A full-width flatness calculation method and device

Abstract

The present application discloses a full-width road surface flatness calculation method and device. The method includes: obtaining the original elevation value point cloud data of the road surface; processing the point cloud data noise; determining the lane category, calculating the wheel track distribution probability, and obtaining the longitudinal section weight; calculating the equivalent longitudinal section elevation value of the lane based on the longitudinal section weight; and calculating the flatness index based on the equivalent longitudinal section elevation value. The present application realizes a comprehensive consideration of the use performance of the entire lane, establishes a new full-width road surface flatness calculation model, and can provide a more accurate and comprehensive road surface flatness evaluation.

Description

Technical Field

[0001] The present invention relates to the field of highways, and in particular to a full-width road surface flatness calculation method and device. Background Art

[0002] The road surface flatness is a key index to measure the road surface performance. It objectively reflects the fluctuation condition of the road surface elevation, thereby revealing the driving quality and maintenance condition of the road surface. The quality of the flatness is directly related to the driving comfort and safety, and at the same time, it also deeply affects the service life of the road surface, the wear degree of the tires, and the fuel consumption efficiency of the vehicle. For this reason, scientifically and accurately detecting the road surface flatness plays an indispensable and important role in ensuring driving safety and improving the road service level.

[0003] There are various definitions of road surface flatness. For example, the Standing Technical Committee of the International Road Conference (PIARC) proposed a classification method for road surface geometric characteristics based on geometric resolution at the 1987 Brussels World Road Congress; the International Organization for Standardization (ISO) defined road surface flatness as the deviation between the surface and the true plane surface in the ISO-13473 standard; the American Society for Testing and Materials (ASTM E 867) defined flatness from the perspective of the influence of vehicle dynamics. In China, the traditional road flatness evaluation method generally follows the "Code for Field Testing of Highway Subgrade and Pavement", which clearly stipulates 3 test methods and corresponding indexes for measuring flatness, namely the three-meter straightedge method, the continuous flatness meter test method, and the vehicle-mounted bump integrator test method. The "Standard for Quality Inspection and Evaluation of Highway Engineering" (JTG F80 / 1-2017) stipulates that the inspection of the flatness indexes of each structural layer of roads at all levels is carried out with the maximum gap h (mm) between the three-meter straightedge and the road surface, the standard deviation σ (mm) measured by the flatness meter, and the international roughness index IRI (unit: m / km) as the standards. JTG 3450-2019 "Code for Field Testing of Highway Subgrade and Pavement" fully explains the test methods for measuring flatness by the three-meter straightedge, the continuous flatness meter, the vehicle-mounted bump integrator, and the vehicle-mounted laser flatness meter. The three commonly used flatness measuring instruments are shown in Figures 1A - 1C .

[0004] The detection methods of pavement evenness are mainly divided into subjective evaluation method, profile-based method and response-based method. Among them, the profile-based method involves using high-precision sensors such as laser profilometers and 3D lidars to calculate the International Roughness Index (IRI) by measuring the elevation changes of the pavement profile, which is the most common method. The response-based method mainly relies on interdisciplinary research because various interaction phenomena occurring during vehicle driving among the road / vehicle / user jointly determine the pavement performance. In this process, complex methodologies are usually introduced to simulate the dynamic response of the vehicle as realistically as possible so as to evaluate its impact. However, whether it is the laser profile-based detection technology or the response-based detection method, they both rely on measuring the elevation changes on one or more measurement lines to determine the pavement evenness index. From a statistical perspective, the elevation changes of the pavement generally conform to a normal distribution with zero mean. However, due to the influence of factors such as vehicle rolling, cross and longitudinal slopes of the road, and weather conditions, the elevation distribution in the lateral direction of the road is not uniform. When different frequencies and different detection vehicles drive on the same road section, since it is difficult to make the detection trajectories of the laser probes completely consistent, this results in a low consistency of the IRI detection results, thus affecting the analysis and evaluation of the evenness of the wide-area road network.

[0005] Since the 1960s, European and American countries have started the research on 3D laser scanning technology and made remarkable progress in the 1990s. For example, the airborne 3D laser scanning technology is used to efficiently obtain large-area topographic information, realizing the industrialization of this technology. The research on pavement 3D laser scanning technology has also matured rapidly, such as the related research of Stanford University and the University of Washington in the United States, as well as the product development of well-known companies such as Leica and Riegl. These devices have been widely used in production, life and scientific research fields, such as ancient building restoration, topographic surveying and mapping, road measurement, etc. Figures 2A - 2B Shows the in-vehicle pavement 3D measurement device in the prior art.

[0006] Further, in terms of three-dimensional road surface measurement, Diaz et al. extracted the road surface measurement line at a distance of 1 m from the outer edge of the road based on the road alignment calculation results, and used the road slice segmentation method and the K-means clustering method to evaluate the road surface flatness by calculating the root mean square value, skewness, and kurtosis of the elevation of this measurement line. Kumar et al. proposed an algorithm that allows users to evaluate the flatness from the point cloud obtained through a three-dimensional measurement system. The purpose is to determine a method that can provide fast, inexpensive, and complete information, and determine the flatness by determining the standard deviation of the elevation value relative to the surface of the interpolation plane. Tawip et al. developed a segmentation algorithm based on point cloud voxelization that can estimate the flatness of the road surface. Debellasis et al. used ordinary Kriging method and inverse distance weighting method to process DSM data and extract the elevation change of the road surface to calculate the IRI value, and compared it with the IRI data measured by the left and right wheels of the laser profiler. The standard deviation was in the range of 1.07 - 1.83, proving that the results obtained by the two methods have a strong correlation. Barbara et al. calculated the road surface flatness of the airport runway using LiDAR three-dimensional point cloud data and DSM, and finely mapped the elevation and slope changes of different transverse measurement lines. Although this study considered the lateral influence of the flatness distribution, it did not analyze whether there are differences in the flatness indices obtained under different acquisition measurement lines from the perspective of flatness parameters.

[0007] Existing studies have established mature and reliable point cloud data processing methods, proving the reliability of laser point cloud measurement technology as a means for detecting road surface flatness. However, limited by the layout conditions of LiDAR and the mapping principle of point cloud data, it is difficult to directly read the elevation change at the position of the vehicle wheel track. In addition, existing studies often use the center line of the road wheel track or the elevation of the measurement line at a fixed width on one side to calculate the flatness value, without fully considering the influence of the lateral flatness distribution of the road surface. Summary of the Invention

[0008] In view of this, the present application proposes a full-width flatness calculation method and device, which can provide a more accurate and comprehensive evaluation of road flatness.

[0009] According to one aspect of the present application, the present application provides a full-width flatness calculation method, which includes:

[0010] Obtain the original elevation value point cloud data of the road surface;

[0011] Process the noise of the point cloud data;

[0012] Determine the lane category, calculate the wheel track distribution probability, and obtain the longitudinal section weight;

[0013] Calculate the equivalent longitudinal section elevation value of the lane based on the longitudinal section weight;

[0014] Calculate the flatness index based on the equivalent longitudinal section elevation value.

[0015] Preferably, processing the point cloud data noise includes:

[0016] Removing the point cloud data located outside the lane boundary;

[0017] Eliminating the abnormal data within the lane.

[0018] Preferably, determining the lane category includes determining the lane as a driving lane, overtaking lane, slow lane or mixed lane; the wheel track distribution probability is the lateral distribution probability of the wheel track;

[0019] The longitudinal section weight f norm (x) is calculated according to the following formula:

[0020]

[0021] where f max and f min represent the maximum and minimum values of f(x), respectively;

[0022]

[0023] where ω i represents the weight when μ i = 0.5L i , L i represents the actual wheelbase, x is the width of the lane along the lateral direction starting from the center of the lane, B i , μ i , and σ i are fitting parameters.

[0024] Preferably, when the lane category is a driving lane, L i is 1.8 m; when the lane category is an overtaking lane, L i is 1.6 m; when the lane category is a slow lane, L i is 2.0 m; when the lane category is a mixed lane, L i is 1.8 m.

[0025] Preferably, the equivalent longitudinal section elevation value P3(n) of the lane pavement is calculated according to the following formula:

[0026]

[0027] P2(x i ,n) = P1(x i ,n)·f norm (x i ), n = 1, 2,..., N;

[0028] P1(x i,n) is the point cloud data of each cross section, n represents the nth cross section along the lateral direction of the lane, and N represents the total number of cross sections covered by the point cloud.

[0029] Preferably, the method further comprises calculating the updated equivalent longitudinal section elevation value P4(n) of the lane according to the following formula:

[0030] P4(n)=a×P3(n) / M;

[0031] Among them, parameter M represents the number of sampling points in the cross section; parameter a is based on f norm The calculation is done by rounding off twice the ratio of the cumulative sum of (x) to M.

[0032] The present application provides a device for calculating the flatness of a whole road width, the device comprising:

[0033] An acquisition module is used to obtain the original elevation value point cloud data of the road surface;

[0034] The calculation module is used to process the point cloud data noise; determine the lane category, calculate the wheel track distribution probability and the longitudinal section weight; calculate the equivalent longitudinal section elevation value of the lane road surface based on the longitudinal section weight; and calculate the flatness index based on the equivalent longitudinal section elevation value.

[0035] Preferably, the computing module is further used to remove point cloud data outside the lane boundary; and to remove abnormal data within the lane.

[0036] Preferably, the calculation module is further used to determine whether the lane is a driving lane, a passing lane, a slow lane or a mixed lane; calculate the lateral distribution probability of the wheel track;

[0037] The longitudinal section weight f is calculated according to the following formula norm (x):

[0038]

[0039] Among them, f max and f min Respectively represent the maximum and minimum values ​​of f(x);

[0040]

[0041] Among them, ω i Represents μ i =0.5L i The weight of L i represents the actual wheelbase, x is the width of the lane in the lateral direction starting from the center of the lane, B i , μ i , and σ i is the fitting parameter.

[0042] Preferably, when the lane type is a driving lane, Li is 1.8 m; when the lane category is the overtaking lane, L i is 1.6 m; when the lane category is the slow lane, L i is 2.0 m; when the lane category is the mixed lane, L i is 1.8 m.

[0043] Preferably, the calculation module is further configured to calculate the pavement equivalent longitudinal section elevation value P3(n) of the lane according to the following formula:

[0044]

[0045] P2(x i ,n) = P1(x i ,n)·f norm (x i ), n = 1, 2,..., N;

[0046] P1(x i ,n) is the point cloud data of each cross-section, n represents the nth cross-section along the transverse direction of the lane, and N represents the total number of cross-sections covered by the point cloud.

[0047] Preferably, the calculation module is further configured to calculate the updated pavement equivalent longitudinal section elevation value P4(n) of the lane according to the following formula:

[0048] P4(n) = a×P3(n) / M;

[0049] wherein, the parameter M represents the number of sampling points of the cross-section; the parameter a is calculated by taking the integer of twice the ratio of the cumulative sum of f norm (x) to M.

[0050] Compared with the traditional technical solution that only focuses on the wheel path, the present application realizes a comprehensive consideration of the use performance of the entire lane, establishes a new full-width road surface flatness calculation model, and can provide a more accurate and comprehensive road surface flatness evaluation. By introducing the statistical principle, a weight model of the road longitudinal section is innovatively established. This model can scientifically and reasonably distribute the weights of each longitudinal section within the full width, generate an equivalent longitudinal section that can represent the full-width road surface flatness through the weight model, and improve the accuracy and comprehensiveness of the flatness evaluation.

[0051] Other features and advantages of the present application will be described in detail in the subsequent specific implementation part. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof are used to explain this application. In the drawings:

[0053] Figures 1A - 1Cis a pavement evenness measuring instrument in the prior art; among which Figure 1A is a three-meter straightedge, Figure 1B is a continuous eight-wheel pavement evenness instrument, Figure 1C is a vehicle-mounted laser evenness instrument;

[0054] Figures 2A - 2B is a vehicle-mounted pavement three-dimensional measurement device in the prior art, among which Figure 2A is a schematic diagram of the device structure and measurement principle, Figure 2B is a pseudo-color map of pavement three-dimensional point cloud data;

[0055] Figure 3 is a flow chart of the full-width evenness calculation method provided by this application;

[0056] Figure 4 is the fitting result of the wheel track center line in BS 5400-10 provided by this application;

[0057] Figure 5 is the wheel track lateral distribution probability curve provided by this application;

[0058] Figure 6 are the vehicle types in each lane under the influence of channelization in the prior art;

[0059] Figure 7 is the ratio of the sum of the weights f norm (x) to the number of sampling points N provided by this application;

[0060] Figure 8 is the mean comparison of IRI calculated based on two elevation values provided by this application;

[0061] Figure 9 is the coefficient of variation comparison of repeatability measurements based on two elevation values provided by this application. Detailed implementation manners

[0062] It should be noted that, without conflict, the implementation manners in this application and the features in each implementation manner can be combined with each other.

[0063] Next, this application will be described in detail with reference to the drawings and in combination with the implementation manners.

[0064] Currently, evenness measurement devices are mainly divided into two-dimensional and three-dimensional types. Two-dimensional detection devices measure elevation values by selecting a longitudinal section of the wheel track belt, but this method has randomness, resulting in poor repeatability of the measured elevation values and the calculated IRI, and may misevaluate the road performance. Three-dimensional detection devices were not initially developed specifically for pavement evenness, but for detecting local damages such as potholes and bumps on the road surface. Although some researchers have proposed using three-dimensional point cloud data to analyze evenness, they still mainly focus on the elevation values of a certain longitudinal section in the wheel track belt.

[0065] For channelized roads, the vehicle driving trajectories are regular, and the wheel track distribution shows an "M" - shaped characteristic, which proves the rationality of calculating the pavement evenness using a certain longitudinal section elevation value within the wheel track belt in the early stage. However, the evenness of a single wheel track belt cannot comprehensively measure the lane usage performance, nor can it represent the impact of the lane on the ride smoothness of the driving vehicles.

[0066] In the early stage, it was challenging to comprehensively measure the elevation values of the entire road - width lane using two - dimensional measurement equipment. However, with the popularization of three - dimensional measurement equipment, it has become possible to comprehensively consider the evenness of the entire lane. After obtaining the point - cloud data of the elevation values of the entire road - width, how to reasonably allocate the weights of each longitudinal section of the lane to reasonably evaluate the lane evenness is an urgent problem to be solved currently.

[0067] This application aims to propose a longitudinal section weight model and the concept of an equivalent longitudinal section of a road based on the three - dimensional point - cloud data of the road surface, and establish a calculation model for the elevation value of the equivalent longitudinal section. Through this model, we will replace the early practice that only relies on the longitudinal section of the wheel track belt and its elevation value, and then construct a calculation model for the evenness of the entire road - width of the lane. This model will be able to more comprehensively and scientifically consider the usage performance of the entire lane, providing a more accurate and comprehensive basis for performance parameters for traffic management departments.

[0068] The flow chart of the calculation method for the evenness of the entire road - width proposed in this application is shown in Figure 3 , and the main operation steps are as follows:

[0069] Step 301: Obtain the original elevation - value point - cloud data of the road surface;

[0070] Step 302: Point - cloud data noise processing, including road boundary recognition and outlier removal within the lane, etc.;

[0071] Step 303: Determine the lane category, calculate the wheel track distribution probability and the longitudinal section weight;

[0072] Step 304: Calculate the equivalent longitudinal section elevation value of the road surface of this lane;

[0073] Step 305: Calculate evenness indexes such as IRI based on the equivalent longitudinal section elevation value.

[0074] First, through three - dimensional laser scanning or other data - collection devices, obtain the original elevation - value point - cloud data of the road surface. Subsequently, identify and determine the boundaries of the road to limit the scope of subsequent evenness analysis. Then, in order to limit the scope of subsequent evenness analysis, it is necessary to identify and determine the boundaries of the road. Subsequently, pre - process these original data, and this step includes identifying and removing the outliers within the lane, which may be caused by equipment errors or foreign objects on the road surface. After cleaning, retain the valid point - cloud data within the lane as a solid foundation for subsequent evenness analysis.

[0075] Further, in order to more accurately analyze the pavement evenness, weights are assigned to the point cloud data of each longitudinal section, and the equivalent elevation values of each longitudinal section are calculated using a weight model. Subsequently, the equivalent elevation values of each cross section are accumulated. On this basis, the maximum and minimum values are found from the original valid point cloud elevation values, and then, taking these two values as the targets, the equivalent elevation values are normalized to reduce the influence of elevation value distortion caused by the aforementioned accumulation.

[0076] Finally, according to the normalized elevation values, the International Roughness Index (IRI) is obtained using a general calculation method, and this index is used as the core index for evaluating pavement evenness and the detection equipment.

[0077] The following analyzes and explains the main steps involved in the above calculation process.

[0078] In step 301, the present application uses the prior art to obtain the original pavement elevation point cloud data, and a three-dimensional point cloud acquisition device is used to collect the pavement elevation information of the test road. When measuring the lane elevation point cloud data, the boundary of the point cloud can cover the entire lane, which is convenient for extracting the lane boundary line from the point cloud data subsequently.

[0079] In step 302, a large amount of useless interference noise is inevitably mixed in the collected original data. These data mainly come from the environment outside the road lane and occasional abnormal points or singular points inside the lane, or are generated by the equipment itself. These noise data may affect the accuracy of subsequent analysis. Therefore, effective data cleaning must be carried out.

[0080] First of all, identifying the lane boundary is a key step in the noise processing of point cloud data. By analyzing the spatial distribution of the point cloud data and the pavement characteristics, the ranges of the lane and non-lane areas are clearly distinguished. This step provides the necessary spatial definition for data cleaning and subsequent data processing. Then, all the point cloud data located outside the lane boundary is removed.

[0081] Further, the abnormal data inside the lane is removed. Considering the continuous characteristic of the point cloud data, it can be used to carefully delete the occasional singular points inside the lane. Such a processing step is crucial for improving the overall quality of the data because it can eliminate the abnormal values that may interfere with data analysis.

[0082] Finally, after screening and cleaning, the retained point cloud data will be limited to the valid information inside the lane. These data will provide an accurate and reliable basis for pavement elevation analysis, evenness evaluation, and road condition monitoring.

[0083] This application does not limit the specific forms and methods for processing noise data of point cloud data. The two core key points are: 1: High-precision identification of the left and right boundaries (edges) of the lane and elimination of the point cloud data outside the lane; 2: Elimination or repair of outliers from the point cloud data within the driving lane based on characteristics such as the continuity of elevation values.

[0084] In step 303, during the actual driving process of the vehicle, the distribution of wheel tracks on the lane cross-section shows a certain regularity. The lateral distribution coefficient of wheel tracks is an index describing the distribution law of vehicle wheel tracks on the road cross-section. It refers to the ratio of the number of action times received per unit width of the road surface to the total number of wheel loads within the lane width range. Factors affecting this coefficient include traffic organization and management, traffic volume, traffic composition, lane width, vehicle speed, and driver's driving habits, etc. These factors result in significant differences in the lateral distribution coefficient of wheel tracks under different highway grades and lane types.

[0085] The lateral distribution probability curve of wheel tracks is a key tool for characterizing the contact frequency of the left and right wheels of a vehicle with each longitudinal section of the ground. This curve reveals the probability of different road surface positions being run over by the wheels during vehicle driving. Specifically, the level of the wheel track distribution probability value is directly related to the compaction degree of the wheel on a specific longitudinal section of the road surface, and conversely, it also shows the excitation effect of a specific longitudinal section of the road on the vehicle. A high probability value indicates that the wheels frequently pass through this longitudinal section, increasing the possibility of forming a wheel track belt; while a low probability value means that the compaction effect of the wheels on this longitudinal section is small. In areas with a small wheel track distribution probability value, not only is the chance of wheel passage reduced, but correspondingly, the reaction force generated by the wheels on the road surface will also decrease. Based on the correlation between the lateral distribution probability curve of wheel tracks and road surface load as well as the reaction on the vehicle, this application selects this probability model as the theoretical basis model for lane longitudinal section weight distribution. This selection is based on an in-depth understanding of vehicle driving behavior and road surface response characteristics. Using the lateral distribution probability model of wheel tracks as the longitudinal section weight distribution model not only reasonably considers the actual effect of the vehicle on the road surface, but also scientifically quantifies the importance of different longitudinal sections, providing accurate weights for the evaluation of road surface smoothness.

[0086] Currently, the European code Eurocode1 and the British code BS 5400-10 involve the distribution of vehicle driving trajectories in the publicly available standard specifications. Among them, the lateral distribution frequency model of the vehicle center line given by the British code BS 5400-10 based on observational statistical data follows a normal distribution (the lateral wheel track loading frequency is given according to a strip width of 0.1 m). The lateral distribution width of the wheel track center lines of both within the lane surface is 1.2 m, as Figure 4 shown. Considering the wide application and applicability of this code, this application designates it as the standard reference. Fitting the statistical probability data corresponding to these 13 strips in the figure according to the normal distribution, the results are shown inFigure 4 The normal distribution model of the probability distribution of the center line of the wheel tracks is shown in formula (1).

[0087]

[0088] Among them, the fitting parameters are B = 0.098674, μ = 0, and σ = 0.226058.

[0089] According to R 2 = 0.9957, it can be seen that the fitting effect is good, indicating that formula (1) has high model generality, and it is scientific to use it for subsequent derivation of the lateral distribution probability of the left and right wheel tracks.

[0090] Considering that within each independent lane, there is a direct relationship between the left and right wheel tracks and the wheelbase of the vehicle. However, the wheelbase of the vehicle is not a unified standard, and its common range covers 1.5 m to 2.2 m. Without loss of generality, this application takes 1.6 m, 1.8 m, and 2.0 m as three typical wheelbase intervals to analyze the probability distribution of the left and right wheel tracks along the lane width direction under the above-mentioned center line distribution frequency of the wheel tracks. The results are shown in Figure 5 .

[0091] In Figure 5 , under the three wheelbase widths, the lateral distribution probabilities of the left and right wheel tracks both show a typical "M" shape. The two-dimensional Gaussian mixture model shown in formula (2) is used to fit it. The parameters to be fitted are B1, B2, μ1, μ2, σ1, and σ2 respectively. Among them, B1 and B2 are the weights of the two Gaussian models f(x) ~ N1(μ1, σ1 2 ) and f(x) ~ N2(μ2, σ2 2 ). Considering the symmetry of the curve in Figure 5 , it can be considered that B1 = B2, μ1 = μ2, and σ1 = σ2. At this time, formula (2) is further simplified to formula (3). Finally, the fitting results are shown in Table 1.

[0092]

[0093] Table 1 Lateral distribution frequencies of the left and right wheel tracks at different wheelbases

[0094] L(m) <![CDATA[B1]]> <![CDATA[μ1]]> <![CDATA[σ1]]> <![CDATA[R 2 > 1.6 0.098669 0.800 0.226 0.9999 1.8 0.098669 0.900 0.226 0.9999 2.0 0.098671 1.000 0.226 0.9999

[0095] As can be seen from Table 1, the fitting parameters B1 and σ1 remain basically unchanged under three different wheelbase conditions, while the parameter μ1 varies with the change of the wheelbase L, and there is a significant 2-fold relationship between the two. Such an observation result indicates that for vehicles with the same wheelbase, the lateral probability distributions of their left and right wheel tracks show consistency. However, it must be pointed out that the above three wheelbases do not represent the wheelbase situations of all vehicles traveling on a specific lane. On an actual road, the formation of the wheel track belt is produced by the joint rolling of all passing vehicles within the traffic lane. This means that the formation process of the wheel track belt involves vehicles with more types of wheelbases. However, due to the non-standardized characteristics of the wheelbase, the specific position and width of the wheel track belt also vary. In this case, the model of the lateral distribution frequency of the wheel track can be summarized as formula (4).

[0096]

[0097] In the formula, ω i represents the weight when μ i = 0.5L i and L i represents the actual wheelbase.

[0098] In addition, the lane channelization effect has a significant impact on the vehicles traveling in each lane, making the vehicle types within each lane relatively universal. Taking the common three-lane on one side as an example, the leftmost lane is usually the overtaking lane, mainly used for high-speed overtaking; the middle lane is the driving lane, which is the main passage for most vehicles to travel; while the rightmost lane is the slow lane, often used for slow-speed driving of heavy trucks, as shown in Figure 6 . This lane channelization effect not only affects the driving behavior of vehicles, but also has an important impact on the formation and characteristics of the wheel track belt.

[0099] Specifically, in the overtaking lane, since it is mainly dominated by small cars with a relatively small wheelbase, the lateral distribution probability curve shapes of the left and right wheel track belts will tend to be the case when the wheelbase L = 1.6m. In the driving lane, the vehicle types are relatively more diverse, but mainly dominated by trucks with a medium wheelbase, which makes the lateral probability distribution curves of the left and right wheel track belts closer to the case when the wheelbase L = 1.8m. In the slow lane, since it is mainly dominated by trucks with a wider wheelbase, the lateral distribution probability curve of the wheel track belt will tend to be the case when the wheelbase L = 2.0m. The difference in vehicle type distribution caused by this lane channelization effect further affects the formation and characteristics of the wheel track belt.

[0100] In a mixed traffic environment, especially in urban internal traffic flows, the vehicle types are more diverse compared to highways, which leads to a higher complexity in the lateral distribution of vehicle wheel tracks. The lateral distribution probability of vehicle wheel tracks is not only related to vehicle types but also affected by factors such as the number of vehicle passages, traffic density, and vehicle behavior patterns. Since the wheelbases of different types of vehicles vary, it becomes challenging to establish a universal weight model to accurately quantify the impact of vehicles with different wheelbases on road surface evenness. Such a model needs to consider the traffic characteristics under specific road conditions, and these characteristics vary from road to road, making it difficult to achieve unified quantification. Despite the above complexities, through on-site observation and analysis of actual roads, visual features such as color and texture can be used to effectively distinguish wheel track bands and non-wheel track bands. Such a physical fact indicates that the wheel track distribution probability model for each lane in a mixed traffic environment also conforms to formula (3).

[0101] In view of this, when using point cloud measurement technology to collect the road surface elevation values of different lanes, the parameter μ i (or L i ) is solidified, and the specific solidified values can be seen in Table 2. By solidifying the parameters, it can be ensured that different devices and manufacturers can have a unified calculation benchmark when processing these point cloud data. In this way, not only can the comparability of the data be effectively improved, but also a more solid and reliable foundation can be provided for subsequent data analysis and decision-making.

[0102] Table 2 Recommended μ i values and L values

[0103] Overtaking lane Traffic lane Slow lane Mixed lane <![CDATA[L i (m)]]> 1.6 1.8 2.0 1.8 <![CDATA[μ i (m)]]> 0.8 0.9 1.0 0.9

[0104] In addition, in order to introduce the weight model f(x) into the subsequent analysis and calculation of cross-section elevation values, f(x) needs to be transformed again. The core objective of this transformation is to strengthen the representation of elevation values on the wheel track band while weakening the influence of elevation values in the middle and on both sides of the lane. To achieve this goal, f(x) should be adjusted to a function with a maximum value of 1 and a minimum value of 0.

[0105] Specifically, the transformation of f(x) should follow the following principle: During vehicle driving, the wheel track band is the key area affecting vehicle dynamic response, so the elevation values in this area need to be clearly represented and retained; while the areas in the middle and on both sides of the lane have less impact on vehicle driving, and the corresponding elevation values can be suppressed by mathematical methods. The general idea of this transformation is to normalize f(x) so that its value range is mapped to the interval [0,1], where 1 represents the maximum weight of the wheel track band and 0 represents the minimum weight in the middle or on both sides of the lane. The normalization process is shown in formula (5).

[0106]

[0107] In the formula, f max and f min respectively represent the maximum and minimum values of f(x).

[0108] Through the above normalization process, f(x) will be effectively converted into a weight function suitable for the analysis of cross-sectional elevation values.

[0109] In step 304, after successfully obtaining the effective point cloud data in the lane and establishing the longitudinal section weight model of the lane, the equivalent elevation value of the lane data can be calculated through the following steps:

[0110] (1) According to the specific type of the detected lane, determine the key parameters L and μ of the longitudinal section weight model, and generate the corresponding data sequence f norm (x i ), where the range of x i is [0, L].

[0111] (2) Along the driving direction, sequentially perform point-by-point multiplication operations on the point cloud data P1(x i , n) of each cross-section and the longitudinal section weight model sequence f norm (x i ) to generate the elevation value P2(x i , n) distributed by weight for each longitudinal section, as shown in formula (6). In the formula, n represents the nth cross-section along the driving direction, and N represents the total number of cross-sections covered by the point cloud.

[0112] (3) Accumulate and sum the cross-section data after the multiplication operation to obtain the comprehensive elevation value P3(n) of each cross-section. The equivalent longitudinal section elevation value P3(n) synthesizes the elevation value P1(x i , n) of each point in the corresponding cross-section and its weight f norm (x i ).

[0113] P2(x i , n) = P1(x i , n) · f norm (x i ), n = 1, 2,..., N (6)

[0114]

[0115] Theoretically, once the equivalent longitudinal section elevation value P3(n) is obtained, indicators such as IRI can be derived according to the general standard flatness calculation method. However, the calculation result may not fully reflect the actual use situation of the road surface, so it may lack sufficient reference value and guiding significance in practical applications. According to the multiplication process described in formula (6), the longitudinal section weight model fnorm (x) has an impact on the elevation point cloud data. Although this process introduces the distribution of weights, it may change the amplitude of the elevation values. In addition, according to the superposition process described by formula (7), a large number of elevation values are comprehensively considered to calculate the equivalent elevation value. Although this process can provide an aggregated representation of the data, it may also over-intensify the aggregation effect of the data, thereby weakening some important detailed information.

[0116] Further analysis reveals that when sampling elevation data for the same cross-section at different sampling rates, the difference in the number of sampling points will cause a significant change in the elevation value P3(n). This fluctuation in the equivalent longitudinal section elevation value caused by the change in sampling density may not accurately reflect the true flatness state of the road surface, and it also contradicts the original intention of this application to accurately evaluate the road surface flatness and provide guidance for road maintenance. Based on this, this application proposes a strategy for scaling the equivalent elevation value P3(n), aiming to map it to a scientifically reasonable numerical range.

[0117] Figure 7 shows the ratio relationship between the cumulative sum of the weight model f norm (x) and the number of sampling points M in the cross-section. Among them, the number of sampling points M is indirectly characterized by the sampling interval Δx. It can be seen from the figure that regardless of how the sampling interval changes, the ratio of the cumulative sum of f norm (x) to M tends to a stable value, that is, 0.3226. In addition, the standard deviation of this ratio is 0.00455, and the coefficient of variation is only 1.41%, indicating that the results under different sampling intervals have high consistency. Based on this significant and stable phenomenon, this application further proposes a scaling method for the equivalent longitudinal section elevation value based on observational data. Specifically, formula (8) is used to scale the equivalent longitudinal section elevation value to ensure that the weight model can maintain the effectiveness and reliability of its analysis under different sampling intervals. This scaling process not only helps to eliminate the influence of the change in sampling interval on the model results, but also improves the generalization ability of the model under different conditions. Finally, the equivalent longitudinal section elevation value of a certain lane is represented by P4(n).

[0118] P4(n) = a × P3(n) / M (8)

[0119] In the formula, the parameter M represents the number of sampling points in the cross-section; the parameter a = 6, which is obtained by rounding up twice the reciprocal of the ratio 0.3226. This calculation method ensures that the scaling process can match the stable ratio observed theoretically, and through the rounding-up process, the operability of the parameter in practical applications is enhanced.

[0120] In step 305, after obtaining the equivalent longitudinal section elevation value of the road surface, a flatness index can be derived according to the publicly available flatness calculation method, such as the International Roughness Index IRI, etc. It should be noted that the present invention does not make any innovation to the method of calculating flatness based on the elevation value, but directly adopts the current prevailing flatness calculation standards and methods.

[0121] In view of the deficiency in the prior art that only the longitudinal section of a single wheel path is used for measurement and evaluation, the present application proposes a new concept of full-width road flatness. A corresponding calculation model and method are developed, which can comprehensively and synthetically evaluate the overall flatness of the measured road, overcoming the limitations of the prior art. The principle of statistics is introduced, and based on the existing specification documents, a weight theoretical model of the road longitudinal section is established. This model can reasonably distribute the weights of each longitudinal section within the full width of the road, ensuring the accuracy and comprehensiveness of the flatness evaluation. By extracting the lane boundary lines in the preprocessing stage, the point cloud range of the entire lane is restricted, enhancing the position accuracy of each longitudinal section of the road. Based on the aforementioned weight model, an equivalent longitudinal section representing the full-width road flatness is generated. The elevation value of the equivalent longitudinal section of the measured lane is calculated as the key index for evaluating the road flatness, replacing the practice in the prior art that only randomly considers a certain longitudinal section within the wheel path.

[0122] A vehicle-mounted three-dimensional point cloud measurement device and a vehicle-mounted laser profiler are used to measure the standard test road of the Ministry of Transport Highway Traffic Test Field in Tongzhou District, Beijing. When collecting the road surface elevation, both vehicles drive along the center line of the lane. The laser profiler selects the data at the longitudinal section V1 of the left wheel path for analysis and marks it as IRI0; after processing the data by the vehicle-mounted three-dimensional point cloud measurement device, it selects two longitudinal sections, namely the longitudinal section V2 of the left wheel path and the equivalent longitudinal section of the lane, for calculation and analysis, which are respectively denoted as IRI V and IRI F . The method of multiple repeated measurements is adopted to ensure the accuracy of the elevation data and the reliability of the proposed flatness method. The test road is measured repeatedly 9 times, and the mean values of the elevation values are used to calculate the corresponding IRI0, IRI V and IRI F .

[0123] Figure 8 The displayed ones are the flatness indexes IRI0, IRI V and IRI F calculated according to the elevation values of the longitudinal section V1, the longitudinal section V2, and the equivalent longitudinal section at every 100 m length.Results. First, compare the measurement results of the two types of devices on the longitudinal profiles V1 and V2 of the left wheel track. The two curves show obvious consistency, but there are local differences, indicating that the elevation values measured by the two measurement methods are indeed different. Due to the lack of standard values, it is impossible to determine which measurement result is the best. However, from the data acquisition sources, the IRI calculated based on the longitudinal profile V2 V is more in line with expectations.

[0124] In addition, from the analysis of the data trend, the pavement evenness indices obtained using two different elevation values in the point cloud data show good consistency, with a correlation coefficient reaching 0.996. However, it is worth noting that the IRI F values are generally higher than those of IRI V . The reason for this phenomenon is that the composition of the equivalent longitudinal profile elevation value is more complex and comprehensive. It not only includes the elevation data of the V2 longitudinal profile but also integrates the elevation information of multiple other longitudinal profiles within the lane. Therefore, compared with relying solely on the elevation data of a single longitudinal profile, the equivalent elevation value undoubtedly provides richer and more comprehensive pavement elevation information. This information gain is reflected in the calculation of the evenness index, resulting in a relatively larger value of the evenness index based on the equivalent elevation value. This phenomenon not only demonstrates the superiority of the equivalent elevation value in reflecting the pavement condition but also provides a more accurate and comprehensive method for pavement evenness evaluation.

[0125] Subsequently, the coefficient of variation Cv is used to analyze the repeatability of the calculation results, and the results are shown in Figure 9 . In the case of the two longitudinal profiles based on the three-dimensional point cloud device, the coefficient of variation Cv is relatively small, less than 3.0% and 4.0% respectively. However, the output results of measuring the V1 longitudinal profile using the vehicle-mounted laser profiler show relatively large variability. This result indicates that using point cloud data for pavement evenness measurement and analysis is a feasible method. In particular, even when only using the elevation values of the V2 longitudinal profile, stable IRI calculation results can be obtained. This stability benefits from the application of three-dimensional point cloud technology. When measuring, the positions of each longitudinal profile within the lane are defined with the road edge as the boundary, thus improving the measurement accuracy and consistency. Compared with the traditional vehicle-mounted laser profiler, the latter randomly collects elevation data within the wheel path range, while the position of the V2 longitudinal profile within the lane is pre-determined, which increases the predictability and accuracy of the measurement.

[0126] In addition, when calculating IRI using the elevation values of the equivalent longitudinal profile, the variability of the data measurement results exactly reflects the performance stability and reliability of the measuring instrument during repeated measurements of the same road section. Since the elevation values of the equivalent longitudinal profile integrate the data of all relevant longitudinal profiles and provide a comprehensive description of the road surface morphology, the stability of the measurement results is affected by the comprehensive data.

[0127] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A full-width flatness calculation method, characterized in that The method includes: Obtaining the original elevation value point cloud data of the road surface; Processing the noise of the point cloud data; Determining the lane category, calculating the wheel track distribution probability, and obtaining the longitudinal section weight; Calculating the equivalent longitudinal section elevation value of the lane based on the longitudinal section weight; Calculating the smoothness index based on the equivalent longitudinal section elevation value; Among them, determining the lane category includes determining the lane as a driving lane, overtaking lane, slow lane or mixed lane; the wheel track distribution probability is the lateral wheel track distribution probability; Longitudinal section weight f norm (x) is calculated according to the following formula: where, f max and f min represent the maximum and minimum values of f(x), respectively; Among them, ω i represents the weight when μ i = 0.5L i , where L i represents the actual track width, x is the width of the lane in the lateral direction starting from the center of the lane, and B i , μ i , and σ i are fitting parameters.

2. The full-width flatness calculation method according to claim 1, wherein Processing the noise of the point cloud data includes: Removing the point cloud data located outside the lane boundary; Eliminating the abnormal data within the lane.

3. The full-width flatness calculation method according to claim 1, wherein When the lane category is the driving lane, L i is 1.8 m; When the lane type is the overtaking lane, L i is 1.6 m; When the lane category is the slow lane, L i is 2.0 m; When the lane type is a mixed lane, L i is 1.8 m.

4. The full-width flatness calculation method according to claim 1, characterized in that, Calculating the equivalent longitudinal section elevation value P3(n) of the lane according to the following formula: P2(x i , n) = P1(x i , n) · f norm (x i ), n = 1, 2,..., N; P1(x i , n) is the point cloud data of each cross-section, where n represents the nth cross-section along the lateral direction of the lane, and N represents the total number of cross-sections covered by the point cloud.

5. The full-width flatness calculation method according to claim 4, characterized in that, The method further includes calculating the updated equivalent longitudinal section elevation value P4(n) of the lane according to the following formula: P4(n) = a × P3(n) / M; Among them, the parameter M represents the number of sampling points of the cross-section; the parameter a is calculated by rounding up twice the ratio of the cumulative sum of f norm (x) to M.

6. A full-width flatness calculation device, characterized in that The device includes: An obtaining module, configured to obtain the original elevation value point cloud data of the road surface; A calculating module, configured to process the noise of the point cloud data; determine the lane category, calculate the wheel track distribution probability and the longitudinal section weight; calculate the equivalent longitudinal section elevation value of the lane based on the longitudinal section weight; calculate the smoothness index based on the equivalent longitudinal section elevation value; Among them, the calculating module is further configured to determine the lane as a driving lane, overtaking lane, slow lane or mixed lane; calculate the lateral wheel track distribution probability; Calculate the longitudinal section weight f according to the following formula norm (x): Among them, f max and f min represent the maximum and minimum values of f(x), respectively; where ω i represents the weight when μ i = 0.5L i and L i represents the actual track width, x is the width of the lane along the lateral direction starting from the center of the lane, and B i , μ i , and σ i are fitting parameters.

7. The full-width flatness calculation device according to claim 6, wherein The calculating module is further configured to remove the point cloud data located outside the lane boundary; eliminate the abnormal data within the lane.

8. The full-width flatness calculation device according to claim 6, characterized in that, When the lane type is a traffic lane, L i is 1.8 m; When the lane category is the overtaking lane, L i is 1.6 m; When the lane type is the slow lane, L i is 2.0 m; When the lane type is a mixed lane, L i is 1.8 m.

9. The full-width flatness calculation device according to claim 6, characterized in that The calculating module is further configured to calculate the equivalent longitudinal section elevation value P3(n) of the lane according to the following formula: P2(x i ,n) = P1(x i ,n)·f norm (x i ), n = 1, 2,..., N; P1(x i , n) is the point cloud data of each cross-section. n represents the nth cross-section along the transverse direction of the lane, and N represents the total number of cross-sections covered by the point cloud.

10. The full-width flatness calculation device according to claim 9, characterized in that, The calculating module is further configured to calculate the updated equivalent longitudinal section elevation value P4(n) of the lane according to the following formula: P4(n) = a × P3(n) / M; Among them, the parameter M represents the number of sampling points of the cross-section; the parameter a is calculated by taking the integer of twice the ratio of the cumulative sum of f norm (x) to M.

Images