subsequent coal pile volume calculation.
In the data transmission process, the topic communication mechanism in ROS (Robot Operating System) communication is utilized to acquire radar point cloud information, so that the real-time property and accuracy of data are ensured. ROS is a flexible framework designed specifically for robotic software development that provides rich libraries and tools that enable robotic developers to easily communicate, hardware abstract, package management, code reuse, etc.
S2: registering the radar point cloud data with the map point cloud data, and outputting positioning information.
Converting the radar point cloud information acquired once into PCL communication information through the ROS topic information; registering radar point cloud data in the radar point cloud information with map point cloud data by adopting PCL.
The scheme of the embodiment does not depend on a specific radar model or scanning mode, and can adapt to radar point cloud data processing requirements in different scenes. Meanwhile, as the PCL (Point Cloud Library) open source library is adopted for point cloud data processing, the point cloud data processing method can be conveniently integrated into other robots or automatic driving systems.
The expression of Lei Dadian cloud data and map point cloud data registration process is:
Wherein, Positioning information for registering ith radar point cloud data and map point cloud data;
Covariance of ith radar point cloud data;
Covariance of map point cloud data;
the error between the ith radar point cloud data and the target point cloud data;
Error of The expression of (2) is:
Wherein, The radar point cloud data obtained for the ith time;
is map point cloud data.
Because the radar is installed on the top of the rake, the radar point cloud information obtained by scanning the radar is composed of radar point cloud data of a detection target and map point cloud data around the rake, and specifically, in this embodiment, the radar point cloud data of the detection target is radar point cloud data of a coal pile, and the map point cloud data around the rake is map point cloud data of a cabin.
In radar point cloud data acquired by radar first scanning, the radar point cloud data of a detection target cannot be distinguished from map point cloud data, namely, when the first point cloud information is aligned, the map point cloud data has no data, and positioning registration cannot be performed; thus, the first time of positioning informationThe pitch angle is 0 when the pitch angle is set as an identity matrix, the rotation heading is 0, the roll is 0, and the pitch angle is 0.
Since the map point cloud data is empty at the time of the first scanning, the scheme of the embodiment can handle the situation of an unknown environment by assuming an initial pose (identity matrix) and taking the point cloud of the first scanning as an initial map. With the increase of the scanning times, the map information is gradually enriched, and the positioning accuracy is also improved.
Storing the first radar point cloud information as map point cloud data, registering the second radar point cloud information with the first radar information, wherein the expression is as follows:
registering the registration positioning information after the second time for the radar point cloud data and the map point cloud data acquired each time, wherein the expression is as follows:
by registering radar point cloud data and map point cloud data, the scheme of the embodiment realizes real-time positioning and map construction (SLAM). With continuous scanning of the radar, the map point cloud data are gradually perfected, so that the positioning accuracy and the positioning robustness are improved.
By continuously iterating and optimizing the registration process, the error between Lei Dadian cloud data and map point cloud data can be gradually reduced, so that the positioning accuracy is improved. In addition, the covariance information of the point cloud data is considered to be helpful for reasonably distributing weights of different point cloud data in the registration process, so that the accuracy of registration is further improved.
In the specific application scene of the push harrow machine, the scheme of the embodiment can realize accurate detection and positioning of the coal pile and provide powerful support for automatic operation of the push harrow machine. Meanwhile, due to the specificity of the cabin environment (such as narrow space, insufficient light and the like), the scheme also shows the advantages of point cloud data processing and positioning in a complex environment.
S3: and fusing the current radar point cloud data with positioning information, and incrementally synthesizing a cabin map.
Because the point cloud information obtained by the radar has sparsity, single radar point cloud data cannot be subjected to volume calculation, map registration is performed by converting single Lei Dadian cloud data into point cloud of a world coordinate system, and the registered point cloud data is inserted into a map to perform increment generation of cabin map data.
And (3) carrying out continuous radar point cloud data acquisition by using radar equipment arranged on the top of the push harrow machine. Because the radar point cloud data has sparsity, the data acquired once cannot be directly used for volume calculation, and therefore, the data needs to be processed through subsequent steps.
The collected radar point cloud data needs to be preprocessed, and the method comprises the steps of removing noise points, filtering invalid data and the like, so that the accuracy of subsequent registration and map generation is ensured.
And converting the single-time acquired radar point cloud data into a world coordinate system, and registering with the existing map point cloud data. In the registration process, an optimization algorithm (such as ICP (inductively coupled plasma), NDT (non-linear transformation) and the like) is utilized to minimize the difference between the radar point cloud and the map point cloud, so that the registered positioning information is obtained。
From the registered positioning informationAnd corresponding radar point cloud dataAccumulating according to the following expression to generate cabin Map data Map, wherein the expression for generating the cabin Map data in an increment mode is as follows:
Wherein, Positioning information for registering ith radar information and map data;
the radar point cloud data acquired for the ith time are acquired;
N is the total number of times of accumulating the registered point cloud data in the present embodiment.
Multiplying the radar point cloud data acquired each time with the corresponding registered positioning information to acquire registered point cloud data of corresponding acquisition times; and accumulating all the registered point cloud data to form cabin Map data Map.
With the continuous collection and registration of radar data, cabin Map data Map is gradually perfected. In order to further improve the accuracy and reliability of the map, a filtering algorithm (such as Gaussian filtering, median filtering and the like) can be adopted to carry out smoothing processing on the map; meanwhile, the accumulated errors in the map are detected and corrected by using loop detection and other technologies.
The generated cabin map data can also be used for various aspects of automatic operation of the push harrow machine, path planning, obstacle detection and the like. By updating the map data in real time, the rake pushing machine can more accurately sense the surrounding environment, and the working efficiency and the safety are improved.
In this embodiment, the second registration of positioning informationThe translation of (2) is: 0.052484270-0.002348415-0.000555588;
Positioning information for nth registration The rotation quaternions of (a) are respectively: 0.002983760-0.001181009 0.193365067 0.981128871.
The cabin environment is usually narrow and the light is insufficient, which brings certain difficulties to the operation of the push harrow machine. The cabin map data generated by the embodiment can clearly display information such as structures, obstacles and the like in the cabin, can calculate the volume of the coal pile more accurately, and simultaneously provides accurate navigation and positioning support for the push harrow machine, so that the push harrow machine can perform efficient and safe operation in a complex environment.
S4: and removing the cloud interference of the bulkhead point of the ship, and identifying the coal pile.
The point cloud data of the cabin wall can interfere with the volume calculation of the coal pile, other point cloud data can be scanned by the MID-360 radar adopted in the embodiment due to the wide pitch angle, the cabin map data are required to be preprocessed, the point cloud outside the cabin is removed, and the point cloud data of the coal pile are reserved.
In this embodiment, through filtering of the Z axis is established, and point cloud data outside the cabin is removed. The method specifically comprises the following steps:
s4.1: and projecting all the radar point cloud data to a Z axis, and screening the radar point cloud data of the bulkhead of the ship by calculating variance.
Dividing a plurality of grids, projecting all the radar point cloud data in the grids to a Z axis respectively, and calculating the Z value variance of all the radar point cloud data in each grid.
Setting a boundary point cloud variance threshold, and screening out a plurality of grids with variances larger than the boundary point cloud variance threshold to form a cabin boundary.
In the embodiment, the shortest vertical length of the cabin wall is 4 meters, and the boundary of the cabin is a coal pile boundary difference point under the condition that other areas are non-vertical, so that the variance is larger, and the threshold value of the variance of the boundary point cloud is set to be 4; the point cloud grid with variance exceeding 4 is taken as the bulkhead point cloud grid.
And calculating the continuity of the screened cabin wall point cloud grids, wherein in the embodiment, the cabin wall point cloud information is only the condition that the three-dimensional grids are continuous.
Furthermore, an iterative method or a sliding window method can be adopted to dynamically adjust the grid size and the boundary point cloud variance threshold value so as to adapt to cabins with different shapes and sizes, and the point cloud data of the cabin walls can be accurately identified.
S4.2: and performing polar coordinate conversion dimension reduction processing on the determined cabin wall point cloud data and the peripheral point cloud data.
And searching the screened and determined cabin wall point cloud information and radar point cloud information within a range of 1m around the screened and determined cabin wall point cloud information, and reducing the dimension according to a polar coordinate method.
The expression of the polar transformation is:
Wherein, A polar representation of radar point cloud data;
a spatial three-dimensional coordinate representation of radar point cloud data;
、 is the minimum resolution.
Can be calculated according to the above formulaIs used, in this embodiment,、The minimum resolutions of (2) are set to 0.5 and 0.5, respectively.
Further, in polar conversion, in addition to using a fixed minimum resolution、In addition, the resolution can be dynamically adjusted according to the density and distribution of the point cloud, so that the conversion accuracy and efficiency can be improved. In addition, in order to ensure the accuracy of the polar coordinate conversion, the original point cloud data needs to be subjected to necessary smoothing processing to reduce the influence of noise and errors.
S4.3: and calculating the inclination angle of the point cloud at the boundary of the cabin, and distinguishing the boundary line of the cabin wall and the coal pile by comparing the inclination angle.
Considering that the cabin wall point cloud and the coal pile point cloud of the cabin have larger vertical inclination change at the boundary, adopting the following steps to respectively calculate the inclination interpolation of the upper point cloud and the lower point cloud at the boundary:
Wherein, Is the inclination angle of the point cloud;
Is the coordinate value of the point cloud.
The larger value of the screening angle is the boundary between the bulkhead and the coal pile, in this embodiment the angle of the bulkheadTo approach 90 degrees, the angle at the boundaryAnd setting the demarcation threshold value of the coal pile to be 50 degrees according to the maximum 50 degrees of the repose angle of the coal pile, and recording the polar coordinate form of the coal pile.
Further, in calculating the tilt angle, more complex algorithms, such as least squares or polynomial fitting, may be used to more accurately fit the point cloud data at the boundary, thereby obtaining a more accurate tilt angle.
Meanwhile, the determination of the dividing line also needs to consider the actual situation of the coal pile, such as the shape, the size, the repose angle and the like of the coal pile, so as to ensure the accuracy and the reliability of the dividing line.
S4.4: and (5) classifying point clouds.
From the boundary point cloud recorded by the polar coordinates in step S4.3, the boundary point cloud is classified up and down, the point cloud above the demarcation line is the cabin wall point cloud, and the point cloud below the demarcation line is the coal pile bottom point cloud, as shown in fig. 2 and 3, the Lei Dadian cloud data of the cabin is divided into the coal pile point cloud data and the cabin wall point cloud data, and the point cloud data of the cabin wall cannot be used for performing volume calculation so as to ensure the accuracy of the volume calculation of the coal pile.
Further, in addition to simple up-down classification, more complex classification algorithms, such as clustering algorithms or machine learning algorithms, may be considered for more accurately identifying point cloud data for coal piles and bulkheads.
In addition, after classification is completed, further post-processing, such as filtering, smoothing, interpolation and the like, can be performed on the point cloud data of the coal pile so as to improve the accuracy and precision of the calculation of the volume of the coal pile.
According to the scheme, the cabin map data are preprocessed, point clouds outside the cabin are removed, and the point cloud data of the coal pile are reserved, so that the accuracy of the calculation of the volume of the coal pile is improved. Through accurate recognition and removal of the cabin wall point cloud, interference of the cabin wall on calculation of the volume of the coal pile is avoided, and therefore accuracy of calculation of the volume of the coal pile is improved. The polar coordinate conversion and dimension reduction processing are adopted, so that the calculated amount is reduced, and the processing efficiency is improved. Meanwhile, parameters and algorithm selection are dynamically adjusted, so that the processing efficiency and accuracy are further improved. The scheme of the embodiment can adapt to cabins of different shapes, sizes and coal pile types, and has strong adaptability and flexibility.
S5: and carrying out Gaussian region prediction on the obtained radar point cloud data, and supplementing a part with Ji Leida point cloud data sparseness.
The point cloud sparsity is a common defect of all radars, is particularly obvious on mechanical radars, and cannot achieve the whole restoration like a camera. Carrying out Gaussian three-dimensional reconstruction on the obtained laser radar point cloud data, selecting the existing radar point cloud data as a sample set and a test set, and generating predicted data distribution, wherein the expression of the predicted point cloud data distribution is as follows:
and extracting point clouds in each voxel grid, dividing the point clouds into prediction data according to the maximum and minimum values, wherein the existing point clouds are training data, a single point cloud cannot be predicted, and grids with less than a certain number cannot be subjected to Gaussian prediction.
Wherein,As a point of the sample,Is a predicted value; specifically, g is the original point cloud data,Traversing each voxel unit to scale into predicted values, the predicted third dimension data is absent, asValues.
The table is normally distributed.
Representing sparse point cloud data,The point cloud data after the filling is obtained; Is the x, y and z values of a point cloud, Representing the error value of the normal distribution.
As the point cloud noise, the point cloud noiseA fixed value is adopted, and a sensor manual can be checked;
In order to be able to predict the value, Is a three-dimensional matrix of samples,Is an identity matrix;
Is the variance of the predicted value;
representing the difference of kernel functions of the predicted point cloud and the training point cloud, Interpolation for training a point cloud kernel function;
As a kernel function to By way of example only,The calculation formula is as follows: The calculation formula is as follows: ；
the surface of the coal pile is perpendicular to the XOY plane in the three-dimensional coordinate system, and the Gaussian processing function is Other 2 axes, Without gaussian processing.
In this embodiment, all the coal pile point clouds are divided into 0.5 cubic grids, gaussian processing is performed, 10 resolutions are acquired according to the size of 0.05 of each grid, prediction processing is performed, and a test set and a prediction set are generated.
And S6, performing grid connection on the reconstructed point cloud, calculating the height data of each grid, and accumulating to obtain the volume of the coal pile.
As shown in the volume calculation schematic diagram of the grids in fig. 4, all grids in the map data are traversed, and the volume of the single columnar grid and the coal pile volume are calculated according to the highest point value and the lowest value.
The calculation expression for calculating the volume of a single columnar grid is:
the volume of the coal pile is calculated by accumulating the volumes of all columnar grids, and the expression of the calculation of the volume of the coal pile is as follows:
Wherein, Is the minimum value of the j-th grid height;
Is the maximum value of the j-th grid height;
The bottom area of the jth grid;
The volume value of the jth grid;
and calculating the volume of the point cloud of the coal pile by grids for the volumes of all the coal piles, and accumulating to obtain the final volume.
According to the scheme, other sensors are not required to be installed on the cabin and the push rake machine, only the point cloud data of the laser radar are required, and the data fusion difficulty caused by multiple devices is avoided. The map data can be generated online, positioning information is generated in real time according to each point cloud input, and cabin maps are generated in an increment mode.
The traditional laser radar wire harness is fixed, point cloud data at the sparse position of the point cloud can be predicted by adopting Gaussian prediction, and accuracy of volume calculation is improved. And finally, the volume calculation is performed by adopting a grid method in the scheme of the embodiment, and the method has higher precision than the grid calculation method.
Specifically, the size of the cabin in the implementation case is 20, 32 and 32 meters, the direct filtering of the Z value is established, the maximum value is 40, the minimum value is-5, and point clouds outside the cabin are removed.
In this embodiment, the second registration of positioning informationThe translation of (2) is: 0.052484270-0.002348415-0.000555588; positioning information for nth registrationThe rotation quaternions of (a) are respectively: 0.002983760-0.001181009 0.193365067 0.981128871.
The solution of this embodiment is used to calculate the volume of the coal pile for the cabin example of the above parameters, and the measurement result diagram of this embodiment as shown in fig. 5 is obtained, where the measurement result is about 950 cubic meters.
It should be understood that the examples are only for illustrating the present application and are not intended to limit the scope of the present application. Furthermore, it should be understood that various changes and modifications can be made by one skilled in the art after reading the teachings of the present application, and such equivalents are intended to fall within the scope of the application as defined in the appended claims.
Claims (10)
1. A Gaussian prediction coal pile volume calculation method based on a point cloud registration of a push harrow machine is characterized by comprising the following steps:
S1: a radar is installed on a rake pushing machine, and Lei Dadian cloud data are obtained;
S2: resolving the radar point cloud data to obtain map point cloud data, registering the radar point cloud data obtained each time with the map point cloud data, and outputting positioning information;
S3: fusing radar point cloud data with positioning information, and incrementally synthesizing a cabin map;
S4: projecting radar point cloud data to a Z axis, screening cabin boundary point cloud data through variance calculation and comparison, and calculating the inclination angle of the cabin boundary point cloud data to distinguish cabin wall point cloud and coal pile point cloud;
S5: and carrying out Gaussian region prediction on the obtained radar point cloud data, dividing network grids for the coal pile point cloud data, and obtaining the coal pile volume in the cabin by calculating the coal pile volume in each network grid.
2. The method for calculating the volume of the coal pile predicted based on the point cloud registration of the push harrow according to claim 1, wherein in the step S4, the specific process of distinguishing the bulkhead and the point cloud data of the coal pile comprises the following steps:
S4.1: projecting all the radar point cloud data to a Z axis, and screening the radar point cloud data of the bulkhead of the ship by calculating variance;
s4.2: taking the point cloud data of the screened bulkhead and the point cloud data in the rated distance of the periphery of the bulkhead as boundary point cloud data, and adopting polar coordinate conversion to perform dimension reduction treatment on the boundary point cloud data;
s4.3: and calculating the inclination angle of the boundary point cloud data, and distinguishing the boundary line between the bulkhead and the coal pile by comparing the inclination angle.
3. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine according to claim 1 or 2, wherein a plurality of grids are divided, all radar point cloud data in the grids are projected to a Z axis respectively, and the Z value variance of all the radar point cloud data in each grid is calculated;
setting a boundary point cloud variance threshold, and screening a plurality of grids with Z value variance larger than the boundary point cloud variance threshold as bulkhead point cloud grids;
And calculating the continuity of the screened cabin wall point cloud grids, and taking the corresponding point cloud data in the cabin wall point cloud grids as radar point cloud data of the cabin wall when the cabin wall point cloud grids are in the continuous condition.
4. The method for calculating the volume of the Gaussian predicted coal pile based on the point cloud registration of the push harrow machine according to claim 3, wherein three-dimensional coordinates of boundary point cloud data are converted into polar coordinates;
Calculating the inclination angle of the boundary point cloud data according to the polar coordinates of the boundary point cloud data;
if the inclination angle of the boundary point cloud data is larger than the maximum repose angle of the coal pile, the boundary point cloud data is cabin wall point cloud, otherwise, the boundary point cloud data is coal pile point cloud.
5. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine according to claim 1 or 4, wherein the expression of Gaussian region prediction is:
；
Wherein, Is a sample point; Is a predicted value;
Representing sparse point cloud data;
is point cloud noise;
Is a predicted value;
Is the variance of the predicted value;
Is a kernel function.
6. The method for calculating the volume of the Gaussian prediction coal pile based on the point cloud registration of the push harrow machine according to claim 5, wherein in the step S5, network grid division is performed on radar point cloud data after Gaussian region prediction reconstruction;
Traversing all network grids in the map data, and calculating the volume of the coal pile in the single columnar network grid according to the value of the highest point and the lowest value of the coal pile point cloud in each network grid;
And accumulating the volumes of the coal piles in all columnar network grids to obtain the total volume of the coal piles in the cabin.
7. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the rake machine according to claim 1, wherein the expression of the registration process of radar point cloud data and map point cloud data is as follows:
；
Wherein, Positioning information for registering ith radar point cloud data and map point cloud data;
Covariance of ith radar point cloud data;
Covariance of map point cloud data;
the error between the ith radar point cloud data and the target point cloud data;
Error of The expression of (2) is:
；
Wherein, The radar point cloud data obtained for the ith time;
is map point cloud data.
8. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine according to claim 1 or 7, wherein the radar point cloud data of the first time is stored as map point cloud data, and the radar point cloud data of the second time is registered with the radar data of the first time;
registering the positioning information after the second time for registering the radar point cloud data and the map point cloud data acquired each time.
9. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine, which is disclosed in claim 8, is characterized in that the first positioning information is set as an identity matrix, the rotation heading is 0, the roll is 0, and the pitch angle is 0.
10. The method for calculating the Gaussian prediction coal pile volume based on the point cloud registration of the push harrow machine according to claim 1, 7 or 9, wherein the radar point cloud data acquired each time is multiplied by the corresponding registration positioning information to acquire the registered point cloud data of corresponding acquisition times; and accumulating all the registered point cloud data to form cabin map data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410806274.3A CN118376290B (en) | 2024-06-21 | Gaussian prediction coal pile volume calculation method based on point cloud registration of push harrow machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410806274.3A CN118376290B (en) | 2024-06-21 | Gaussian prediction coal pile volume calculation method based on point cloud registration of push harrow machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118376290A true CN118376290A (en) | 2024-07-23 |
| CN118376290B CN118376290B (en) | 2024-09-27 |
Family
ID=
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109556673A (en) * | 2018-11-22 | 2019-04-02 | 中铁四局集团有限公司 | Earthwork calculation method and system based on unmanned aerial vehicle |
| US20190244420A1 (en) * | 2018-02-08 | 2019-08-08 | Covidien Lp | Imaging reconstruction system and method |
| KR102038097B1 (en) * | 2018-10-24 | 2019-10-29 | 한국중부발전(주) | Amount of coal measure system using drone with lidar |
| WO2019242174A1 (en) * | 2018-06-21 | 2019-12-26 | 华南理工大学 | Method for automatically detecting building structure and generating 3d model based on laser radar |
| CN111681173A (en) * | 2020-05-25 | 2020-09-18 | 浙江大学 | Inclined image correction method based on combination of minimum area bounding box and projection method |
| WO2022017131A1 (en) * | 2020-07-22 | 2022-01-27 | 商汤集团有限公司 | Point cloud data processing method and device, and intelligent driving control method and device |
| CN114581619A (en) * | 2022-02-25 | 2022-06-03 | 浙江工业大学 | Coal bunker modeling method based on three-dimensional positioning and two-dimensional mapping |
| CN114862932A (en) * | 2022-06-20 | 2022-08-05 | 安徽建筑大学 | BIM global positioning-based pose correction method and motion distortion correction method |
| CN116642417A (en) * | 2023-04-17 | 2023-08-25 | 华能国际电力股份有限公司丹东电厂 | Method, system, computer equipment and storage medium for rapid analysis of coal quantity |
| CN117788838A (en) * | 2023-12-18 | 2024-03-29 | 东南大学 | Automatic ship unloading system and method based on laser radar scanner |
| US20240159146A1 (en) * | 2021-03-26 | 2024-05-16 | Hummingbird Solutions Inc. | Mining productivity improvement methods and systems |
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190244420A1 (en) * | 2018-02-08 | 2019-08-08 | Covidien Lp | Imaging reconstruction system and method |
| WO2019242174A1 (en) * | 2018-06-21 | 2019-12-26 | 华南理工大学 | Method for automatically detecting building structure and generating 3d model based on laser radar |
| KR102038097B1 (en) * | 2018-10-24 | 2019-10-29 | 한국중부발전(주) | Amount of coal measure system using drone with lidar |
| CN109556673A (en) * | 2018-11-22 | 2019-04-02 | 中铁四局集团有限公司 | Earthwork calculation method and system based on unmanned aerial vehicle |
| CN111681173A (en) * | 2020-05-25 | 2020-09-18 | 浙江大学 | Inclined image correction method based on combination of minimum area bounding box and projection method |
| WO2022017131A1 (en) * | 2020-07-22 | 2022-01-27 | 商汤集团有限公司 | Point cloud data processing method and device, and intelligent driving control method and device |
| US20240159146A1 (en) * | 2021-03-26 | 2024-05-16 | Hummingbird Solutions Inc. | Mining productivity improvement methods and systems |
| CN114581619A (en) * | 2022-02-25 | 2022-06-03 | 浙江工业大学 | Coal bunker modeling method based on three-dimensional positioning and two-dimensional mapping |
| CN114862932A (en) * | 2022-06-20 | 2022-08-05 | 安徽建筑大学 | BIM global positioning-based pose correction method and motion distortion correction method |
| CN116642417A (en) * | 2023-04-17 | 2023-08-25 | 华能国际电力股份有限公司丹东电厂 | Method, system, computer equipment and storage medium for rapid analysis of coal quantity |
| CN117788838A (en) * | 2023-12-18 | 2024-03-29 | 东南大学 | Automatic ship unloading system and method based on laser radar scanner |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3683721B1 (en) | A material handling method, apparatus, and system for identification of a region-of-interest | |
| AU2010265789B2 (en) | Autonomous loading | |
| CN108458715B (en) | Robot positioning initialization method based on laser map | |
| Tóvári et al. | Segmentation based robust interpolation-a new approach to laser data filtering | |
| Oehler et al. | Efficient multi-resolution plane segmentation of 3D point clouds | |
| CN112529958B (en) | Single laser radar bulk cargo ship hatch position identification method | |
| JP2004188562A (en) | Workpiece taking out device | |
| CN113340287B (en) | Cabin hatch identification method for ship loader | |
| CN112669376B (en) | Method and system for calculating layer height of residual stack of material pile | |
| CN115856923B (en) | Method, device, equipment and storage medium for measuring ore card unloading time | |
| CN114219770A (en) | Ground detection method, ground detection device, electronic equipment and storage medium | |
| CN118376290A (en) | Gaussian prediction coal pile volume calculation method based on point cloud registration of push harrow machine | |
| Jeong et al. | Point cloud segmentation of crane parts using dynamic graph CNN for crane collision avoidance | |
| CN117011362A (en) | Method for calculating cargo volume and method for dynamically calculating volume rate | |
| CN115937817A (en) | Target detection method and system and excavator | |
| CN115971004A (en) | Intelligent putty spraying method and system for carriage | |
| Steinemann et al. | 3D outline contours of vehicles in 3D-LIDAR-measurements for tracking extended targets | |
| CN114202631A (en) | Method for determining rock working face and working point in secondary rock crushing operation | |
| CN118225074B (en) | Self-adaptive map updating method and device for bulk cargo ship cabin cleaning robot | |
| Ngo et al. | Design of an AI Model for a Fully Automatic Grab-Type Ship Unloader System | |
| CN117152672B (en) | Cabin driving area dividing method and system based on point cloud polygon fitting | |
| CN115979273A (en) | Unmanned aerial vehicle full coverage path planning method and system based on self-adaptive viewpoint sampling | |
| Zhang et al. | Weld Seam Detection Method with Rotational Region Proposal Network | |
| Stößel et al. | Feature and viewpoint selection for industrial car assembly | |
| CN117687408A (en) | Intelligent control method, system and device for ship loader and storage medium |
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 |