CN114541222A - Road network grade road surface evenness detection method based on multi-vehicle crowd-funded vibration data - Google Patents
Road network grade road surface evenness detection method based on multi-vehicle crowd-funded vibration data Download PDFInfo
- Publication number
- CN114541222A CN114541222A CN202210144895.0A CN202210144895A CN114541222A CN 114541222 A CN114541222 A CN 114541222A CN 202210144895 A CN202210144895 A CN 202210144895A CN 114541222 A CN114541222 A CN 114541222A
- Authority
- CN
- China
- Prior art keywords
- road
- vehicle
- flatness
- vibration data
- iri
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 37
- 238000012360 testing method Methods 0.000 claims abstract description 51
- 238000000034 method Methods 0.000 claims abstract description 39
- 238000004364 calculation method Methods 0.000 claims abstract description 26
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000003595 spectral effect Effects 0.000 claims description 17
- 238000009826 distribution Methods 0.000 claims description 16
- 238000005520 cutting process Methods 0.000 claims description 6
- 239000000725 suspension Substances 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000004971 IR microspectroscopy Methods 0.000 description 48
- 238000012060 immune response imaging Methods 0.000 description 48
- 230000008569 process Effects 0.000 description 7
- 238000005259 measurement Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000005065 mining Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000003746 surface roughness Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004092 self-diagnosis Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- E—FIXED CONSTRUCTIONS
- E01—CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
- E01C—CONSTRUCTION OF, OR SURFACES FOR, ROADS, SPORTS GROUNDS, OR THE LIKE; MACHINES OR AUXILIARY TOOLS FOR CONSTRUCTION OR REPAIR
- E01C23/00—Auxiliary devices or arrangements for constructing, repairing, reconditioning, or taking-up road or like surfaces
- E01C23/01—Devices or auxiliary means for setting-out or checking the configuration of new surfacing, e.g. templates, screed or reference line supports; Applications of apparatus for measuring, indicating, or recording the surface configuration of existing surfacing, e.g. profilographs
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Architecture (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention relates to a road network grade road surface evenness detection method based on multi-vehicle crowd funding vibration data, which comprises the following steps of: obtaining vibration data of a test vehicle and preprocessing the vibration data; taking the road section with the flatness index gradient larger than a preset threshold value in the road network to be detected as a road section with known flatness, and acquiring the flatness index of the road section with known flatness; acquiring a running track of a test vehicle, acquiring a vehicle number threshold value according to the running track of the test vehicle and known flatness road section information, extracting vehicles passing through the known flatness road section for more than or equal to the vehicle number threshold value as the iterative calculation vehicle, acquiring vibration data of the calculation vehicle and estimating vehicle parameters of the calculation vehicle; calculating a flatness index of the road section with unknown flatness based on the estimated vehicle parameters; and repeating the iteration for a plurality of times until the flatness indexes of all road sections in the road network to be detected are obtained. Compared with the prior art, the method has the advantages of high accuracy, good stability, low cost and the like.
Description
Technical Field
The invention relates to the field of pavement quality detection, in particular to a road network grade pavement evenness detection method based on multi-vehicle crowd-funded vibration data.
Background
The road surface evenness is an important index for reflecting the service performance and the driving comfort of the road. The traditional laser section flatness detection method can guarantee high measurement precision, but is expensive, limited in coverage range and needs to carry out specialized transformation on vehicles. With the development of sensing technology, mobile terminals represented by smart phones become important data acquisition approaches. However, the flatness measurement based on the vibration of the mobile phone of the bicycle has no essential difference from the testing mechanism of the traditional reaction method, and the special test vehicle still needs to be subjected to complex parameter calibration, so that the measurement efficiency is limited, the result discreteness is high, and the popularization and the application of the method are limited.
The measuring method of the road surface evenness is mainly divided into three types: subjective evaluation method, section method and reaction method. The subjective evaluation method adopts an expert scoring mode, has strong subjectivity and only serves as a road section evaluation reference. The section method is to measure the elevation change of the road surface under the running track of the vehicle, generally deploy precise sensing instruments such as a vehicle-mounted section plane instrument, a three-dimensional LiDAR and the like to a special detection vehicle, and calculate the international flatness index according to the section of the road surface. The section method has higher measurement accuracy, but the equipment has high production and application cost and limits the measurement conditions, is mainly suitable for measuring the performance of high-grade highways such as expressways, national and provincial trunk roads and the like, is only used for evaluating and sampling and detecting roads in urban roads, and is not suitable for the flatness inspection of large-scale and high-frequency urban road networks. The response method is to measure the flatness of the road surface by measuring the vibration response of the vehicle, such as a bump integrator, a BPR meter, and the like. In addition, it also belongs to a reaction method to obtain vehicle vibration through an accelerometer, a smart phone and the like. The reaction method has low cost and high efficiency, can be used as an important supplement of the section method, is sensitive to a mutation value, has high data processing difficulty, is easily influenced by parameters of a vehicle, a driving environment and a running state, and has low stability of a result.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a road network level road surface evenness detection method based on multi-car crowd-funded vibration data.
The purpose of the invention can be realized by the following technical scheme:
a road network level road surface flatness detection method based on multi-vehicle crowd funding vibration data comprises the following steps:
s1: obtaining vibration data of a test vehicle and preprocessing the vibration data;
s2: taking the road section with the flatness index gradient larger than a preset threshold value in the road network to be detected as a road section with known flatness, and acquiring the flatness index of the road section with known flatness;
s3: acquiring the running track of a test vehicle, acquiring a vehicle number threshold value according to the running track of the test vehicle and the known flatness road section information,
s4: extracting vehicles passing through the road section with the known flatness for more than or equal to a vehicle number threshold value to serve as calculation vehicles of the iteration, obtaining vibration data of the calculation vehicles and estimating vehicle parameters of the calculation vehicles;
s5: calculating the flatness index of the road section with unknown flatness based on the estimated vehicle parameters;
s6: and repeating the steps S3-S5 for a plurality of iterations until the flatness indexes of all road sections in the road network to be detected are obtained.
Preferably, the step S1 specifically includes:
s11: collecting Z-axis vibration data of a test vehicle, setting a reference sampling frequency f, reducing the frequency of the Z-axis vibration data to f when the frequency of the vibration data is higher than f, and increasing the frequency of the Z-axis vibration data to f when the frequency of the vibration data is lower than f;
s12: and calculating the power spectral density of the Z-axis vibration acceleration data of each test vehicle in a calculation period T, cutting the frequency band by an octave distance of the delta l distance, and calculating the power spectral density integral value in each octave.
Preferably, the method for acquiring the flatness index of the road section with the known flatness in the step S2 is laser detection, vibration detection or level meter detection.
Preferably, in the step S2, the road segment with a large flatness index gradient of 10% is selected as the road segment with known flatness, and the selecting step specifically includes:
when 10% of the total number of the road sections of the road network G to be detected is a non-integer, the total number num (R) of the road R with the known flatness is rounded upwards;
when the historical flatness data of the road sections in the road network G are known, selecting the section IRI with the highest flatness in the road networkmaxWith minimum distance IRIminAnd with (IRI)max-IRImin) /(num (R) -1) is interval, and IRI are extracted respectivelymin+N*(IRImax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2;
when the historical flatness data of the road sections in the road network G is unknown, the road section ACC with the maximum average vibration amplitude of the vehicle Z axis in the road network is selectedmaxWith the smallest road section ACCminAnd is combined with (ACC)max-ACCmin) /(num (R) -1) is interval, and IRI and ACC are respectively extractedmin+N*(ACCmax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2.
Preferably, the step S3 specifically includes:
acquiring the driving track of the test vehicle, and classifying the test vehicle set V according to the quantity passing through the known road sections: v ═ V1,V2,V3,...,Vmax},V1,V2,V3,...,VmaxRespectively, the vehicles pass through 1, 2 and 3 … max known road sections, wherein a frequency distribution histogram of the classified test vehicle set is drawn, and the number of the road sections which pass through the known flatness and correspond to the preset vehicle quantile value q is calculated as a vehicle number threshold value MqWhen M isqWhen the number is non-integer, rounding upwards, wherein M is more than or equal to 2q≤max。
Preferably, the step S4 specifically includes:
extraction ofThe vehicle is used as a calculation vehicle, and Z-axis vibration data of the calculation vehicle is acquired and countedCalculating vehicle parameters, saidThe number of the passing road sections is more than or equal to MqThe vehicle of (1).
Preferably, the vehicle parameters in step S4 include vehicle suspension parameters P, Q and vehicle model intercept b, and the vehicle parameters are obtained based on least square fitting in step S4:
wherein IRI is flatness index, K is total number of octaves, omega is angular velocity, delta l is distance of octaves, Sa(ω) is the power spectral density and,is the power spectral density integral value within each octave.
Preferably, the step S5 specifically includes:
s51: calculating based on estimated vehicle parameters and its power spectral density integral valueFlatness index IRI of middle vehicle passing through road section with unknown flatnessjWhen a plurality of vehicles pass through the same road section, calculating the estimated flatness standard deviation sigma of the flatness indexes of a plurality of test vehicles passing through the same road sectionj;
S52: according to the formula:screening the road sections with the largest number of passing vehicles and the smallest estimated flatness standard deviationWherein, size (IRI)j) The number of vehicles passed by the section j,
s53: computing a pass throughTaking the average value of flatness indexes of a plurality of vehicles on a road section asFlatness index of road sectionThe road segments add to the road segments of known flatness.
Preferably, after each iteration of step S6, the difference percentage of the parameter P, Q, b of each vehicle after two consecutive iterations is compared:
wherein, χP、χq、χbRespectively P, Q, b difference percentage between the current iteration and the last iteration, d is iteration number, if xP、χq、χbWhen any item exceeds the set threshold value, the vehicle elimination test vehicle set is subjected to d +1 iteration, and the eliminated vehicle is added into the test vehicle set again in the d +2 iteration.
Preferably, the vehicle speed of the test vehicle is not lower than 20km/h, and the vehicle speed variation coefficient of the test vehicle on different road sections within a preset time threshold value is not more than 10%.
Compared with the prior art, the invention has the following advantages:
(1) the method can realize the rapid estimation method of the large-scale road network surface evenness based on the crowd-funded vibration data and track information of the test vehicles and the small-sample road section evenness parameters of the crowd-funded vibration data of the test vehicles, can effectively solve the problems of time and labor consumption, high price and the like of the traditional detection method, improves the road surface evenness updating period, finds abnormal jolt and damage of the road in time, reduces the manpower and material resource consumption of a single vehicle complex calibration process, and has important significance for high-frequency digital detection of large-scale road surface performance.
(2) The invention is based on the data of a plurality of test vehicles for detection, has low cost, and based on the iterative crowd-funding detection process, by superposing the vibration data of different vehicles running on the same road section and mining the stable characteristics of the vibration signals in time-frequency distribution, the random error generated by single-vehicle single-frequency detection can be avoided, and the influence of abnormal data on the result can be reduced; through the vibration data trend of comparing that same vehicle traveles on different roads of road network, can the roughness distribution difference between the road section of analysis, improve the degree of accuracy and the confidence coefficient of model, can reduce the sensitivity to the sudden change value, effectively improve the stability and the reliability that road surface roughness detected, need not to carry out complicated parameter calibration to dedicated test vehicle, detection efficiency is high, road surface roughness perception method based on crowd funding data can effectively improve coverage and measurement efficiency, benefit from the data magnanimity simultaneously, the high frequency's characteristic, can further promote the roughness and predict the precision, reduce the result discreteness.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of a semi-supervised machine learning model employed in the present invention;
FIG. 3 is a schematic diagram illustrating the number of vehicles traveling on a road section in the embodiment;
FIG. 4 is a schematic diagram illustrating the number of road sections traveled by the vehicle in the embodiment;
FIG. 5 is a plot of vehicle power spectral density integral versus IRI in the present embodiment;
FIG. 6 is a diagram of the distribution of the estimated flatness parameters of each road section in this embodiment;
fig. 7 is a diagram illustrating the result of semi-supervised machine learning in this embodiment.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. Note that the following description of the embodiments is merely a substantial example, and the present invention is not intended to be limited to the application or the use thereof, and is not limited to the following embodiments.
Examples
Along with the popularization of smart phone equipment, vibration and position information of vehicles are acquired by utilizing a three-axis acceleration sensor, a gyroscope, a positioning module and the like which are arranged on a smart phone, and the data directly or indirectly reflect the smoothness level of a road surface, so that possibility is provided for large-scale and high-frequency road network level flatness sensing. Along with the popularization of the 'internet plus' travel service mode, a large amount of mass vehicle-mounted mobile phone crowd-funding data with high frequency, wide coverage and low cost is promoted in a large number of traffic trips. Although crowd funded data is lower than professional equipment in single-point acquisition accuracy, the mass of the crowd funded data is huge, and the crowd funded data can reflect stable characteristics in the environment. Compared with the flatness measuring method based on the single-vehicle mobile phone, the road flatness sensing method based on crowd funding data can effectively improve the coverage and the measuring efficiency, meanwhile, the characteristics of mass data and high frequency are benefited, the flatness estimation precision is expected to be further improved, and the result discreteness is reduced.
The crowd funding vibration data is a new carrier for traffic analysis and information mining, and the improvement of the quality of the data through the volume of the data is one of important means for data analysis. A large amount of sensing data are accumulated in a vehicle during running, and although the acquisition conditions of partial data and vehicle parameters are unknown, crowd funding data are obviously related in space-time distribution. For example, by superposing vibration data of different vehicles running on the same road section and mining the stable characteristics of the vibration signals in time-frequency distribution, random errors generated by single-vehicle single-frequency detection can be avoided, and the influence of abnormal data on results is reduced; by comparing the variation trends of the vibration data of the same vehicle running on different roads of the road network, the flatness distribution difference between road sections can be analyzed, and the accuracy and the confidence coefficient of the model are improved. Therefore, how to extract stable flatness parameters from high-dispersion crowd-funded vibration data is a key for performing road network level road surface flatness collaborative perception.
A road network level road surface flatness detection method based on multi-vehicle crowd-funded vibration data mainly estimates suspension parameters of vehicles on a road section through roads based on a small part of known flatness parameters, and when the vehicles with the estimated parameters run to other unknown roads, the flatness of the road section is calculated by a reactive method. In each cycle, the road section with the highest calculation confidence degree is selected, the road section is switched from an unknown road section set to a known road section level, the process is repeated until the road of the unknown road set is an empty set, namely all the roads are calculated, the rapid detection of the road surface evenness of the road network in a large range is realized, and an interactive semi-supervised learning model of road marking, vehicle positioning and road calculation is constructed.
In the description of this embodiment, IRI is used to refer to road flatness, and in practical application, other parameters may be used for the road flatness parameter. As shown in fig. 1, the present invention comprises the steps of:
s1: the method includes the steps of obtaining vibration data of a test vehicle and preprocessing the vibration data, setting a set of the test vehicle as a test vehicle set V in the embodiment, and carrying devices capable of obtaining the vibration data of the vehicle, such as a mobile phone, a vehicle sensor and an external sensor, on the test vehicle. In this embodiment, the test vehicles are not lower than 20km/h, when an individual vehicle has a speed of 20km/h, that is, when the individual vehicle runs at a low speed, the individual vehicle is removed from the vehicle set V, the vehicle speed of the road section is relatively stable, the variation coefficient of the vehicle speed of the road section is not more than 10% within the time T, the road section of the road network to be detected is a passable lane between adjacent nodes in the road network, different lanes are regarded as different road sections between the same nodes, for example, an urban road can be used as a boundary for cutting the road section at an intersection, an expressway can be used as a boundary for cutting the road section at a stake number and a toll station, and an elevated road can be used as a boundary for cutting the road section at a ramp position, etc.
Step S1 specifically includes:
s11: the method comprises the steps of collecting vibration data of a test vehicle, respectively marking the data as X-axis data, Y-axis data and Z-axis data according to the axial direction of a sensor, extracting the Z-axis vibration data of the test vehicle, setting a reference sampling frequency f, reducing the frequency of the Z-axis vibration data to f by adopting an anti-aliasing frequency reduction method when the frequency of the vibration data is higher than f, and increasing the frequency of the Z-axis vibration data to f by adopting a linear interpolation method when the frequency of the vibration data is lower than f. Meanwhile, a flatness alternation calculation period T is set.
The vehicle vibration data acquisition can be realized in the modes of mobile phone application acquisition, peripheral vibration sensor acquisition, vehicle-mounted self-diagnosis system reading and the like, the data frequency of the Z axis is not lower than 1Hz, and the data frequency of the Z axis is more than 20Hz in the embodiment. The anti-aliasing downconversion adopts a finite long impulse response (FIR) digital low-pass filter, the calculation period T is selected according to the road network scale and is not less than 30 minutes, and the calculation period is 24 hours in the embodiment.
In order to avoid the influence of the sample size on the detection result, the sample capacity of the known road section can be adjusted in the actual operation, and if the traffic flow of the road network is small and the vehicle vibration data is small, the sample capacity of the known road section is increased to 20% or more; when the known sample is limited, the calculation period T is expanded, so that the data volume requirement is met.
S12: and calculating the power spectral density of the Z-axis vibration acceleration data of each test vehicle in a calculation period T, cutting a frequency band by an octave distance of a delta l distance, wherein the length of the octave does not exceed 10Hz, and calculating the power spectral density integral value in each octave.
The vehicle parameters are calibrated, and a small amount of road network section flatness parameters are required to be known, so that S2: and taking the road sections with the flatness index gradient larger than a preset threshold value in the road network G to be detected as the road sections with the known flatness, wherein the set of the road sections is R, and the set of the other road sections with unknown flatness is U, and acquiring the flatness index of the road sections with the known flatness R.
In this embodiment, S2 selects a road segment with a large flatness index gradient of 10% as a known flatness road segment, and obtains the flatness index by using laser detection, vibration detection, level detection, and other methods, where the flatness index includes, but is not limited to, parameters that can objectively evaluate the smoothness of a road surface, such as an international flatness index and a driving comfort index, and the selecting step of the known flatness road segment specifically includes:
when 10% of the total number of the road sections of the road network G to be detected is a non-integer, the total number num (R) of the road R with the known flatness is rounded upwards;
when the historical flatness data of the road sections in the road network G are known, selecting the section IRI with the highest flatness in the road networkmaxWith minimum distance IRIminAnd with (IRI)max-IRImin) /(num (R) -1) is interval, and IRI are extracted respectivelymin+N*(IRImax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2;
when the historical flatness data of the road sections in the road network G is unknown, the road section ACC with the maximum average vibration amplitude of the vehicle Z axis in the road network is selectedmaxWith the smallest road section ACCminAnd is combined with (ACC)max-ACCmin) /(num (R) -1) is interval, and IRI and ACC are respectively extractedmin+N*(ACCmax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2.
S3: the method includes the steps of obtaining a running track of a test vehicle, obtaining a vehicle number threshold value according to the running track of the test vehicle and known flatness road section information, and specifically includes the following steps:
acquiring the driving track of the test vehicle, and classifying the test vehicle set V according to the quantity passing through the known road sections: v ═ V1,V2,V3,...,Vmax},V1,V2,V3,...,VmaxRespectively, the vehicles pass through 1, 2 and 3 … max known road sections, wherein, a frequency distribution histogram of the classified test vehicle set is drawn, and the number of road sections passing through the known flatness corresponding to the preset vehicle quantile value q is calculated as a vehicle number threshold value MqWhen M isqWhen the number is non-integer, rounding up, M is more than or equal to 2q≤max。
In this example, q is 85%.
S4: extraction ofThe vehicle in (1) is used as a calculation vehicle, the Z-axis vibration data of the calculation vehicle is obtained, vehicle parameters are calculated by a calculator,the number of the passing road sections is more than or equal to MqThe vehicle of (1).
The vehicle parameters comprise vehicle suspension parameters P, Q and vehicle model intercept b, wherein b mainly represents the influence of a vehicle engine at a certain stable speed, and the vehicle parameters are obtained based on least square fitting in step S4:
wherein IRI is flatness index, K is total number of octaves, omega is angular velocity, delta l is distance of octaves, Sa(ω) is the power spectral density and,is the power spectral density integral value within each octave.
In the above steps S1-4, the vehicle parameters are estimated based on the small sample known flatness road section, and the vehicle vibration is the most direct and common expression of road surface flatness, but is affected by the suspension parameters of the vehicle and the performance of the engine, and the same road surface flatness may generate different vibration characteristics, so the estimated vehicle parameters are the first step of road network level road surface flatness detection, and therefore the above formula is a linear equation with three parameters of P, Q, and b, and the model parameters can be fitted by using the least square method, so that the vehicle at least passes through three known road sections to obtain IRI and vibration parameters of the three road sections, and the three parameters of P, Q, and b can be calculated. According to the statistical principle, when the number of road sections passed by the vehicle is larger, namely the sample size of the fitting model is larger, the confidence coefficient of the fitting result is higher. Furthermore, the method calculates the number of the roads passing through the known label corresponding to the 85% vehicle number quantile, so as to ensure that the estimated vehicle parameter confidence is at a higher level and avoid the influence of the parameter estimation error on the subsequent calculation.
S5: calculating the flatness index of the road section with unknown flatness based on the estimated vehicle parameters, which specifically comprises the following steps:
s51: calculating based on estimated vehicle parameters and its power spectral density integral valueFlatness index IRI of middle vehicle passing through road section with unknown flatnessjWhen a plurality of vehicles pass through the same road section, calculating the estimated flatness standard deviation sigma of the flatness indexes of a plurality of test vehicles passing through the same road sectionj;
S52: according to the formula:screening the road sections with the largest number of passing vehicles and the smallest estimated flatness standard deviationWherein, size (IRI)j) The number of vehicles passed by the section j,
s53: computing a pass throughAs a mean value of the flatness index IRI of a plurality of vehicles of a road sectionFlatness index of road sectionThe road segment is added to the road segment R of known flatness and removed from U.
In step S5, road parameters of the flatness position are calculated based on the estimated vehicle parameters, after S1-S4, multiple vehicle parameters in the road network are all estimated, when the vehicles enter the road section with unknown flatness parameters, vibration data of the vehicles are collected, the IRI of the position road section can be calculated by using the IRI formula, and since the same unknown road section is driven by multiple vehicles with known parameters, the IRI can be estimated for multiple times. Due to errors in the estimation of suspension parameters of the vehicle itself, anThe speeds of multiple vehicles passing through the same road section may have differences, and the estimated IRI is also different. To improve the confidence level of model estimation, consensus is basedAnd selecting the road with the most acquisition times and the lowest distribution dispersion in all the estimated road sections, namely the road with the most stable result.
S6: and repeating the steps S3-S5 for a plurality of iterations until U is an empty set. After each iteration, only one road segment with the highest confidence in the result is moved from the unknown set to the known set. Therefore, after each iteration, the number of unknown road segments is-1, the known road segment data is +1, the iteration process of the method is detected based on a semi-supervised machine learning model, and if the total number of the unknown road segments in the road network is num (U), num (U) iterations are required to calculate the flatness parameters of all the road segments, and the iteration process is shown in fig. 2.
In each iteration, the vehicle parameters and the flatness of the unknown road section need to be recalculated. In this embodiment, after each iteration of S6, the difference percentage of the parameter P, Q, b of each vehicle after two consecutive iterations is compared:
wherein, χP、χq、χbRespectively P, Q, b difference percentage between the current iteration and the last iteration, d is iteration number, if xP、χq、χbWhen any item exceeds the set threshold value, the vehicle elimination test vehicle set is subjected to the (d + 1) th iteration, and the eliminated vehicles are added again in the (d + 2) th iterationGo to the test vehicle set.
In this embodiment, the specific implementation process of the scheme of the present invention is as follows:
(1) the vibration data of 500 vehicles in the road network are collected through the vehicle-mounted mobile phone APP and the fixed support, the data collection frequency is 200Hz, 50 road sections are contained in the road network, the number is 1-50, the flatness of the known road sections is shown in the table 1, and all known flatness parameters are international flatness index IRI data collected by the laser detection vehicle.
TABLE 1 road segment with known flatness parameters
Number of | 4 | 9 | 16 | 28 | 29 |
IRI | 3.987 | 4.911 | 4.152 | 2.917 | 5.264 |
|
30 | 35 | 39 | 44 | 48 |
IRI | 5.913 | 6.539 | 9.860 | 2.762 | 3.398 |
(2) The distribution of the vehicles passing through each road section and the number of the parts of the road section passing through the vehicles are counted, as shown in fig. 3 and 4, M can be obtained by accumulating the distribution map85%For 3, i.e. all vehicles passing a known road section greater than 3 are selected, the vibration data collected by the mobile phone APP, fig. 5 shows the power spectral density integral of three vehicles numbered 3, 280, 400 in relation to the known IRI, and the model parameters P, Q, b are estimated by means of least squares using the multiple linear regression equation (1).
(3) The road flatness IRI of the vehicle with the estimated parameters is calculated based on equation (1) using the vibration data of the vehicle passing through the unknown IRI section, as shown in fig. 7. Calculating the estimated IRI standard deviation of each road section,and assigning the IRI of the road section, and iterating to finally obtain the traversal calculation results of all road sections of the road network, as shown in FIG. 6.
(4) Model accuracy evaluation
The average relative error of the flatness of the road network surface calculated according to the model algorithm is 9.71%, the relative error of the highest road section appears on the number 40 and is 46.5%, the absolute error is 1.15m/km, the IRI of the position is considered to be low, most of IRIs of the trained road sections are distributed around 4.5m/km, and therefore the expressive force is poor at the position with low IRI. As can be seen from fig. 6, the semi-supervised learning model has excellent overall performance, and basically conforms to the actual IRI distribution condition of the road surface, most of the relative errors can be kept within 10%, and the fall-down state of the errors occurs at the higher position and the lower position of the IRI, considering that the IRI distribution of the training samples simulates the actual normal distribution form of the road, so that the volumes of samples with too high IRI and too low IRI are less, and are difficult to be fully expressed in the training process, but the overall trend still conforms to the real road section distribution condition.
The above embodiments are merely examples and do not limit the scope of the present invention. These embodiments may be implemented in other various manners, and various omissions, substitutions, and changes may be made without departing from the technical spirit of the present invention.
Claims (10)
1. A road network level road surface flatness detection method based on multi-vehicle crowd funding vibration data is characterized by comprising the following steps:
s1: obtaining vibration data of a test vehicle and preprocessing the vibration data;
s2: taking the road section with the flatness index gradient larger than a preset threshold value in the road network to be detected as a road section with known flatness, and acquiring the flatness index of the road section with known flatness;
s3: acquiring a running track of a test vehicle, and acquiring a vehicle number threshold according to the running track of the test vehicle and the known flatness road section information;
s4: extracting vehicles passing through the road section with the known flatness for more than or equal to a vehicle number threshold value to serve as calculation vehicles of the iteration, obtaining vibration data of the calculation vehicles and estimating vehicle parameters of the calculation vehicles;
s5: calculating a flatness index of the road section with unknown flatness based on the estimated vehicle parameters;
s6: and repeating the steps S3-S5 for a plurality of iterations until the flatness indexes of all road sections in the road network to be detected are obtained.
2. The road network level road surface evenness detection method based on multi-car crowd-funded vibration data according to claim 1, wherein the step S1 specifically includes:
s11: collecting Z-axis vibration data of a test vehicle, setting a reference sampling frequency f, reducing the frequency of the Z-axis vibration data to f when the frequency of the vibration data is higher than f, and increasing the frequency of the Z-axis vibration data to f when the frequency of the vibration data is lower than f;
s12: and calculating the power spectral density of the Z-axis vibration acceleration data of each test vehicle in a calculation period T, cutting the frequency band by an octave distance of the delta l distance, and calculating the power spectral density integral value in each octave.
3. The road network level road surface evenness detection method based on multi-car crowd-funded vibration data as claimed in claim 1, wherein the method for obtaining the evenness index of the road section with known evenness in step S2 is laser detection, vibration detection or level meter detection.
4. The road network level road surface evenness detection method based on the multi-car crowd-funded vibration data as claimed in claim 1, wherein the step S2 is to select the road segment with a larger evenness index gradient by 10% as the known evenness road segment, and the step of selecting specifically includes:
when 10% of the total number of the road sections of the road network G to be detected is a non-integer, the total number num (R) of the road R with the known flatness is rounded upwards;
when the historical flatness data of the road sections in the road network G are known, selecting the section IRI with the highest flatness in the road networkmaxWith minimum distance IRIminAnd with (IRI)max-IRImin) /(num (R) -1) is interval, and IRI are extracted respectivelymin+N*(IRImax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2;
when the historical flatness data of the road sections in the road network G is unknown, the road section ACC with the maximum average vibration amplitude of the vehicle Z axis in the road network is selectedmaxWith the smallest road section ACCminAnd is combined with (ACC)max-ACCmin) /(num (R) -1) is interval, and IRI and ACC are respectively extractedmin+N*(ACCmax-IRImin) (num (R) -1) nearest road segment as known road segment, N is a natural integer from 1 to num (R) -2.
5. The method for detecting the road network level road surface evenness based on the multi-car crowd funding vibration data according to claim 1, wherein the step S3 specifically includes:
acquiring the driving track of the test vehicle, and classifying the test vehicle set V according to the quantity passing through the known road sections: v ═ V1,V2,V3,...,Vmax},V1,V2,V3,...,VmaxRespectively, the vehicles pass through 1, 2 and 3 … max known road sections, wherein, a frequency distribution histogram of the classified test vehicle set is drawn, and the number of road sections passing through the known flatness corresponding to the preset vehicle quantile value q is calculated as a vehicle number threshold value MqWhen M isqWhen the number is non-integer, rounding upwards, wherein M is more than or equal to 2q≤max。
6. The road network level road surface evenness detection method based on multi-car crowd-funded vibration data according to claim 1, wherein the step S4 specifically includes:
7. The method as claimed in claim 1, wherein the vehicle parameters in step S4 include vehicle suspension parameters P, Q and vehicle model intercept b, and the vehicle parameters are obtained based on least square fitting in step S4:
8. The road network level road surface evenness detection method based on multi-car crowd-funded vibration data according to claim 1, wherein the step S5 specifically includes:
s51: calculating based on estimated vehicle parameters and its power spectral density integral valueFlatness index IRI of middle vehicle passing through road section with unknown flatnessjWhen a plurality of vehicles pass through the same road section, calculating the estimated flatness standard deviation sigma of the flatness indexes of a plurality of test vehicles passing through the same road sectionj;
S52: according to the formula:screening the road sections with the largest number of passing vehicles and the smallest estimated flatness standard deviationWherein, size (IRI)j) The number of vehicles passing by the road segment j,
9. The road network level road surface evenness detection method based on the multi-car crowd-funded vibration data as claimed in claim 1, wherein after each iteration of step S6, the difference percentage of the parameter P, Q, b of each car after two consecutive iterations is compared:
wherein, χP、χq、χbRespectively P, Q, b difference percentage between the current iteration and the last iteration, d is iteration number, if xP、χq、χbWhen any item exceeds the set threshold value, the vehicle elimination test vehicle set is subjected to d +1 iteration, and the eliminated vehicle is added into the test vehicle set again in the d +2 iteration.
10. The method as claimed in claim 1, wherein the vehicle speed of the test vehicle is not lower than 20km/h, and the vehicle speed variation coefficient of the test vehicle on different road sections within a preset time threshold is not more than 10%.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210144895.0A CN114541222B (en) | 2022-02-17 | 2022-02-17 | Road network grade pavement flatness detection method based on multi-vehicle crowd funding vibration data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210144895.0A CN114541222B (en) | 2022-02-17 | 2022-02-17 | Road network grade pavement flatness detection method based on multi-vehicle crowd funding vibration data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114541222A true CN114541222A (en) | 2022-05-27 |
CN114541222B CN114541222B (en) | 2024-01-26 |
Family
ID=81675291
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210144895.0A Active CN114541222B (en) | 2022-02-17 | 2022-02-17 | Road network grade pavement flatness detection method based on multi-vehicle crowd funding vibration data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114541222B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104120644A (en) * | 2013-04-26 | 2014-10-29 | 同济大学 | Gravitational acceleration sensor based road surface flatness detection method |
JP2015028456A (en) * | 2013-07-31 | 2015-02-12 | 国立大学法人 東京大学 | Road surface evaluation device and method |
CN108839657A (en) * | 2018-06-05 | 2018-11-20 | 吉林大学 | A method of online recognition road roughness information is responded based on automobile vibration |
CN111275363A (en) * | 2020-03-27 | 2020-06-12 | 山东智路信息科技有限公司 | Road network flatness evaluation system and method based on vehicle vibration data |
CN113434954A (en) * | 2021-06-15 | 2021-09-24 | 同济大学 | Calibration method of vibrating type pavement flatness test vehicle |
-
2022
- 2022-02-17 CN CN202210144895.0A patent/CN114541222B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104120644A (en) * | 2013-04-26 | 2014-10-29 | 同济大学 | Gravitational acceleration sensor based road surface flatness detection method |
JP2015028456A (en) * | 2013-07-31 | 2015-02-12 | 国立大学法人 東京大学 | Road surface evaluation device and method |
CN108839657A (en) * | 2018-06-05 | 2018-11-20 | 吉林大学 | A method of online recognition road roughness information is responded based on automobile vibration |
CN111275363A (en) * | 2020-03-27 | 2020-06-12 | 山东智路信息科技有限公司 | Road network flatness evaluation system and method based on vehicle vibration data |
CN113434954A (en) * | 2021-06-15 | 2021-09-24 | 同济大学 | Calibration method of vibrating type pavement flatness test vehicle |
Also Published As
Publication number | Publication date |
---|---|
CN114541222B (en) | 2024-01-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104164829B (en) | Detection method of road-surface evenness and intelligent information of road surface real-time monitoring system based on mobile terminal | |
Alessandroni et al. | Smartroadsense: Collaborative road surface condition monitoring | |
CN107330217B (en) | Mesoscopic oil consumption prediction method based on RBFNN | |
CN104792937B (en) | Bridge head bump detection evaluation method based on vehicle-mounted gravitational acceleration sensor | |
CN105005694B (en) | A kind of bridge fatigue life frequency-domain analysis method based on dynamic weighing system | |
CN107207009A (en) | Pavement state determination methods | |
CN104233935B (en) | A kind of pavement quality grade discrimination method based on profile of road information | |
Pawar et al. | IRI (international roughness index): an indicator of vehicle response | |
CN101275900A (en) | Method for recognizing road surface types based on vehicle wheel vibration | |
CN109829252B (en) | Influence line identification-based bridge condition rapid rating method | |
Song et al. | Characteristics of low-speed vehicle-specific power distributions on urban restricted-access roadways in beijing | |
CN110852542B (en) | Road flatness calculation method and system | |
CN111127906B (en) | Intelligent road surface management system and method based on Internet of things | |
He et al. | Freeway multisensor data fusion approach integrating data from cellphone probes and fixed sensors | |
CN103290768A (en) | Flatness Measuring Instrument and Measuring Method | |
D'Hooge et al. | Application of real-world wind conditions for assessing aerodynamic drag for on-road range prediction | |
Daraghmi et al. | Crowdsourcing-based road surface evaluation and indexing | |
CN112100721B (en) | Bridge structure modal identification method based on mobile crowd sensing and deep learning | |
CN113340392B (en) | Vehicle load detection method and device based on acceleration sensor | |
CN110867075A (en) | Method for evaluating influence of road speed meter on reaction behavior of driver under rainy condition | |
CN114541222B (en) | Road network grade pavement flatness detection method based on multi-vehicle crowd funding vibration data | |
Shaw et al. | Instantaneous fuel consumption estimation using smartphones | |
CN115544763A (en) | Road surface flatness prediction method, system and medium | |
CN114323512B (en) | Heavy-load vehicle identification method and system | |
Yunusov et al. | Signal processing and conditioning tools and methods for road profile assessment |
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 |