CN107092020A - Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image - Google Patents
Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image Download PDFInfo
- Publication number
- CN107092020A CN107092020A CN201710257172.0A CN201710257172A CN107092020A CN 107092020 A CN107092020 A CN 107092020A CN 201710257172 A CN201710257172 A CN 201710257172A CN 107092020 A CN107092020 A CN 107092020A
- Authority
- CN
- China
- Prior art keywords
- data
- cloud
- point cloud
- road
- point
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
- G01C11/04—Interpretation of pictures
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/86—Combinations of lidar systems with systems other than lidar, radar or sonar, e.g. with direction finders
Abstract
The invention discloses a kind of fusion unmanned plane LiDAR and high-resolution remote sensing image highroad pavement planeness monitoring method, including:The LiDAR point cloud and high score image data obtained to unmanned plane is pre-processed;The multiple dimensioned geometric properties of the spectral signature of high score image and cloud data are carried out to construct validity feature variables set, so as to realize the Optimum Classification of road surface point cloud and non-road surface point cloud;Further more accurate road surface cloud data is obtained using Filtering Analysis;Build fine road surface model;Calculated using IRI indexes and obtain surface evenness.Existing surface evenness monitoring method inefficiency, the automaticity of the invention of effectively overcoming is low, is unsuitable for a wide range of system-wide section monitoring, can only obtain the road vertical section information of only a few, it is impossible to reflect the defects such as road surfaces three-dimensional structure information comprehensively;It is practical, it can be widely applied to urban road and the flatness monitoring of other inferior grade roads.
Description
Technical field
It is to utilize unmanned plane laser radar (Light the present invention relates to highroad pavement planeness remote sensing monitoring technology
Detection And Ranging, LiDAR) and high-definition remote sensing technology for highway pavement carry out a wide range of, high efficiency,
The surface evenness monitoring and evaluation of low cost.
Background technology
Airborne laser radar (Airborne LiDAR) is a kind of active earth observation systems, collection global navigational satellite system
System, inertial navigation, laser ranging and computer processing technology, can directly obtain high accuracy, high-resolution numeral
The three-dimensional spatial information of ground model and atural object, the superiority that can not replace with traditional photography measuring method, and in landform
The fields such as mapping, Road Alignment Survey Design, vegetation parameter inverting, disaster monitoring, urban planning and three-dimensional modeling have obtained more wide
General application.Highway plays more and more important effect as a kind of transport channel of modernization in social economy, to edge
The logistics of line, development of resources, invite outside investment, the effect actively promoted is played in industry restructuring, inter-regional economic cooperation etc., thus
Its quality directly influences the economic development of area and country.Document [1] (Sun L, Zhang Z M, Ruth
J.Modeling indirect statistics of surface roughness[J].Journal Of
Transportation Engineering-ASCE,2001,127(2):In 105-111), nineteen sixty National Highway with
The actual road test that communications and transportation administration federation carries out shows:About 95% road surface service performance depends on the flat of road surface
Whole degree.Surface evenness, as one of leading indicator for evaluating pavement quality, is also the important of execution maintenance of highway pavement management
Decision-making foundation.According to《2015 year traffic carrier statistical communique of developments》Statistics, Chinese Highway total kilometrage reaches within 2015
457.73 ten thousand kilometers, 446.56 ten thousand kilometers of highway maintenance mileage accounts for total mileage of highway 97.6%, and become in rapid growth year by year
Gesture.With developing rapidly that highway in China is built, a wide range of, high efficiency, the highway pavement health status of low cost how are realized
Monitoring just turns into a urgent problem to be solved.
Flatness monitoring is one of conventional highway pavement health monitoring and the important indicator evaluated.Flatness is evaluated
The measurement of index has 3 meters of rulers to determine maximal clearance (h), international roughness index (International Roughness
Index, IRI), exercise performance figure (Running Quality Index, RQI), jolt accumulated value (Bump
Cumulative Value, VBI), flatness standard deviation (σ), power spectral density (Power Spectral Density, PSD)
Deng.Existing traditional surface evenness monitoring method has:3m rulers, continous way smoothness measuring equipment, hand propelled profiler, vehicular
Jolt accumulating instrument and the laser evenness detector (road detection vehicle for including integrated laser smoothness measuring equipment) that occurs in recent years,
Its each the characteristics of be summarized in table 1.One of general character of these methods is exactly that can only be obtained by ground survey as the seldom road of number
Road vertical section information, it is also relatively rough to the description of road evenness, and can not quick obtaining road surface three-dimensional geometry letter
Breath so that understanding of the administrative department to road integral status is very limited.
The conventional surface evenness monitoring method of table 1.
The appearance of low latitude unmanned plane LiDAR technologies be not only obtain road surface three-dimensional information provide may, and should
Technology is less to be limited by landform, high accuracy, high efficiency, low cost (especially a wide range of monitoring), can be while obtaining ground remote sensing
The advantage of image is even more to provide a kind of new effective solution for surface evenness monitoring.Enter currently with LiDAR technologies
The CROSS REFERENCE of walking along the street surface evenness monitoring and research are rare.Document [2] (Chin A, Olsen M J.Evaluation of
Technologies for Road Profile Capture,Analysis,and Evaluation[J].Journal of
Surveying Engineering,2014,141(1):4014011) describe using ground laser radar technology to public at a high speed
Road has carried out flatness monitoring, still, and the do not satisfy the need spatial visualization method of surface evenness of this method is attempted, also not
The construction method of a cloud classification method and road surface model is explored, another the heel of Achilles of the technology is exactly
Small range (radius can only be obtained in some fixing point position vicinity<Cloud data 50m).If obtained on a large scale
Information of road surface then needs a series of complex, the errors such as cloud, coordinate adjustment, point cloud registering, Coordinate Conversion constantly to accumulate at substantial amounts of
Tired last handling process, not only flow is complicated but also very poorly efficient.Document [3] (path formations of the Guo Jiao based on vehicle-mounted laser
Spend detecting system research [D] Capital Normal Universitys, 2013) describe and put down using mobile lidar technical Analysis road section
Whole degree feature simultaneously positions the exceeded position of road damage, and still, this method is explored also without to a cloud classification method,
But work of classifying is completed by traditional points cloud processing platform.In addition, this method relies solely on single number carry out table
Reach, the spatial distribution characteristic for failing road pavement quality is explored, and require that detection car traveling velocity-stabilization, gps signal are steady
It is fixed, and this is some buildings are blocked, rugged complicated highway section is difficult to apply.
In summary, there is inefficiency, automaticity in existing traditional highroad pavement planeness monitoring method mostly
It is low, it is impossible to carry out a wide range of system-wide section monitoring, and can only obtain as several seldom road vertical section information, it is impossible to obtain
The three-dimensional structure information of road surfaces.The flatness monitoring method of existing laser radar technique then has that measurement range is small, operation
Efficiency is low, lack the defect that spatial visualization is expressed, data source is single, and can be limited by orographic condition.But highway road
Surface evenness monitoring improves a wide range of surface evenness monitoring efficiency and accuracy in the urgent need to exploring new technological approaches.
The content of the invention
For the deficiency in existing highway surface evenness monitoring method, the present invention proposes a kind of fusion unmanned plane LiDAR
With the surface evenness monitoring method and its embodiment of high resolution image.First, unmanned plane LiDAR systems are obtained
Cloud data and high score image data are pre-processed, and then extract spectral signature from high score image data, cloud data is carried
Take multiple dimensioned geometric properties.Using resulting geometric properties and spectral signature as point cloud classifications characteristic variable, and to these
Characteristic variable carries out de-redundancy and dimensionality reduction operation, two class data is merged in feature rank, using in machine learning
Random forest sorting algorithm carries out point cloud classifications;Further corrected by filtering operation and rough error and obtain comparing accurately road
Waypoint cloud data set.Then high-precision fine road surface digital surface model (Digital is built using three-dimensional interpolation method
Surface Model,DSM).Surface evenness is calculated based on IRI indexes principle, recycles GIS to carry out spatial visualization expression
With flatness quality evaluation.Whole section can only be expressed compared to conventional surface evenness monitoring technology with single IRI values
For information, party's rule can provide the spatial distribution of the different track IRI values of certain road different sections of highway, improve maintenance of surface
The science of management work decision-making, effectively increases practicality of the unmanned aerial vehicle remote sensing technology in highway maintenance management application, can
It is widely used in urban road and the surface evenness monitoring needs of other inferior grade roads.
The present invention principle be:Highway and periphery atural object are included in the cloud data that unmanned plane LiDAR remote sensing systems are obtained
Three-dimensional coordinate information, can effectively remove noise jamming by the suitable point cloud denoising means of selection, and the high score that the same period obtains
Resolution image data is rich in the spectral information of atural object and is highly susceptible to visual interpretation.Cloud data then because point set it is scattered without
Interpretation easily directly perceived is this, and merged high score image with cloud data just can be while by spectral information and space geometry information
Progress is integrated, and more available informations are provided for point cloud classifications.At present, there is one for the terrain classification of high score image
Challenge, exactly cannot be distinguished by the ground object target of " foreign matter is with spectrum ", and the space geometry information of cloud data can then rise to this
To good supplement.Likewise, cloud data itself is very discrete and cause in some demarcation lines and several due to point set
Obscure between what the closely similar atural object of architectural feature, and high score image then can be to this good booster action of progress.Base
In this, on the other hand one aspect of the present invention extracts allusion quotation from the spectral signature of high score Extraction of Image typical feature classification from a cloud
The geometric properties of type atural object classification, strengthen the spatial shape difference of type of ground objects by building multiple dimensioned geometric properties, then
A cloud is classified using efficient and the anti-noise sound intensity machine learning classification algorithm, then sorted cloud carried out thick
Difference is corrected and the i.e. available more accurately road cloud data of classification filtering, and fine road is built in conjunction with a cloud interpolation algorithm
Road DSM.On this basis, a series of vertical profile surface curves are chosen from road surface, segmentation calculates international flatness value and carries out sky
Between Visualization, finally to flatness value carry out grade determination, so as to realize the evaluation to road evenness quality.This hair
Bright method proves to can be utilized for the rapid automatized flatness quality condition prison of a wide range of highway pavement by precision evaluation
Survey.
The technical scheme that the present invention is provided:
Fusion unmanned plane LiDAR and high score image a kind of surface evenness monitoring method, including:Data acquisition and pre- place
The sport technique segments such as reason, feature extraction are with merging, point cloud classifications, the structure of road surface DSM models and Evaluation of Pavement Evenness.
Wherein, data acquisition and the purpose of pretreatment are that the cloud data that low latitude unmanned plane LiDAR systems are obtained is carried out
Denoising, carries out registration and splicing to the high score image that the same period obtains, obtains the orthography data of test block, be data fusion
Prepare.The purpose of data fusion is to carry out the spectral information of the geometric space information of cloud data and high score image data
It is integrated, so that cloud data is provided simultaneously with space geometry information and spectral information.The purpose of point cloud classifications is to obtain
Road cloud data, the process of data fusion provides available information for point cloud classifications, and typically species are carried out to test block
Do not divide, build a series of geometrical characteristic parameters to express the difference of type of ground objects, and just constitute with reference to spectral signature parameter
The characteristic variable of grader, then realizes the classification to cloud data using random forests algorithm, obtains road surface cloud data.
The structure of road surface DSM models can provide basis, and each grid of DSM models for the expression of road surface three-D space structure
Unit storage is all elevation information, is easy to flatness to calculate, and also provides condition for the spatial visualization expression of flatness.
The purpose of Evaluation of Pavement Evenness is exactly that the flatness situation progress quantization by choosing certain roughness index road pavement is retouched
State, and set up grade determination system and objective evaluation is carried out to the condition of road surface of test block, so that the current integral status of road pavement
Provide prospective analysis and decision-making.
The surface evenness monitoring method that the present invention is provided specifically includes following steps:
1) using low latitude unmanned plane LiDAR systems (including:Unmanned aerial vehicle platform, LiDAR scanners, multispectral camera, posture
The units such as positioning unit, automatic pilot it is integrated), obtain test block LiDAR point cloud data and high resolution image number
According to;
2) the LiDAR point cloud data to acquisition are converted to current general cloud data file format (* .las), then right
The high resolution image data that cloud data and unmanned plane are shot are pre-processed, including the denoising of a cloud, the splicing of image and
The generation of registration process and orthography data;
3) to step 2) in the obtained wave band of orthography analyzed and extract the spectral signature of reflection atural object difference
Parameter, together with the original wave band of high resolution image, constitutes multiple Spectral Properties comprising spectral information and collects;
4) Fusion Features are carried out to obtained spectrum characteristic data and cloud data, and entered by georeferencing benchmark of a cloud
Row geo-spatial registration;
5) the major surface features type to test block is divided, and in step 4) on the basis of obtained cloud data by taking out
Dilute cloud data vacuated, then chooses a series of geometrical characteristic parameter to the cloud data vacuated and calculates, and
The cloud data comprising multiple dimensioned geometrical characteristic parameter and spectral signature parameter is obtained by changing the yardstick of spatial analysis, is connect
The cloud data of the corresponding type of ground objects of therefrom selected part as the training sample of follow-up point cloud classifications, is vacuated remaining
Cloud data is as treating divided data.The process that vacuates of point cloud reduces the operand in geometrical characteristic parameter calculating process, simultaneously
The data volume of processing, substantially increases data-handling efficiency needed for also reducing classification;
6) step 5) in obtained characteristic variable would generally be a lot, inevitably occur between some variables exist compared with
Strong correlation, causes characteristic variable redundancy.In order to find key characteristic variables, so that data are played with dimensionality reduction effect, and then
Simplify the complexity of grader building process, in feature rank, merge the spectral information and laser point cloud number of multispectral image
According to space geometry information, constitute new point cloud classifications feature set;
7) using Random Forest model be based on step 6) feature set to step 5) training sample carry out model training, lead to
Cross constantly adjustment model parameter to obtain after satisfied classifying quality, classified applied to sample cloud data to be sorted, so
The category attribute of sample point cloud is assigned to the cloud data after denoising according to space arest neighbors Interpolation Principle afterwards, so as to complete
All put cloud classification.Classification results will may now enter also comprising the point cloud for being not belonging to road surface to sorted cloud data
Row hand inspection and rough error are corrected, and further obtain road cloud data by category attribute filtering;
8) to step 7) the obtained elevation information of road cloud data, risen and fallen and become using TIN simulated roadway
Change, the DSM for the fine pavement of road of high accuracy that regular grid is then built by natural neighbor interpolation method;
9) from step 8) in choose a series of buttock line in obtained DSM model datas, run IRI exponential model meters
Calculate the corresponding IRI values of each hatching and by rasterizing processing, obtain the spatial distribution map of the IRI values on road surface;
10) to step 9) in obtained IRI values, the grade determination standard according to road evenness is classified, so that must
To the corresponding flatness quality evaluation result in section to be measured.
For above-mentioned highroad pavement planeness monitoring method, further, step 2) in the LiDAR point cloud data that are related to,
Specially meet the cloud data of following 4 conditions:
A) coordinate system:The coordinate system of cloud data is WGS84 coordinate systems (longitude, latitude) or the ground matched with locality
Square plane projection coordinate system (X, Y);
B) attribute information:Cloud data attribute information should also include elevation, scan angle, reflection by force in addition to longitude, latitude
Degree;
C) cloud density is put:Point cloud density (i.e. laser footpoint radius is within 50mm) more than 400 points/square metre;
For above-mentioned flatness monitoring method, further, to step 2) in the process of data preprocessing that is related to, mainly
Refer to following 4 kinds:The noise spot filtering of LiDAR point cloud, the abnormal elevation filtering of LiDAR point cloud, the scan angle of LiDAR point cloud
Filter with the splicing of high resolution image with it is registering.For above-mentioned flatness monitoring method, further, to step 2) in relate to
And high resolution image data, refer to the high-definition digital photo on the market captured by multispectral digital camera of good performance,
And unmanned function provides the posture matched and positioning (POS) message file.For above-mentioned flatness monitoring method, enter one
Step ground, to step 3) in the spectral signature parameter that is related to, be following 3 class:
A) gray value of the red, green, blue wave band of high score image.
B) standard deviation of the gray value of the red, green, blue wave band of high score image;
C) (laser pulse of LiDAR scanners transmitting falls within one kind of electromagnetic wave to the reflection intensity values of laser point cloud, together
Sample has wavelength information).
For above-mentioned flatness monitoring method, further, to step 4) in the Fusion Features that are related to, specifically refer to point
Cloud data are merged with the spectral signature parameter of high resolution image data, and the process merged is with cloud data
On the basis of georeferencing, the spectral value of each pixel on image is matched by cloud data using affine inverse transformation model and sat
Mark system;
For above-mentioned flatness monitoring method, further, to step 5) in the geometrical characteristic parameter that is related to, be following 3
Class:
A) local roughness (Local Degree of Roughness, LDR).The concrete meaning of this feature variable refers to
The best-fitting plane that the neighborhood point under some space scale where some reference point to the reference point in point cloud is formed
Distance;
B) local dimension feature (Local Dimension Feature, LDF).Document [4] (Brodu N, Lague
D.3D terrestrial lidar data classification of complex natural scenes using a
multi-scale dimensionality criterion: Applications in geomorphology[J].ISPRS
Journal Of Photogrammetry and Remote Sensing, 2012,68:121-134) describe local dimension
Feature LDF, the concrete meaning of this feature variable be give directions some reference point in cloud and this refer between neighborhood of a point point in sky
Between on be, with one-dimensional, two-dimentional or distributed in three dimensions characteristic measure value, to belong to these three dimensions with a numerical value to describe it
Size degree, the value is exactly local dimension characteristic parameter;
C) local depth displacement (Local Height Difference, LHD).The concrete meaning of this feature variable is to give directions
The reference point of some in cloud is concentrated with all points that neighborhood point of the reference point under some space scale is constituted, and reaches the point
The difference of the maximum and minimum range in the formed best fit face of collection, is local depth displacement, the bigger plane value of fluctuating
It is bigger.
For above-mentioned flatness monitoring method, further, to step 5) in the multiple dimensioned geometric properties that are related to, its is specific
Implication refers to, when calculating some geometrical characteristic, its spatial analysis yardstick has that multiple (fixing some spatial analysis yardstick can obtain
To a corresponding space geometry characteristic parameter), different types of ground objects its geometrical characteristic under different spatial analysis yardsticks
Have difference.
For above-mentioned flatness monitoring method, further, to step 5) in the cloud data classification that is related to, mainly include
Road waypoint cloud (contain traffic marking) and short vegetation (crops and meadow), soil, trees, move target, power line (including
Electric pole) this 6 major class.
For above-mentioned flatness monitoring method, further, to step 5) in the point cloud that is related to vacuate process, be in space
In sampled according to certain space interval, the process will ensure distribution of a cloud in space for original point cloud
It is uniform (rather than messy distribution).
For above-mentioned flatness monitoring method, further, to step 6) in the feature selecting that is related to, mainly including feature
Two processes of the search of subset and the evaluation of character subset.The search of character subset uses Greedy strategy, i.e.,:Initial
Optimal subset is sky, and all characteristic variables to be assessed then constitute alternative features collection, and screening interpretability is concentrated from alternative features
Most strong characteristic variable is added in optimal feature subset, and one cycle only adds a variable;A certain characteristic variable addition knot
Need to concentrate from alternative features after beam and reject this feature variable, circulate next time again.The evaluation criterion of optimal feature subset:
Correlation between the characteristic variable newly added and the characteristic variable chosen is relatively low.
For above-mentioned flatness monitoring method, further, to step 8) in the resolution ratio of DSM models that is related to will to the greatest extent can
Can be high, and its height accuracy should be controlled within 15mm.Specific level and height accuracy assessment level are:From scene
Select some that there is the object of obvious characteristic (such as pit) in road surface and record its corresponding gps coordinate, grown by measuring it
Wide high attribute, then finds corresponding object in high resolution image, utilizes merging for image and LiDAR point cloud, it is easy to
The correspondence position of the target is found in DSM models, and then interactively determines using dimensional measuring instrument the length of the target
It is wide high, then statistical analysis completion is to the level and the accuracy evaluation of elevation of DSM models.
For above-mentioned flatness monitoring method, further, to step 9) in the road buttock line that is related to, its spacing gets over
Small then corresponding surface evenness spatial distribution is finer.The theoretical foundation of IRI exponential models is a quarter vehicle model,
The length (10m or 20m) of strict control sampling interval (within 50mm) and measuring unit is needed in calculating process.Optimal adopts
Sample interval and measuring unit length can by Monte-Carlo Simulation Method quantitative analysis the timing IRI of vertical error one error
Determined with the changing rule of sampling interval and measuring unit length.
For above-mentioned flatness monitoring method, further, step 9) in the rasterizing that is related to refer to road surface vertical section
The IRI values of line are assigned to the grid cell of surrounding according to closest principle, so as to obtain the flatness spatial distribution in section to be measured
Figure.
It is further, a series of comprising different smooth by being selected in section to be measured for above-mentioned flatness monitoring method
The road surface sample of credit rating is spent as reference data, can carry out accuracy assessment to the evaluation result of this method, specific steps are such as
Under:
1) according to the flatness quality classification standard of setting, the same period is a series of (comprising flatness matter in section to be measured selection
Amount preferably with poor, covering health and Damage Types, every kind of road surface selects a number of ground actual measurement sample) road surface sample
This is used as reference data;
2) choosing outward appearance, surface roughness, rutting depth, disease area, damaged condition this 5 indexs to reference data makes
Flatness quality evaluation is carried out to it with expert point rating method, and is used as reference value;
3) actual measurement sample progress flatness value in this road surface is calculated with this method is same, according to flatness quality
Grade scale, obtains the evaluation result of the surface evenness quality of this method;
4) evaluation to flatness quality results can be analogous to the evaluation of classification of remote-sensing images result, and confusion matrix is
For evaluating the most common form of classification of remote-sensing images precision, therefore this method is also satisfied the need surface evenness matter using confusion matrix
Amount result is evaluated.Precision evaluation index overall accuracy (Overall Accuracy, OA) and Kappa coefficients are introduced to knot
Fruit is analyzed, and obtains the precision of this method.
Compared with prior art, the beneficial effects of the invention are as follows:
LiDAR data and high-resolution remote sensing image that the present invention is obtained based on low latitude unmanned aerial vehicle platform, propose to melt first
The method that the spectral signature of picture of taking a group photo and the multiple dimensioned geometric properties of LiDAR point cloud build point cloud classifications feature set, is solved current
The problem of road surface point cloud and non-road surface point cloud classifications;Road surface point cloud is recycled to build subtle three-dimensional road surface digital surface model
(DSM) it is quick to calculate the flatness parameter for obtaining highway pavement, by introducing the Computing Principle of international roughness index, realize
The quick monitoring and evaluation of pavement quality.Present invention firstly provides a kind of fusion low latitude unmanned plane LiDAR and high score image
The flatness monitoring method of highway pavement rapidly and efficiently, efficiently solve inefficiency in existing surface evenness monitoring method,
Automaticity is relatively low, is not suitable for a wide range of system-wide section monitoring, and can only obtain as several seldom road vertical profile surface evenness letters
Breath, it is impossible to obtain the deficiencies such as road three-dimensional structure information.This method can quickly grasp large-scale surface evenness situation, from
And provide scientific basis for pavement maintenance & rehabilitation decision-making.Meanwhile, the Point Cloud Processing flow that the present invention is designed is simple easily real
Existing, effect is notable.In addition, the flatness mass space distribution map that the present invention is provided can reflect the flatness of highway pavement comprehensively
Situation, by single number can only express the flatness information in whole section compared to conventional flatness monitoring technology and
Speech, effectively promotes the informationization and the development that becomes more meticulous of road maintenance management, improves pavement maintenance management job acceptance decision
It is scientific.
Fusion evaluation and LiDAR of the present invention carry out surface evenness monitoring, and compared with prior art, one is to employ low latitude
Unmanned plane LiDAR data source, and propose fusion spectral signature information first with multiple dimensioned geometric properties information to optimize waypoint
Cloud classification, builds fine DSM on this basis, and rapid extraction has the surface evenness information of spatial distribution;Two be data
Processing procedure succinctly easily realizes that the mathematical modeling used is it can be readily appreciated that effect is notable, and all operations can be in personal common electricity
Completed on brain, to the less demanding of hardware;Three be that in the operating process for providing feasibility, furthermore present optimization data
The method of processing, and checking is completed by testing, with good repeatability.
Brief description of the drawings
The flow chart for the surface evenness monitoring method that Fig. 1 provides for the present invention
Fig. 2 is local roughness computational methods schematic diagram.
Fig. 3 simplifies expression schematic diagram for local dimension feature;
Wherein, (a) is one-dimensional space distribution characteristics, and (b) is two-dimensional space distribution characteristics, and (c) is that three-dimensional spatial distribution is special
Levy.
Fig. 4 is that local depth displacement calculates schematic diagram.
Fig. 5 is the natural neighbor interpolation algorithm schematic diagram of progress point cloud interpolation in this method.
Fig. 6 is the algorithm schematic diagram of the international roughness index (IRI) used in this method;
Wherein, ZsRepresent the vertical displacement of spring carried mass, ZuRepresent the vertical displacement of nonspring carried mass, msRepresent spring charge material
The size of amount, muRepresent the size of nonspring carried mass, KsRepresent the stiffness system of the spring of connection spring carried mass and nonspring carried mass
Number, CsRepresent the linear damping coefficient of connection spring carried mass and nonspring carried mass, KsRepresent the precision system of the spring of connection tire
Number, can simulate the damping effect produced by tire, and y (x) represents road surface elevation, and what b was represented is containing length of the tire to ground
(part that tire is contacted with ground).
Fig. 7 is the road surface sample photo of the different flatness features of ground acquisition in the embodiment of the present invention
Wherein, (a1) pit, is characterized as building stones exposure and in big block distribution, shape is approximately circular or ellipse,
Flatness quality evaluation is extreme difference (Failed);(a2) pit, is characterized as that building stones are exposed but area is relatively small, shape is approximate
Circular or ellipse, flatness quality evaluation is extreme difference (Failed);(b1) collapse, be characterized as ribbon distribution and depth compared with
(building stones do not expose), area coverage are big greatly, and flatness quality evaluation is extreme difference (Failed);(b2) collapse, be characterized as band
Shape is distributed and depth smaller (building stones do not expose), area coverage are small, and flatness quality evaluation is poor (Poor) or extreme difference
(Failed) depression depth and area coverage, are specifically dependent upon;(c) crack (including minute crack and thick crack), is characterized as narrow
Long shape distribution, crack gaps are within 10mm, and flatness quality evaluation is medium (Fair) or poor (Poor), is specifically dependent upon
Fracture width and area;(d) pitted skin, is characterized as that road surface is very coarse or has many trickle pits, flatness quality evaluation
For medium (Fair) or poor (Poor), the degree and area of surface breakdown are specifically dependent upon;(e) sandstone are covered, and are characterized as road surface
Occur block damaged and covered by sandstone, flatness quality evaluation is poor (Poor);(f) the general old pavement of flatness, feature
Rougher but integral surface is smooth for pavement texture, flatness quality evaluation is medium (Fair);(g) flatness is preferably old
Road surface, is characterized as that road surface color and luster more light, texture densification, surface are very smooth, flatness quality evaluation is preferably (Good).
Fig. 8 changes LiDAR point cloud data to be obtained in test block in the embodiment of the present invention and passing through form.
Fig. 9 is the multispectral orthography in test block of the embodiment of the present invention through image joint and registration process.
Figure 10 is that (blue ripple is shown in the LiDAR point cloud data of fusion evaluation band class information in the embodiment of the present invention in figure
Section).
Figure 11 is to carrying out feature using 5 spectral informations and 36 multiple dimensioned geometric properties information in the embodiment of the present invention
Property value characteristic distributions of 23 key characteristic variables and training sample retained after selection under these characteristic variables;
Wherein, 5 light such as reflected intensity, R-G-B wave bands gray standard deviation, R-G-B band grey datas of (a) different atural objects
Spectrum signature;(b) local roughness (LDR) and local dimension feature (LDF) of different atural objects;(c) the local dimension of different atural objects
Spend feature (LDF) and local depth displacement (LHD);(d) the local depth displacement (LHD) of different atural objects.Length digital generation in parantheses
Table space analyzes scale size, and LDF_1 represents one-dimensional local dimension feature, and LDF_1 represents the local dimension feature of two dimension
The DSM that Figure 12 expresses for the millet cake cloud interpolation generation TIN that satisfied the need in the embodiment of the present invention.
Figure 13 is the road DSM that is obtained using natural neighbor interpolation algorithm in the embodiment of the present invention.
Figure 14 is that the embodiment of the present invention calculates obtained surface evenness spatial distribution map using IRI exponential models.
Figure 15 is that the surface evenness mass space that the embodiment of the present invention to flatness value obtain after reclassification is distributed
Figure.
Figure 16 is the spatial distribution statistical chart of different pavement quality types in the embodiment of the present invention.
The confusion matrix schematic diagram for the flatness quality evaluation result that Figure 17 is set up by assessment case study on implementation of the present invention.
Embodiment
Below in conjunction with the accompanying drawings, and by instantiation the specific implementation process of the present invention is further described, but not with any
Mode limits the scope of application of the present invention.
Fig. 1 is the flow of the surface evenness monitoring method of fusion unmanned plane LiDAR proposed by the present invention and high score image
Block diagram, comprises the following steps:
Step 1:LiDAR point cloud data in section to be measured and high resolution image are obtained using unmanned plane LiDAR systems simultaneously
Data;
Step 2:(production firm is mostly for the cloud data form difference provided due to different LiDAR systems production firms
Have a set of customized storage format), in order to be able to be handled on some common softwares, it is therefore desirable to LiDAR point cloud number
According to row format conversion is entered, the field attribute of selection and the reservation of projected coordinate system is noted in output;
Step 3:Denoising is carried out to the universe point cloud file (* .las) obtained in step 2, interference noise is tentatively removed
Information, thus obtain it is dry after cloud data.Specific denoising mode includes:
A) noise spot is filtered.Mainly the refraction in LiDAR point cloud due to laser pulse or multipath effect are produced
Abnormity point rejected;
B) abnormal elevation is filtered.Local abnormal elevation (rough error) is mainly rejected, specific determination methods are:Set
Certain search radius threshold value r, if the elevation average value of the neighborhood point where certain point and the point in radius r is more than 3
Standard deviation again, then can be considered abnormal elevation and rejected;
C) angular filter is scanned.Due to when unmanned aerial vehicle flight path is planned, in order to be able to the surface information in Overall Acquisition section, nobody
The flight path of machine should directly over section, therefore using point cloud scan angle attribute can to data carry out filtering so as to
Data processing amount is reduced, unnecessary data processing time is saved.Assuming that the horizontal range immediately below certain point to aircraft is
D, the flying height of aircraft is h, then the scan angle theta of the point is defined as formula 1:
θ=arctan (d/h) (formula 1)
The maximum scan angle θ of road can be determined according to the bounds of roadmax, therefore suitable scan angle threshold is set
Value can separate road area.Have in view of course line needs as far as possible to set scan angle threshold value in certain deviation, experiment
It is somewhat larger;
Step 4:Utilize the unmanned plane image joint software (business developed by Pix4D companies of Switzerland used in such as this example
With software Pix4D Mapper), image joint is carried out to the high resolution image data that unmanned plane in section to be measured is shot and matched somebody with somebody
Quasi- processing, can obtain section to be measured just penetrates remote sensing image;
Step 5:Spectra feature extraction is carried out to high resolution image, the spectral signature parameter that the present invention is used includes following
4:The gray value of red, green, blue wave band and the standard deviation of these three wave bands, the mathematical modeling of standard deviation is formula 2:
Wherein, std. represents standard deviation, gi(i=1,2,3) be the wave band of red, green, blue three gray value.
Step 6:To carrying out Fusion Features comprising multiband high score image and cloud data, with the geographical space of cloud data
With reference on the basis of, the pixel coordinate of image is mapped to the geographical coordinate of the pixel center point by affine transformation, so that by shadow
As upper corresponding spectral information matches cloud data.The mathematical modeling of affine transformation is formula 3:
But in practical operation, because the pixel quantity of image is far longer than the data volume of a cloud, and both are also endless
Full weight is closed, therefore in order to reduce amount of calculation, the present invention uses the process that image coordinate is transformed into from geographical coordinate, that is, imitates
Inverse transformation is penetrated, formula 4 is expressed as:
In formula 3, formula 4, G is affine transformation matrix, ai,bi,ci(i=1,2) is affine transformation coefficient, mapX, mapY points
It is not X-coordinate, the Y-coordinate of a cloud, row, col are the row, column value of image respectively.GeoTiff or other with geographical coordinate believe
Typically just there is affine coefficients information in the metadata of the image file of breath, can also pass through 4 border point coordinates on image in addition
And its corresponding ranks number (pixel coordinate) information is solved by building equation and obtained to affine coefficients.
Step 7:Octree is built to cloud data to store, using the length of Octree grid unit as between minimum space
Vacuated every to a cloud, the cloud data vacuated.Then the meter of geometrical characteristic parameter is carried out to the cloud data vacuated
Calculate, the geometrical characteristic parameter used in the present invention includes three below:
A) local roughness (Local Degree of Roughness, LDR), as illustrated in fig. 2, it is assumed that whole three-dimensional point
Cloud point integrates as C={ pi|pi=(xi,yi,zi), i=1,2,3 ..., n }, the current point that calculates is p=(x, y, z) ∈ C, three-dimensional space
Between in its radius be R neighborhood point set be P={ pj|||pj-p||≤R,pj≠ p, j=1,2,3 ..., n }, if plane T:Ax+
By+Cz+D=0 is point set Q={ P ∪ p } optimal fitting plane, then LDR mathematical form can be expressed as formula 5:
A in formula 5, B, C, D represent the basic parameter of plane equation respectively, and x, y, z represents the x of current search point respectively,
Y, z coordinate.When optimal fitting plane T in the present invention refers to that all points of concentration are projected in x-y plane so that z value errors
Quadratic sum S is minimum plane, and specific mathematical form is formula 6:
X in formula 6i,yi,ziRepresent i-th point of X respectively, Y, Z coordinate, other specification implication is with formula 5;
B) local dimension feature (Local Dimension Feature, LDF), as shown in figure 3, three-dimensional point cloud point set and
Neighborhood point set is configured similarly to local roughness, it is assumed that current search point is p=(x, y, z) ∈ C, makes point set Q={ P ∪ p }
=[X Y Z], carries out principal component transform to Q first, obtains matrix Q three principal component coefficient μ1,μ2,μ3(μ1≥μ2≥μ3),
Further these three principal component coefficients are normalized by formula 7:
Wherein, λ1、λ2、λ3Correspond respectively to current search point and one-dimensional, two-dimentional, three-dimensional spatial distribution is obeyed in the neighborhood
Degree, namely one-dimensional, two-dimentional, three-dimensional feature value, due to λ1+λ2+λ3=1, therefore only need to the first two characteristic parameter
The local dimension feature of current search point is expressed, the geometric properties computational efficiency of cloud data can be so improved;
C) local depth displacement (Local Height Difference, LHD), as shown in figure 4, wherein optimal fitting plane
Definition and local roughness calculate in optimal fitting plane definition it is consistent, it is assumed that current search point is p=(x, y, z) ∈
The neighborhood point set that its radius is R in C, three dimensions is P={ pj|||pj-p||≤R,pj≠ p, j=1,2,3 ..., n }, point
Collection Q={ P ∪ p } is to include current search neighborhood of a point point set, then LHD mathematical form can be expressed as formula 8:
In formula 9The i-th, distance of j neighborhood point to optimal fitting plane is represented respectively, and other specification implication is same
Formula 5;
In experimentation in order to accelerate calculate Searching point local depth displacement speed, to Q carry out SVD decompose can be quick
Plane T 4 key parameters A, B, C, D are obtained, are then looked for apart from point maximum plane T and apart from click-through minimum plane T
Row distance asks poor, so as to obtain the local depth displacement of Searching point.
By the size for the radius R for constantly changing search neighborhood, and calculate successively each geometric properties (LDR, LDF and
LHD) the respective value under different search space radius R, can obtain one group of column vector on these three geometric properties, this
Group column vector is exactly the multiple dimensioned geometrical characteristic parameter to ground object target (object where Searching point).
Step 8:After being merged to high resolution image with cloud data, cloud data now contains substantial amounts of spy
Variable (geometrical characteristic parameter and spectral signature parameter) is levied, the matrix for forming a higher-dimension (each puts inherently one 3-dimensional
Vector), correlation is also likely to be present between these characteristic variables in itself, causes information to produce certain redundancy, redundancy is removed
The process of information is exactly the process of feature selecting.The feature selecting strategy that the present invention is used is divided into two processes:Character subset is searched
Rope and character subset evaluation.The strategy of character subset search is sweep forward, and the evaluation of character subset is then mainly by phase
Relation number assesses the degree of relevancy of intra-subset.Concrete implementation process is as follows:
1) data structure of two aggregate forms is initially set up, it is assumed that respectively P and Q, P represent candidate subset, Q is represented most
Excellent subset, category attribute variable is Y, and other attributes in addition to category attribute are X={ X1,X2,X3,...,Xn}.By in addition to Y
Other characteristic variables are added in candidate subset, and the relevance threshold for setting intra-subset is δ;
If 2)Into step 4), each characteristic variable X in X is read in otherwise circulationi, and calculate it by formula 10
With Y relative coefficient, the characteristic variable X maximum wherein with Y relative coefficient is foundc, and make δ=max { corr (Xi,
Y)}(Xi∈ X), subsequently into 3);
In formula 10, Xi, Y represent ith feature, categorical attribute respectively,Ith feature, categorical attribute are represented respectively
Average value, andThe ith feature and category attribute value of j-th of sample data are then represented respectively.
3) by characteristic variable X to be selectedcRelativity evaluation is carried out with each characteristic variable in Q, if XcWith it is any in Q
The correlation of one characteristic variable is respectively less than δ, then by XcIt is added in Q, while deleting this feature variable out of P, otherwise directly
This feature variable is deleted out of P, continues step 2);
4) Q is exported, terminates character subset search.
Special instruction, can also be omitted if characteristic variable number is not a lot (within such as 30) steps.
Step 9:After cloud data after being vacuated, an interactive selection part therein trains sample as training sample
This requirement covers all types of ground objects, and training sample can not be concentrated excessively, where being suitably dispersed in a cloud in space,
The training samples number of every kind of type of ground objects should not very little, and otherwise grader can not build disaggregated model.Random Forest model has
Have adaptable (can be while handling discrete type, numeric type data) to data set, training speed is fast, noise resisting ability is strong,
The advantages of being not easy to be absorbed in few over-fitting, model parameter, good classification effect and as most popular calculation in machine learning field
One of method, the present invention uses for reference its advantage and come pair while the cloud data for merging multiple dimensioned geometric properties and spectral signature carries out ground
Thing category classification.
The basic thought of the algorithm is:Assuming that the number of input sample data is N (each sample correspondence a line), feature becomes
The number of amount is M (each characteristic variable correspondence one is arranged, i.e., raw sample data size is N*M matrix), is had at random when initial
That puts back to selects n sample as the training of one tree (every one tree is all binary tree in random forest) from N number of sample
Collection, the root node of the tree is exactly training set;M characteristic variable is randomly selected from M feature, is then become from this m feature
Enter line splitting according to the minimum principle selection characteristic variable of purity in amount, so as to which the tree is divided into two nodes, the two sections
A part for each self-contained training subset of point, it is so recursive to be divided, until the tree can not continue division and (reach y-bend
The maximum depth of recursion of tree) or this m feature use untill finishing.Repetition built just now tree process can obtain containing times
The forest (random forest) of meaning tree, and the training sample of each tree is all not exclusively the same, and the characteristic variable used is also endless
It is complete the same.In the prediction classification stage, when inputting a sample to be sorted, each tree in forest can all be obtained to the sample
One decision-making classification, the distribution frequency that just can obtain sample to be sorted under each prediction classification is counted to decision-making classification,
The maximum classification of output wherein frequency predicts classification as it, so as to complete classification.
The purity Measure Indexes used in the present invention are Gini coefficient, and using document [5], (Chen Huazhou, Chen Fu, Shi Kai wait
Fish meal protein Near-Infrared Quantitative Analysis [J] agricultural mechanical journals based on random forest, 2015 (05):233-238) record
Computational methods, its mathematical modeling is:
1) assume that training sample set includes P classification, the sample number ratio of i-th of classification is pi, then its Gini coefficient
Gini (P) value is:
M subset is divided into according to some characteristic variable to the training sample set, the number of samples of i-th of subset is ni,
The total sample number of the training sample set is n, then the Gini coefficient Gini of subset after dividingsplit(P) it is:
Step 9:Point cloud after to vacuating is classified, and filtering is carried out to it according to category attribute can obtain road surface point cloud number
According to the point cloud interpolation algorithm for further passing through correlation can be obtained by the DSM data in section to be measured.To reduce elevation noise pair
The influence that surface model effect and follow-up flatness are calculated, this method generates dimensional topography first with TIN (TIN)
Data, then build the DSM models of road using natural neighbor interpolation method.Document [6] (Sibson R.A brief
description of natural neighbour interpolation[J].Interpreting Multivariate
Data,1981:21-36) the natural neighbor interpolation method recorded calculates the height value for obtaining interpolation point by formula 13:
W in formula 13iRepresent the weight coefficient of i-th of elevational point in neighborhood, ZiIt is then the height of i-th of elevational point in neighborhood
Journey, i represents i-th of elevational point in neighborhood, and Z is the height value of interpolation point.This method determines the principle of neighbor point and apart from nothing
Close, but proximity relations is determined with whether Tyson (Voronoi) polygon of interpolation point has common factor, and neighbor point
Weight be then interpolation point Thiessen polygon and neighbor point Thiessen polygon public area proportion (such as Fig. 5).Need
It should be noted that:Consider in the Thiessen polygon building process of interpolation point be a little, and the Thiessen polygon of neighbor point
Then interpolation point is foreclosed in building process.The basic characteristics of the interpolation method are that it has locality, using only to be inserted
A sample point subset around value point, weight coefficient ensure that result that interpolation obtains within the scope of used sample value.Its
Advantage is to be not in extrapolation trend and will not generate mountain peak, lowland, ridge or mountain valley that input sample point is not yet represented,
And obtained surface is smooth in other all positions in addition to input sample position;
Step 10:A series of buttock line of determining deviations (such as 0.2m) is chosen using obtained road DSM, is then utilized
International roughness index model tests the flatness value in section to express.In present invention specific implementation, using document [7]
(Sayers M W,Gillespie T D,Queiroz C A.The international road roughness
Experiment [J] .1986) record method calculate obtain section flatness value.The central principle of IRI indexes be four/
One vehicle model (such as Fig. 6), when a quarter vehicle is travelled with certain speed along road vertical section, in swashing for road gradient
Encourage under effect, system will produce vibration, calculate it and hung with (such as 1km) after fixing speed (such as 80km/h) traveling certain distance
The accumulation shift value of system is IRI, and unit is m/km.To solve the relative displacement of the suspension, second order vibration is set up micro-
Divide equation such as 14~formula of formula 15:
Wherein, Y is road surface elevation, ZsAnd ZuSpring carried mass displacement and nonspring carried mass displacement, μ, C, K are represented respectively1、K2
It is coefficient, and is recorded according to document [7], is had:
Wherein, msRepresent the size of spring carried mass, muRepresent the size of nonspring carried mass, KsRepresent connection spring carried mass and
The stiffness factor of the spring of nonspring carried mass, CsRepresent the linear damping coefficient of connection spring carried mass and nonspring carried mass, KsTable
Show the quality coefficient of the spring of connection tire.
Structural regime variableThen full scale equation can be turned to:
Wherein:
Formula 16 is a nonhomogeneous linear differential equation, and its corresponding state equation is:
Wherein,
PR=A-1(ST-I) B, (formula 19)
I in formula 18, formula 19 is unit matrix, with reference toFirst state of value and the gradientSequence, passes through the side of recursion
Formula can be tried to achieve any time successivelyAgain:It is rightKey can be tried to achieve by being integrated
Variable ZsAnd Zu, can be calculated by formula 20 and obtain IRI:
Document [8] (Wang Jianfeng, Song Hongxun, Ma Ronggui Evaluation of Pavement Evenness indexs IRI influence factor [J] Chongqing
University of communications's journal (natural science edition), 2012,31 (06):Above-mentioned calculating IRI method 1145-1148) is described, but is made
Data are traditional ground survey data.It is the calculating process for realizing IRI indexes in this example, it is necessary first to by DSM data
Carry out equidistant intervals sampling processing and obtain road surface elevation point sequence Y={ y0,y1,...,yn, and as the defeated of model
Enter data.Then state matrix ST and PR are calculated according to matrix A and B, then carrying out first-order difference computing to Y obtainsAnd make
Z (0)=[b 0 of b 0]T, b is the velocity amplitude (if road section length is less than 11m, taking 0m) at 11 meters away from starting point, so that sharp
It can be calculated with formula (17) and obtain t=0,1,2 ... the moment is correspondingValue, so it is rightIt is cumulative i.e. available by the period
The integral displacement of Each point in time, and to the displacement of Each point in time by carrying out asking difference to take absolute value and sum again formula 20 Suo Shi,
Finally average and solve IRI.
Step 11:The result of calculation of step 6 is only the IRI value sequences for having obtained single hatching, whole in order to be able to simulate
The flatness value of individual road surfaces, can be to each grid cell on road DSM according to the corresponding IRI of closest principle distribution
Value, the error size of this processing mode depends primarily on the spacing size of buttock line, and the smaller then error of spacing is also smaller.
The flatness spatial distribution in whole section to be measured is just can obtain by said process.
Step 12:Obtained road surface IRI values are classified with reference to following standard (such as table 2), that is, complete road pavement smooth
Spend the evaluation of quality.The classification sets standard to need to combine China《Asphalt highway maintenance technology specification》Maintenance of surface
The related request of quality is set, and this method is to combine《Asphalt highway maintenance technology specification》(JTJ 073.2-2001)
Specification is determined.Its corresponding distribution proportion can be calculated by counting the pixel number of each credit rating of section to be measured, and
The related specifications repaired with reference to road maintenance, so as to need which kind of maintenance measure to make science decision in the section.
The surface evenness quality grading of table 2 sets standard
In order to which the reliability to this method is evaluated, using following step to this method surface evenness quality-monitoring knot
Fruit carries out precision evaluation.
1) according to the flatness quality classification standard of setting, section to be measured selection it is a series of (comprising flatness quality compared with
It is good with it is poor, include health and Damage Types, every kind of road surface selects a number of ground to survey sample data) road surface sample
This is used as reference data;
2) choosing outward appearance, surface roughness, rutting depth, disease area, damaged condition this 5 indexs to reference data makes
Flatness quality evaluation is carried out to it with expert point rating method, and is used as reference value;
3) this road surface sample progress flatness value is calculated with this method is same, according to flatness quality grading
Standard, obtains the corresponding evaluation result of this method;
4) precision evaluation to flatness quality results can be analogous to the precision evaluation of classification of remote-sensing images result, and mix
The matrix that confuses is the most common form for evaluating classification of remote-sensing images precision.Therefore this method also uses confusion matrix road pavement
Flatness quality results are evaluated.Result is carried out by nicety of grading evaluation index overall accuracy (OA) and Kappa coefficients
Analysis, obtains the precision of this method, specific precision evaluation index calculating method is:
In formula 21, formula 22, N is total sample number, niiThe evaluation result of this method is represented with being i grades with reference to evaluation result
Sample size, ni.Then represent with reference to the total sample number that evaluation result is i grades, n.iThen represent the evaluation result of this method as
The total sample number of i grades.
Illustrate this method specific implementation and precision evaluation process by taking one, certain city country highway as an example below.
(1) data acquisition
Unmanned plane LiDAR point cloud data
The Highway house at county level is that road width is about near 44 ° 24 ' 47 of north latitude ", east longitude 85 ° 53 ' 47 " in geographical position
8m.The laser scanning that the Scout B1-100 depopulated helicopters that researcher produces on June 23rd, 2016 by Switzerland are carried
Instrument system (Rigel VUX-1LR) carries out experimental data collection, the basic parameter letter of the laser scanner directly over test block
Breath (source as shown in table 3:http://www.riegl.com/products/newriegl-vux-1-series/
newriegl-vux-1lr/).In order that the cloud data density that must be obtained is sufficiently high, unmanned plane during flying apart from road in experiment
Face be highly set to 30m, flying speed be 5m/s, scan angle be 110 °, scan frequency be 550KHz, the laser finally obtained
Pin point density is 300-600pts, scanning breadth sunny nothing (few) cloud on the day of being 85.7m, data acquisition.
The critical parameter information of the laser scanning instrument system of table 3
Ground data acquisition
Obtaining LiDAR point cloud data on the same day, researcher carries out characters of ground object point and road surface in flight range
The collection of sample data.Characters of ground object point data is mainly gathered along road periphery, by choosing some crucial ground object targets
(such as vehicle body, markings, pit, bulk crack) carries out physical dimension measurement, is positioned simultaneously using real time differential GPS equipment
Carry out notes archive;Then different types of road surface sample is chosen on flight range area road surface and (includes flatness quality
Preferably with poor, covering health and Damage Types) it is sampled and takes pictures and make feature description and statistic of classification, all data
Collecting work all carries out position record and preservation of taking pictures.This research acquires 10 crucial atural object characteristic points, 9 kinds of differences altogether
The road surface sample of type (sample point sum 52, every kind of road surface sample is about 5-7).Road surface sample type includes:Pit
(point big block and small bulk), collapse (a point building stones are completely exposed not to be exposed with building stones), crack (large fracture and gap), fiber crops
Face, sandstone are covered, surface evenness is general and flatness is good, as shown in Figure 7.
(2) data processing
The specific process step of data is as shown in figure 1, specific step implementation procedure is as follows.
The first step:Had as the attribute that the LiDAR point cloud packet acquired in Rigel VUX-1LR laser scanners contains:Three
Dimension coordinate, scan angle, echo quantity, echo times, laser intensity.The point cloud file format of original point cloud file and non-universal,
The software kit provided by Rigel companies enters row format to cloud data and is converted to general .las forms, such as Fig. 8 institutes
Show;
Second step:Visual identification is first passed through to the progress denoising of above-mentioned LiDAR point cloud data, in experimentation tentatively will
Noise spot is rejected, and it is that 0.2m is rejected to local abnormal elevation then to set search radius, is had a lot of social connections then according to road
The scanning angular region for the road surface point cloud that degree and unmanned plane during flying high computational are obtained is in ± 8 °, it is contemplated that course line has certain
Set scan angle threshold value to be ± 24 ° in deviation, experimentation to filter cloud data;
3rd step:The commercial Unmanned Aerial Vehicle Data processing software Pix4D Mapper developed using Pix4D companies of Switzerland are treated
Survey after the high resolution image data that unmanned plane is shot in section carry out image joint and registration process and can obtain section to be measured
Orthophotoquad (Fig. 9).Then feature extraction is carried out to the orthography, the gray value standard for calculating the wave band of RGB three is poor
As new characteristic wave bands, and with the band combination of RGB three, obtain 4 characteristic wave bands containing spectral information;
4th step:The point in a cloud is matched with the pixel in image using affine Transform Model, then by image
On band value be fused in cloud data, as a result as shown in Figure 10 (by taking blue band class information as an example);
5th step:Cloud data is carried out after Fusion Features, it is necessary to enter to the fused data with high resolution image data
The multiple dimensioned geometrical characteristic parameter of row is calculated.Because cloud data amount is very big, to save the time, while not influenceing result again
Reliability, this example is first optimized to the storage organization of cloud data, that is, builds Octree storage organization, is set one and is compared
Suitable numerical value (0.3m is used in this example) carries out resampling, the point after being vacuated as space memory cell length
Cloud data;
6th step:Centered on each point of above-mentioned cloud data, set search radius in [0.2m, 1.0m] and with
0.1m scans for for spacer unit, and the neighborhood point of search is then apart from small in cloud data after denoising with current search point
In the point set equal to search radius, local roughness (LDR) then respectively under calculating current scale, local dimension feature
(LDF), local depth displacement (LHD).With the adjustment of space search yardstick, you can obtain different scale geometric properties information
Cloud data;
7th step:The cloud data obtained to step 6 selects the partial dot cloud for covering all types of ground objects by visual observation
As training sample, dimensionality reduction (such as Figure 11) is then carried out to characteristic variable by feature selecting, it is easy to see that many after feature selecting
Separating capacities of the yardstick geometrical characteristic parameter LHD to all kinds of atural objects for other characteristic variables is most strong.Further by
Random Forest model carries out grader structure.This example sets the number of tree to be 200 in building process, and the depth of tree is 40.Its
It is secondary using constructed Random Forest model to the cloud data after vacuating is classified, so as to complete to vacuate minute of a cloud
Class process;
8th step:The point cloud vacuated is classified, it is necessary to take certain side due to up to the present having only completed
Method is classified to whole point clouds after denoising.Because two nearer object its attributes of space length are more similar, therefore this example
The class classified a little nearest apart from the point is assigned to the category attribute of the cloud data after denoising according to space nearest neighbouring rule
Other property value;
9th step:Classification screening is carried out to sorted cloud data, road cloud data is obtained, then pass through irregular three
Angle net builds road surface TIN models (Figure 12), and the DSM models of road surface are further obtained using natural neighbor interpolation, are
Meet IRI indexes and calculate needs while take into account data processing needs, raster resolution is set to 50mm, as a result such as Figure 13;
Tenth step:By interaction vector quantization mode in Arcgis10, inside road along road direction digital into
At intervals of a series of 0.125m buttock lines, the IRI values on these hatchings are then calculated using IRI exponential models.In order to
The difference reflected on each section longitudinal direction, need to set suitable computing unit length, this example is set to 10m when calculating IRI.
In order to which the IRI values of calculating are carried out the flatness spatial distribution result that rasterizing obtains whole section by effect of visualization again, such as scheme
14;
11st step:Standard is set according to the flatness quality grading that this method is set, to obtained surface evenness sky
Between distribution map carry out reclassification, the flatness quality distribution diagram on road surface is obtained, as shown in figure 15, by different quality grade
Grid cell carry out statistical analysis, it may be determined that each credit rating distribution proportion statistical chart in the section, as shown in figure 16;
(3) interpretation of result
In general, it is main based on " Fair " for the quality evaluation result of the road south of road section the characteristics of aspect of longitudinal direction, it is few
Subregion is " Failed ";Quality evaluation in the middle part of road is worst, and evaluation result is mainly " Poor " or " Failed ";And
The quality evaluation result of road north of a road section is then in the majority with " Fair " or " Failed ", and this is consistent with the distribution trend of flatness.
The characteristics of lateral aspects is then that the credit rating of approximate centerline is lower than both sides driveway, based on " Failed ", and is driven a vehicle
Evaluation result near road is then main based on " Fair " or " Poor ", and the pavement quality grade in road boundary area is then relative
It is relatively low, it is " Poor " or " Failed " that this subregion causes rough surface not due to reasons such as pavement depression, sand mixing
It is flat.From the point of view of each credit rating distribution map (Figure 15), it is known that the overall flatness quality in the section based on " Fair ", its
Distribution proportion is up to 41.1%, and secondly quality occupies for the section distribution proportion of " Poor " or " Failed ", is respectively 27.4% He
24.9%, and quality is minimum for the section proportion of " Good ", only 6.6%.And quality is " Poor " or " Failed "
Distribution proportion summation account for whole section more than half, this indicates that the overall mass deviation in the section, it is necessary to make great
Maintenance.
(4) precision evaluation
Using 52 road surface samples gathering on the spot as analyze data, by choose outward appearance, surface roughness, rutting depth,
This 5 indexs of disease area, damaged condition carry out flatness quality evaluation using expert point rating method to it, and as ginseng
Examine value.Then to these road surface samples are according to the GPS point coordinate of field survey and aid in high resolution image data in DSM
Orientation on map is carried out on model, then according to surface evenness space quality spatial distribution map to the smooth of these road surface samples
Degree credit rating is identified, then the recognition result of all road surface samples is counted, and thus obtains being used to carry out precision
The confusion matrix (such as Figure 17) of evaluation.Finally calculate the overall classification accuracy for having obtained this method respectively using formula 21 and formula 22
It is 0.65 for 75%, Kappa coefficients, shows that the overall evaluation result of this method is substantially the same with actual result, and uniformity journey
Degree is higher.
It should be noted that the purpose for publicizing and implementing example is that help further understands the present invention, but the skill of this area
Art personnel are appreciated that:Do not departing from the present invention and spirit and scope of the appended claims, various substitutions and modifications are all can
Can.Therefore, the present invention should not be limited to embodiment disclosure of that, and the scope of protection of present invention is with claim
The scope that book is defined is defined.
Claims (10)
1. a kind of fusion unmanned plane LiDAR and high-resolution remote sensing image surface evenness monitoring method, it is characterized in that, it is based on
LiDAR data and high-resolution remote sensing image that low latitude unmanned aerial vehicle platform is obtained, the spectral signature and LiDAR point cloud of fusion evaluation
Multiple dimensioned geometric properties build characteristic of division collection;Road surface point cloud is recycled to build subtle three-dimensional road surface digital surface model
(DSM) quick to calculate the flatness parameter for obtaining highway pavement, by international roughness index computational methods, combined ground is tested
The evaluation that data carry out surface evenness grade is demonstrate,proved, the quick monitoring and evaluation of pavement quality is realized;Specifically include following steps:
1.1) first with integrated LiDAR and multispectral camera low latitude experiment of UAV remote sensing system, obtain LiDAR point cloud data and
High-definition remote sensing image data;The LiDAR point cloud data are converted into .las file formats;To LiDAR point cloud data and
Remote sensing image data is pre-processed, and is then based on image data and is extracted spectral signature;
1.2) spectral signature is fused in the attribute of the cloud data, multiple dimensioned geometry is carried out to the cloud data
Feature extraction;By resulting geometric properties information and spectral signature information combination into point cloud classifications characteristic variable collection;
1.3) characteristic variable to the point cloud classifications carries out de-redundancy and dimensionality reduction, includes the search procedure and feature of character subset
The evaluation procedure of subset, obtains optimal feature subset;The suitable machine learning classification method of reselection carries out point cloud classifications, enters one
Step filtering and rough error, which are corrected, obtains more accurately road cloud data;
1.4) high-precision fine road surface model (DSM) is built by three-dimensional interpolation;
1.5) a series of buttock lines are chosen from obtained DSM model datas, international roughness index (IRI) model meter is utilized
Calculation obtains the corresponding IRI values of each hatching, and passes through the spatial distribution map that rasterizing handles the IRI values for obtaining road surface;
1.6) combined ground measured data is demarcated, and the surface evenness that satisfies the need distinguishes different grades, represents pavement quality from difference
To good type, remote sensing monitoring and the assessment of road pavement quality are realized.
2. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.1) especially by UAV flight
LiDAR scanner devices are in flying overhead gathered data near the ground, and obtaining data includes LiDAR point cloud data and the high score of the same period
Resolution image data;The high resolution image data are that digital camera shoots obtained high-definition digital photo, and with nobody
The corresponding posture that machine is provided and location information file;The LiDAR point cloud data at least meet following condition:
2.1) coordinate system of the LiDAR point cloud data is WGS84 coordinate systems or the local plane projection matched with locality
Coordinate system;
2.2) attribute information that the LiDAR point cloud packet contains includes longitude, latitude, elevation, scan angle, reflected intensity;
2.3) the point cloud density of the LiDAR point cloud data is more than 400 points/square metre.
3. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.1) pretreatment includes some cloud numbers
According to pretreatment and high resolution image data pretreatment;Main pretreatment process includes:The noise spot mistake of LiDAR point cloud
Filter;The abnormal elevation filtering of LiDAR point cloud;The scanning angular filter of LiDAR point cloud;The splicing of high resolution image with it is registering.
4. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.1) parameter of the spectral signature is
Reflect the parameter of type of ground objects difference, including:The gray value of the red, green, blue wave band of high resolution image;High resolution image
The standard deviation of the gray value of red, green, blue wave band;The reflection intensity values of laser point cloud.
5. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.2) data fusion is with a cloud
On the basis of the georeferencing of data, the spectral signature of image is fused in a cloud genera;Specifically use affine inverse transformation model
By in the band value information matches of each pixel on image to the cloud data coordinate of correspondence position.
6. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.3) geometric properties include it is local
Roughness (LDR), local dimension feature (LDF) and local depth displacement (LHD);The local roughness (LDR) is given directions in cloud
Some reference point to the reference point where some space scale under the distance of best-fitting plane that is formed of neighborhood point;
The local dimension feature (LDF) be give directions cloud in some reference point and this refer to neighborhood of a point point between be spatially with
One-dimensional, two-dimentional or distributed in three dimensions characteristic measure value;The local depth displacement (LHD) be give directions cloud in some reference point with should
All points that neighborhood point of the reference point under some space scale is constituted are concentrated, and are reached the best fit that the point set formed and are put down
The ultimate range in face and the difference of minimum range.
7. surface evenness monitoring method as claimed in claim 1, it is characterized in that, the multiple dimensioned geometric properties are specially:
When calculating some geometrical characteristic, multiple spatial analysis yardsticks are set;Fix some spatial analysis yardstick to be calculated, obtain one
Individual corresponding space geometry characteristic parameter;Geometric properties of the different types of ground objects under different spatial analysis yardsticks are different;By
This obtains multiple dimensioned geometric properties collection.
8. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.3) in, specifically, the dimensionality reduction enters
The search of row character subset uses Greedy strategy;To in character subset, when the characteristic variable newly added is not only between classified variable
Correlation it is high, while when the correlation again between the characteristic variable inside current subnet is relatively low, obtaining optimal feature subset;
The machine learning classification method is classified using random forest sorting algorithm to sample point cloud, then by the classification of sample point cloud
Attribute enters row interpolation by a cloud interpolation algorithm, obtains the classification of overall cloud data;When the classification of the overall cloud data
When as a result comprising the point cloud for being not belonging to road surface, hand inspection is carried out to sorted cloud data and rough error is corrected, and further
Filtered by category attribute, obtain more accurately road cloud data;Preferably, described cloud interpolation algorithm is adjacent using nature
Domain method.
9. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.4) the fine road surface digital surface
The raster resolution of model (DSM) should be controlled within 50mm, and height accuracy is within 15mm;Step 1.5) using international flat
When whole degree index (IRI) model calculates evenness of road surface angle value, the sampling interval control within 50mm, measuring unit length for 10~
20m。
10. surface evenness monitoring method as claimed in claim 1, it is characterized in that, step 1.6) obtain surface evenness i.e. IRI
After value, flatness quality classification standard is set, classified according to road evenness quality classification standard road pavement;Further
Precision evaluation is carried out using the method for classification of remote-sensing images precision evaluation, commented so as to obtain the corresponding flatness quality in section to be measured
Valency result;The method of the precision evaluation comprises the following steps:
10.1) according to the flatness quality classification standard, in section to be measured, selection same period road surface sample is used as reference data;Institute
State the same period road surface sample of selection comprising flatness quality preferably with poor, covering health and the road surface sample of Damage Types;Often
Plant road surface sample and randomly choose multiple samples, be used as reference data;
10.2) outward appearance, surface roughness, rutting depth, disease area, damaged condition index are chosen to reference data, uses layer
Fractional analysis (AHP) expert point rating method carries out flatness quality evaluation, will evaluate score value and is used as reference value;
10.3) the surface evenness monitoring method is used, the road surface sample is calculated and obtains flatness value;According to what is set
Flatness quality classification standard, obtains the corresponding flatness classification results in road surface;
10.4) result is calculated using precision evaluation index, finally gives the monitoring of the surface evenness monitoring method
Precision;The precision evaluation index includes overall accuracy (OA) or/and Kappa coefficients, and the value of the two is bigger to represent that precision is higher.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710257172.0A CN107092020B (en) | 2017-04-19 | 2017-04-19 | Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710257172.0A CN107092020B (en) | 2017-04-19 | 2017-04-19 | Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107092020A true CN107092020A (en) | 2017-08-25 |
CN107092020B CN107092020B (en) | 2019-09-13 |
Family
ID=59636957
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710257172.0A Active CN107092020B (en) | 2017-04-19 | 2017-04-19 | Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107092020B (en) |
Cited By (54)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107576960A (en) * | 2017-09-04 | 2018-01-12 | 苏州驾驶宝智能科技有限公司 | The object detection method and system of vision radar Spatial-temporal Information Fusion |
CN107657636A (en) * | 2017-10-16 | 2018-02-02 | 南京市测绘勘察研究院股份有限公司 | A kind of method that route topography figure elevational point is automatically extracted based on mobile lidar data |
CN108198230A (en) * | 2018-02-05 | 2018-06-22 | 西北农林科技大学 | A kind of crop and fruit three-dimensional point cloud extraction system based on image at random |
CN108492329A (en) * | 2018-03-19 | 2018-09-04 | 北京航空航天大学 | A kind of Three-dimensional Gravity is laid foundations cloud precision and integrity degree evaluation method |
CN108562913A (en) * | 2018-04-19 | 2018-09-21 | 武汉大学 | A kind of unmanned boat decoy detection method based on three-dimensional laser radar |
CN108564096A (en) * | 2018-04-26 | 2018-09-21 | 电子科技大学 | A kind of neighborhood fitting RCS sequence characteristic extracting methods |
CN108595553A (en) * | 2018-04-10 | 2018-09-28 | 红云红河烟草(集团)有限责任公司 | A kind of industrial number based on relevant database adopts time series data compression storage and decompression querying method |
CN109073744A (en) * | 2017-12-18 | 2018-12-21 | 深圳市大疆创新科技有限公司 | Landform prediction technique, equipment, system and unmanned plane |
CN109190673A (en) * | 2018-08-03 | 2019-01-11 | 西安电子科技大学 | The ground target classification method sentenced is refused based on random forest and data |
CN109446983A (en) * | 2018-10-26 | 2019-03-08 | 福州大学 | A kind of coniferous forest felling accumulation evaluation method based on two phase unmanned plane images |
CN109459759A (en) * | 2018-11-13 | 2019-03-12 | 中国科学院合肥物质科学研究院 | City Terrain three-dimensional rebuilding method based on quadrotor drone laser radar system |
CN109508508A (en) * | 2018-12-08 | 2019-03-22 | 河北省地矿局国土资源勘查中心 | Open-pit mine treatment and exploration design method |
CN109684929A (en) * | 2018-11-23 | 2019-04-26 | 中国电建集团成都勘测设计研究院有限公司 | Terrestrial plant ECOLOGICAL ENVIRONMENTAL MONITORING method based on multi-sources RS data fusion |
CN109829425A (en) * | 2019-01-31 | 2019-05-31 | 沈阳农业大学 | A kind of small scale terrain classification method and system of Farmland Landscape |
CN109945853A (en) * | 2019-03-26 | 2019-06-28 | 西安因诺航空科技有限公司 | A kind of geographical coordinate positioning system and method based on 3D point cloud Aerial Images |
CN110009054A (en) * | 2019-04-12 | 2019-07-12 | 南京大学 | A kind of airborne LiDAR point cloud classification method by different level using geometry and strength characteristic |
CN110070544A (en) * | 2019-06-06 | 2019-07-30 | 江苏省农业科学院 | One planting fruit-trees target three-dimensional data compensation method and compensation system |
CN110084116A (en) * | 2019-03-22 | 2019-08-02 | 深圳市速腾聚创科技有限公司 | Pavement detection method, apparatus, computer equipment and storage medium |
CN110197223A (en) * | 2019-05-29 | 2019-09-03 | 北方民族大学 | Point cloud data classification method based on deep learning |
CN110210500A (en) * | 2019-06-06 | 2019-09-06 | 上海黑塞智能科技有限公司 | A kind of point cloud classifications method based on the insertion of multiple dimensioned local feature |
CN110490888A (en) * | 2019-07-29 | 2019-11-22 | 武汉大学 | Freeway geometry Characteristic Vectors based on airborne laser point cloud quantify extracting method |
CN110516653A (en) * | 2019-09-03 | 2019-11-29 | 武汉天擎空间信息技术有限公司 | A kind of method for extracting roads based on multispectral airborne laser radar point cloud data |
CN110658850A (en) * | 2019-11-12 | 2020-01-07 | 重庆大学 | Greedy strategy-based flight path planning method for unmanned aerial vehicle |
CN110670461A (en) * | 2019-11-14 | 2020-01-10 | 上海宝冶建筑工程有限公司 | Method for detecting flatness of airport pavement |
CN110726998A (en) * | 2019-10-24 | 2020-01-24 | 西安科技大学 | Method for measuring mining subsidence basin in mining area through laser radar scanning |
CN110763223A (en) * | 2019-10-31 | 2020-02-07 | 苏州大学 | Sliding window based indoor three-dimensional grid map feature point extraction method |
CN110880202A (en) * | 2019-12-02 | 2020-03-13 | 中电科特种飞机系统工程有限公司 | Three-dimensional terrain model creating method, device, equipment and storage medium |
CN111381248A (en) * | 2020-03-23 | 2020-07-07 | 湖南大学 | Obstacle detection method and system considering vehicle bump |
CN111402387A (en) * | 2019-01-03 | 2020-07-10 | 迪普迈普有限公司 | Removing short timepoints from a point cloud of a high definition map for navigating an autonomous vehicle |
CN111638185A (en) * | 2020-05-09 | 2020-09-08 | 哈尔滨工业大学 | Remote sensing detection method based on unmanned aerial vehicle platform |
CN111784966A (en) * | 2020-06-15 | 2020-10-16 | 武汉烽火众智数字技术有限责任公司 | Personnel management and control method and system based on machine learning |
WO2020253171A1 (en) * | 2019-06-21 | 2020-12-24 | 哈尔滨工业大学 | Terrain modeling method and system amalgamating geometric characteristics and mechanical characteristics |
CN112446343A (en) * | 2020-12-07 | 2021-03-05 | 苏州工业园区测绘地理信息有限公司 | Vehicle-mounted point cloud road rod-shaped object machine learning automatic extraction method integrating multi-scale features |
CN112470032A (en) * | 2019-11-04 | 2021-03-09 | 深圳市大疆创新科技有限公司 | Topographic prediction method and device for undulating ground, radar, unmanned aerial vehicle and operation control method |
CN112633092A (en) * | 2020-12-09 | 2021-04-09 | 西南交通大学 | Road information extraction method based on vehicle-mounted laser scanning point cloud |
CN112686396A (en) * | 2020-12-15 | 2021-04-20 | 中公高科养护科技股份有限公司 | Method, medium and system for selecting pavement maintenance property based on disease number |
CN112819273A (en) * | 2020-12-31 | 2021-05-18 | 东北大学 | Regional highway network quality evaluation method and system |
CN113326865A (en) * | 2021-04-15 | 2021-08-31 | 东南大学 | Highway road surface disease three-dimensional information detecting system based on deep learning |
CN113325440A (en) * | 2021-05-06 | 2021-08-31 | 武汉大学 | Polarization laser radar data inversion method and system based on image recognition and signal characteristic decomposition |
CN113551636A (en) * | 2021-07-02 | 2021-10-26 | 武汉光谷卓越科技股份有限公司 | Flatness detection method based on abnormal data correction |
CN113689565A (en) * | 2021-10-21 | 2021-11-23 | 北京中科慧眼科技有限公司 | Road flatness grade detection method and system based on binocular stereo vision and intelligent terminal |
CN113962301A (en) * | 2021-10-20 | 2022-01-21 | 北京理工大学 | Multi-source input signal fused pavement quality detection method and system |
CN114155447A (en) * | 2021-12-02 | 2022-03-08 | 北京中科智易科技有限公司 | Artificial intelligence big data acquisition system |
CN114283070A (en) * | 2022-03-07 | 2022-04-05 | 中国铁路设计集团有限公司 | Method for manufacturing terrain section by fusing unmanned aerial vehicle image and laser point cloud |
CN114646305A (en) * | 2022-03-03 | 2022-06-21 | 湖南省测绘科技研究所 | Intelligent identification method for surveying and mapping behaviors of unmanned aerial vehicle |
CN114863695A (en) * | 2022-05-30 | 2022-08-05 | 中邮建技术有限公司 | Overproof vehicle detection system and method based on vehicle-mounted laser and camera |
CN114858214A (en) * | 2022-04-27 | 2022-08-05 | 中徽建技术有限公司 | Urban road performance monitoring system |
CN115168975A (en) * | 2022-08-17 | 2022-10-11 | 中国建筑第二工程局有限公司 | BIM technology-based road flatness quality control method and device |
CN115290012A (en) * | 2022-08-09 | 2022-11-04 | 南京市计量监督检测院 | Road surface flatness standard field magnitude tracing method based on laser point cloud data |
CN115323876A (en) * | 2022-07-15 | 2022-11-11 | 东南大学 | Airport cement concrete pavement flatness detection system |
US11618502B2 (en) * | 2019-03-28 | 2023-04-04 | Volkswagen Aktiengesellschaft | On-road localization methodologies and equipment utilizing road surface characteristics |
CN116612400A (en) * | 2023-05-30 | 2023-08-18 | 衡水金湖交通发展集团有限公司 | Road management method and system based on road flatness |
US11769238B2 (en) | 2020-05-18 | 2023-09-26 | Roadbotics, Inc. | Systems and methods for creating and/or analyzing three-dimensional models of infrastructure assets |
CN117437368A (en) * | 2023-12-22 | 2024-01-23 | 深圳大学 | Unmanned plane-based pavement evenness measuring method, system, terminal and medium |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH09264743A (en) * | 1996-03-29 | 1997-10-07 | Pub Works Res Inst Ministry Of Constr | Control method for finished form of pavement |
CN102103202A (en) * | 2010-12-01 | 2011-06-22 | 武汉大学 | Semi-supervised classification method for airborne laser radar data fusing images |
CN102163342A (en) * | 2011-04-07 | 2011-08-24 | 北京农业信息技术研究中心 | Fruit three morphological structure three-dimensional (3D) reconstruction method based on multi-scale measurement data |
CN102902883A (en) * | 2012-09-24 | 2013-01-30 | 北京师范大学 | Method for establishing bidirectional reflectance distribution function (BRDF) prototype library based on multi-angular measurement |
CN103017739A (en) * | 2012-11-20 | 2013-04-03 | 武汉大学 | Manufacturing method of true digital ortho map (TDOM) based on light detection and ranging (LiDAR) point cloud and aerial image |
CN104233935A (en) * | 2014-08-28 | 2014-12-24 | 吉林大学 | Identification method for pavement quality grade on basis of information of longitudinal section of road |
CN104867180A (en) * | 2015-05-28 | 2015-08-26 | 南京林业大学 | UAV and LiDAR integrated forest stand characteristic inversion method |
CN104933708A (en) * | 2015-06-07 | 2015-09-23 | 浙江大学 | Barrier detection method in vegetation environment based on multispectral and 3D feature fusion |
CN105488770A (en) * | 2015-12-11 | 2016-04-13 | 中国测绘科学研究院 | Object-oriented airborne laser radar point cloud filtering method |
CN106056591A (en) * | 2016-05-25 | 2016-10-26 | 哈尔滨工业大学 | Method for estimating urban density through fusion of optical spectrum image and laser radar data |
CN106127771A (en) * | 2016-06-28 | 2016-11-16 | 上海数联空间科技有限公司 | Tunnel orthography system and method is obtained based on laser radar LIDAR cloud data |
CN106296814A (en) * | 2015-05-26 | 2017-01-04 | 中国公路工程咨询集团有限公司 | Highway maintenance detection and virtual interactive interface method and system |
CN106355643A (en) * | 2016-08-31 | 2017-01-25 | 武汉理工大学 | Method for generating three-dimensional real scene road model of highway |
-
2017
- 2017-04-19 CN CN201710257172.0A patent/CN107092020B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH09264743A (en) * | 1996-03-29 | 1997-10-07 | Pub Works Res Inst Ministry Of Constr | Control method for finished form of pavement |
CN102103202A (en) * | 2010-12-01 | 2011-06-22 | 武汉大学 | Semi-supervised classification method for airborne laser radar data fusing images |
CN102163342A (en) * | 2011-04-07 | 2011-08-24 | 北京农业信息技术研究中心 | Fruit three morphological structure three-dimensional (3D) reconstruction method based on multi-scale measurement data |
CN102902883A (en) * | 2012-09-24 | 2013-01-30 | 北京师范大学 | Method for establishing bidirectional reflectance distribution function (BRDF) prototype library based on multi-angular measurement |
CN103017739A (en) * | 2012-11-20 | 2013-04-03 | 武汉大学 | Manufacturing method of true digital ortho map (TDOM) based on light detection and ranging (LiDAR) point cloud and aerial image |
CN104233935A (en) * | 2014-08-28 | 2014-12-24 | 吉林大学 | Identification method for pavement quality grade on basis of information of longitudinal section of road |
CN106296814A (en) * | 2015-05-26 | 2017-01-04 | 中国公路工程咨询集团有限公司 | Highway maintenance detection and virtual interactive interface method and system |
CN104867180A (en) * | 2015-05-28 | 2015-08-26 | 南京林业大学 | UAV and LiDAR integrated forest stand characteristic inversion method |
CN104933708A (en) * | 2015-06-07 | 2015-09-23 | 浙江大学 | Barrier detection method in vegetation environment based on multispectral and 3D feature fusion |
CN105488770A (en) * | 2015-12-11 | 2016-04-13 | 中国测绘科学研究院 | Object-oriented airborne laser radar point cloud filtering method |
CN106056591A (en) * | 2016-05-25 | 2016-10-26 | 哈尔滨工业大学 | Method for estimating urban density through fusion of optical spectrum image and laser radar data |
CN106127771A (en) * | 2016-06-28 | 2016-11-16 | 上海数联空间科技有限公司 | Tunnel orthography system and method is obtained based on laser radar LIDAR cloud data |
CN106355643A (en) * | 2016-08-31 | 2017-01-25 | 武汉理工大学 | Method for generating three-dimensional real scene road model of highway |
Non-Patent Citations (1)
Title |
---|
尚大帅等: "机载LiDAR点云数据与影像数据融合", 《测绘工程》 * |
Cited By (86)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107576960B (en) * | 2017-09-04 | 2021-03-16 | 赵建辉 | Target detection method and system for visual radar space-time information fusion |
CN107576960A (en) * | 2017-09-04 | 2018-01-12 | 苏州驾驶宝智能科技有限公司 | The object detection method and system of vision radar Spatial-temporal Information Fusion |
CN107657636A (en) * | 2017-10-16 | 2018-02-02 | 南京市测绘勘察研究院股份有限公司 | A kind of method that route topography figure elevational point is automatically extracted based on mobile lidar data |
CN109073744A (en) * | 2017-12-18 | 2018-12-21 | 深圳市大疆创新科技有限公司 | Landform prediction technique, equipment, system and unmanned plane |
WO2019119184A1 (en) * | 2017-12-18 | 2019-06-27 | 深圳市大疆创新科技有限公司 | Terrain prediction method, device and system, and drone |
CN108198230A (en) * | 2018-02-05 | 2018-06-22 | 西北农林科技大学 | A kind of crop and fruit three-dimensional point cloud extraction system based on image at random |
CN108492329B (en) * | 2018-03-19 | 2022-04-12 | 北京航空航天大学 | Three-dimensional reconstruction point cloud precision and integrity evaluation method |
CN108492329A (en) * | 2018-03-19 | 2018-09-04 | 北京航空航天大学 | A kind of Three-dimensional Gravity is laid foundations cloud precision and integrity degree evaluation method |
CN108595553A (en) * | 2018-04-10 | 2018-09-28 | 红云红河烟草(集团)有限责任公司 | A kind of industrial number based on relevant database adopts time series data compression storage and decompression querying method |
CN108595553B (en) * | 2018-04-10 | 2022-02-08 | 红云红河烟草(集团)有限责任公司 | Industrial data acquisition time sequence data compression storage and decompression query method based on relational database |
CN108562913A (en) * | 2018-04-19 | 2018-09-21 | 武汉大学 | A kind of unmanned boat decoy detection method based on three-dimensional laser radar |
CN108564096A (en) * | 2018-04-26 | 2018-09-21 | 电子科技大学 | A kind of neighborhood fitting RCS sequence characteristic extracting methods |
CN109190673B (en) * | 2018-08-03 | 2021-09-03 | 西安电子科技大学 | Ground target classification method based on random forest and data rejection |
CN109190673A (en) * | 2018-08-03 | 2019-01-11 | 西安电子科技大学 | The ground target classification method sentenced is refused based on random forest and data |
CN109446983A (en) * | 2018-10-26 | 2019-03-08 | 福州大学 | A kind of coniferous forest felling accumulation evaluation method based on two phase unmanned plane images |
CN109459759A (en) * | 2018-11-13 | 2019-03-12 | 中国科学院合肥物质科学研究院 | City Terrain three-dimensional rebuilding method based on quadrotor drone laser radar system |
CN109459759B (en) * | 2018-11-13 | 2020-06-30 | 中国科学院合肥物质科学研究院 | Urban terrain three-dimensional reconstruction method based on quad-rotor unmanned aerial vehicle laser radar system |
CN109684929A (en) * | 2018-11-23 | 2019-04-26 | 中国电建集团成都勘测设计研究院有限公司 | Terrestrial plant ECOLOGICAL ENVIRONMENTAL MONITORING method based on multi-sources RS data fusion |
CN109508508A (en) * | 2018-12-08 | 2019-03-22 | 河北省地矿局国土资源勘查中心 | Open-pit mine treatment and exploration design method |
CN109508508B (en) * | 2018-12-08 | 2024-03-22 | 河北省地质矿产勘查开发局国土资源勘查中心(河北省矿山和地质灾害应急救援中心) | Surface mine governance investigation design method |
CN111402387B (en) * | 2019-01-03 | 2023-09-12 | 辉达公司 | Removing short-time points from a point cloud for navigating a high-definition map of an autonomous vehicle |
CN111402387A (en) * | 2019-01-03 | 2020-07-10 | 迪普迈普有限公司 | Removing short timepoints from a point cloud of a high definition map for navigating an autonomous vehicle |
CN109829425A (en) * | 2019-01-31 | 2019-05-31 | 沈阳农业大学 | A kind of small scale terrain classification method and system of Farmland Landscape |
CN110084116B (en) * | 2019-03-22 | 2022-02-01 | 深圳市速腾聚创科技有限公司 | Road surface detection method, road surface detection device, computer equipment and storage medium |
CN110084116A (en) * | 2019-03-22 | 2019-08-02 | 深圳市速腾聚创科技有限公司 | Pavement detection method, apparatus, computer equipment and storage medium |
CN109945853A (en) * | 2019-03-26 | 2019-06-28 | 西安因诺航空科技有限公司 | A kind of geographical coordinate positioning system and method based on 3D point cloud Aerial Images |
CN109945853B (en) * | 2019-03-26 | 2023-08-15 | 西安因诺航空科技有限公司 | Geographic coordinate positioning system and method based on 3D point cloud aerial image |
US11618502B2 (en) * | 2019-03-28 | 2023-04-04 | Volkswagen Aktiengesellschaft | On-road localization methodologies and equipment utilizing road surface characteristics |
CN110009054A (en) * | 2019-04-12 | 2019-07-12 | 南京大学 | A kind of airborne LiDAR point cloud classification method by different level using geometry and strength characteristic |
CN110009054B (en) * | 2019-04-12 | 2021-01-29 | 南京大学 | Hierarchical airborne LiDAR point cloud classification method utilizing geometric and intensity features |
CN110197223A (en) * | 2019-05-29 | 2019-09-03 | 北方民族大学 | Point cloud data classification method based on deep learning |
CN110070544A (en) * | 2019-06-06 | 2019-07-30 | 江苏省农业科学院 | One planting fruit-trees target three-dimensional data compensation method and compensation system |
CN110210500A (en) * | 2019-06-06 | 2019-09-06 | 上海黑塞智能科技有限公司 | A kind of point cloud classifications method based on the insertion of multiple dimensioned local feature |
WO2020253171A1 (en) * | 2019-06-21 | 2020-12-24 | 哈尔滨工业大学 | Terrain modeling method and system amalgamating geometric characteristics and mechanical characteristics |
CN110490888A (en) * | 2019-07-29 | 2019-11-22 | 武汉大学 | Freeway geometry Characteristic Vectors based on airborne laser point cloud quantify extracting method |
CN110490888B (en) * | 2019-07-29 | 2022-08-30 | 武汉大学 | Highway geometric feature vectorization extraction method based on airborne laser point cloud |
CN110516653A (en) * | 2019-09-03 | 2019-11-29 | 武汉天擎空间信息技术有限公司 | A kind of method for extracting roads based on multispectral airborne laser radar point cloud data |
CN110726998A (en) * | 2019-10-24 | 2020-01-24 | 西安科技大学 | Method for measuring mining subsidence basin in mining area through laser radar scanning |
CN110726998B (en) * | 2019-10-24 | 2020-08-07 | 西安科技大学 | Method for measuring mining subsidence basin in mining area through laser radar scanning |
CN110763223A (en) * | 2019-10-31 | 2020-02-07 | 苏州大学 | Sliding window based indoor three-dimensional grid map feature point extraction method |
CN112470032A (en) * | 2019-11-04 | 2021-03-09 | 深圳市大疆创新科技有限公司 | Topographic prediction method and device for undulating ground, radar, unmanned aerial vehicle and operation control method |
WO2021087701A1 (en) * | 2019-11-04 | 2021-05-14 | 深圳市大疆创新科技有限公司 | Terrain prediction method and apparatus for undulating ground, and radar, unmanned aerial vehicle and operating control method |
CN110658850A (en) * | 2019-11-12 | 2020-01-07 | 重庆大学 | Greedy strategy-based flight path planning method for unmanned aerial vehicle |
CN110658850B (en) * | 2019-11-12 | 2022-07-12 | 重庆大学 | Greedy strategy-based flight path planning method for unmanned aerial vehicle |
CN110670461A (en) * | 2019-11-14 | 2020-01-10 | 上海宝冶建筑工程有限公司 | Method for detecting flatness of airport pavement |
CN110880202B (en) * | 2019-12-02 | 2023-03-21 | 中电科特种飞机系统工程有限公司 | Three-dimensional terrain model creating method, device, equipment and storage medium |
CN110880202A (en) * | 2019-12-02 | 2020-03-13 | 中电科特种飞机系统工程有限公司 | Three-dimensional terrain model creating method, device, equipment and storage medium |
CN111381248B (en) * | 2020-03-23 | 2023-06-09 | 湖南大学 | Obstacle detection method and system considering vehicle bump |
CN111381248A (en) * | 2020-03-23 | 2020-07-07 | 湖南大学 | Obstacle detection method and system considering vehicle bump |
CN111638185A (en) * | 2020-05-09 | 2020-09-08 | 哈尔滨工业大学 | Remote sensing detection method based on unmanned aerial vehicle platform |
CN111638185B (en) * | 2020-05-09 | 2022-05-17 | 哈尔滨工业大学 | Remote sensing detection method based on unmanned aerial vehicle platform |
US11769238B2 (en) | 2020-05-18 | 2023-09-26 | Roadbotics, Inc. | Systems and methods for creating and/or analyzing three-dimensional models of infrastructure assets |
CN111784966A (en) * | 2020-06-15 | 2020-10-16 | 武汉烽火众智数字技术有限责任公司 | Personnel management and control method and system based on machine learning |
CN112446343B (en) * | 2020-12-07 | 2024-03-15 | 园测信息科技股份有限公司 | Vehicle-mounted point cloud road shaft-shaped object machine learning automatic extraction method integrating multi-scale features |
CN112446343A (en) * | 2020-12-07 | 2021-03-05 | 苏州工业园区测绘地理信息有限公司 | Vehicle-mounted point cloud road rod-shaped object machine learning automatic extraction method integrating multi-scale features |
CN112633092A (en) * | 2020-12-09 | 2021-04-09 | 西南交通大学 | Road information extraction method based on vehicle-mounted laser scanning point cloud |
CN112686396B (en) * | 2020-12-15 | 2024-02-23 | 中公高科养护科技股份有限公司 | Pavement maintenance property selection method, medium and system based on disease quantity |
CN112686396A (en) * | 2020-12-15 | 2021-04-20 | 中公高科养护科技股份有限公司 | Method, medium and system for selecting pavement maintenance property based on disease number |
CN112819273A (en) * | 2020-12-31 | 2021-05-18 | 东北大学 | Regional highway network quality evaluation method and system |
CN112819273B (en) * | 2020-12-31 | 2023-11-28 | 东北大学 | Regional highway network quality evaluation method and system |
CN113326865A (en) * | 2021-04-15 | 2021-08-31 | 东南大学 | Highway road surface disease three-dimensional information detecting system based on deep learning |
CN113325440A (en) * | 2021-05-06 | 2021-08-31 | 武汉大学 | Polarization laser radar data inversion method and system based on image recognition and signal characteristic decomposition |
CN113325440B (en) * | 2021-05-06 | 2022-08-16 | 武汉大学 | Polarization laser radar data inversion method and system based on image recognition and signal characteristic decomposition |
CN113551636A (en) * | 2021-07-02 | 2021-10-26 | 武汉光谷卓越科技股份有限公司 | Flatness detection method based on abnormal data correction |
CN113962301A (en) * | 2021-10-20 | 2022-01-21 | 北京理工大学 | Multi-source input signal fused pavement quality detection method and system |
CN113962301B (en) * | 2021-10-20 | 2022-06-17 | 北京理工大学 | Multi-source input signal fused pavement quality detection method and system |
CN113689565B (en) * | 2021-10-21 | 2022-03-18 | 北京中科慧眼科技有限公司 | Road flatness grade detection method and system based on binocular stereo vision and intelligent terminal |
CN113689565A (en) * | 2021-10-21 | 2021-11-23 | 北京中科慧眼科技有限公司 | Road flatness grade detection method and system based on binocular stereo vision and intelligent terminal |
CN114155447B (en) * | 2021-12-02 | 2022-06-24 | 北京中科智易科技有限公司 | Artificial intelligence big data acquisition system |
CN114155447A (en) * | 2021-12-02 | 2022-03-08 | 北京中科智易科技有限公司 | Artificial intelligence big data acquisition system |
CN114646305B (en) * | 2022-03-03 | 2024-04-02 | 湖南省测绘科技研究所 | Intelligent recognition method for unmanned aerial vehicle mapping behavior |
CN114646305A (en) * | 2022-03-03 | 2022-06-21 | 湖南省测绘科技研究所 | Intelligent identification method for surveying and mapping behaviors of unmanned aerial vehicle |
CN114283070B (en) * | 2022-03-07 | 2022-05-03 | 中国铁路设计集团有限公司 | Method for manufacturing terrain section by fusing unmanned aerial vehicle image and laser point cloud |
CN114283070A (en) * | 2022-03-07 | 2022-04-05 | 中国铁路设计集团有限公司 | Method for manufacturing terrain section by fusing unmanned aerial vehicle image and laser point cloud |
CN114858214A (en) * | 2022-04-27 | 2022-08-05 | 中徽建技术有限公司 | Urban road performance monitoring system |
CN114858214B (en) * | 2022-04-27 | 2023-08-25 | 中徽建技术有限公司 | Urban road performance monitoring system |
CN114863695A (en) * | 2022-05-30 | 2022-08-05 | 中邮建技术有限公司 | Overproof vehicle detection system and method based on vehicle-mounted laser and camera |
CN115323876A (en) * | 2022-07-15 | 2022-11-11 | 东南大学 | Airport cement concrete pavement flatness detection system |
CN115290012A (en) * | 2022-08-09 | 2022-11-04 | 南京市计量监督检测院 | Road surface flatness standard field magnitude tracing method based on laser point cloud data |
CN115290012B (en) * | 2022-08-09 | 2023-03-31 | 南京市计量监督检测院 | Road surface flatness standard field magnitude tracing method based on laser point cloud data |
CN115168975B (en) * | 2022-08-17 | 2024-02-09 | 中国建筑第二工程局有限公司 | BIM technology-based pavement evenness quality control method and device |
CN115168975A (en) * | 2022-08-17 | 2022-10-11 | 中国建筑第二工程局有限公司 | BIM technology-based road flatness quality control method and device |
CN116612400B (en) * | 2023-05-30 | 2024-03-19 | 衡水金湖交通发展集团有限公司 | Road management method and system based on road flatness |
CN116612400A (en) * | 2023-05-30 | 2023-08-18 | 衡水金湖交通发展集团有限公司 | Road management method and system based on road flatness |
CN117437368A (en) * | 2023-12-22 | 2024-01-23 | 深圳大学 | Unmanned plane-based pavement evenness measuring method, system, terminal and medium |
CN117437368B (en) * | 2023-12-22 | 2024-04-26 | 深圳大学 | Unmanned plane-based pavement evenness measuring method, system, terminal and medium |
Also Published As
Publication number | Publication date |
---|---|
CN107092020B (en) | 2019-09-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107092020B (en) | Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image | |
Pan et al. | Detection of asphalt pavement potholes and cracks based on the unmanned aerial vehicle multispectral imagery | |
Biçici et al. | An approach for the automated extraction of road surface distress from a UAV-derived point cloud | |
Rastiveis et al. | Automated extraction of lane markings from mobile LiDAR point clouds based on fuzzy inference | |
CN106124454A (en) | A kind of bituminous paving aging performance monitoring method based on remote sensing image | |
Liu et al. | Pattern identification and analysis for the traditional village using low altitude UAV-borne remote sensing: Multifeatured geospatial data to support rural landscape investigation, documentation and management | |
CN109299673A (en) | The green degree spatial extraction method of group of cities and medium | |
CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
Yadav et al. | Identification of trees and their trunks from mobile laser scanning data of roadway scenes | |
Lin et al. | Leveraging optical and SAR data with a UU-Net for large-scale road extraction | |
Karantanellis et al. | 3D hazard analysis and object-based characterization of landslide motion mechanism using UAV imagery | |
Badenko et al. | Comparison of software for airborne laser scanning data processing in smart city applications | |
Zhao et al. | Combining ICESat-2 photons and Google Earth Satellite images for building height extraction | |
Zeybek et al. | Geometric feature extraction of road from UAV based point cloud data | |
Alshaiba et al. | Automatic manhole extraction from MMS data to update basemaps | |
Rutzinger et al. | Detection of high urban vegetation with airborne laser scanning data | |
Yadav et al. | Automatic urban road extraction from high resolution satellite data using object based ımage analysis: A fuzzy classification approach | |
Zhu | A pipeline of 3D scene reconstruction from point clouds | |
Wang et al. | GPS trajectory-based segmentation and multi-filter-based extraction of expressway curbs and markings from mobile laser scanning data | |
Wang et al. | Building extraction from LiDAR and aerial images and its accuracy evaluation | |
Liu et al. | Framework for Runway's True Heading Extraction in Remote Sensing Images Based on Deep Learning and Semantic Constraints | |
Bettineschi et al. | Clearence cairnfields forever: combining AI and LiDAR data in the Marcesina upland (northern Italy) | |
Huang et al. | Ground filtering algorithm for mobile LIDAR using order and neighborhood point information | |
Zhao et al. | An efficient pavement distress detection scheme through drone–ground vehicle coordination | |
DOWAJY et al. | Comparative Analysis of Road Scanning Techniques |
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 |