Assessment of asphalt pavement aging condition based on GF-2 high-resolution remote sensing image

Abstract. For the evaluation of traffic infrastructure, asphalt pavement aging conditions are crucial. Due to the complexity of identifying and monitoring asphalt pavement aging conditions, many current studies tend to use satellite remote sensing methods. We conducted an extraction experiment on the aging status of asphalt pavement using on-site measured Pavement Surface Condition Index data and Gaofen-2 satellite (GF-2) high-resolution remote sensing images based on comprehensive references to previous research results. Based on our experimental results, the difference health index, ratio health index, and normalized difference health index can reflect asphalt pavement aging to varying degrees, but the correlation is relatively weak. The purpose of this paper is to propose a new asphalt pavement aging index (PAI), namely the PAI, based on sufficient experimental analysis. In addition to identifying asphalt pavement aging perfectly, PAI has a good ability to discriminate between road interference information, such as shadows and vehicles, after it has been verified. There is a significant linear relationship between its correlation coefficient R and pavement surface condition index, which is 0.894. The evaluation results of three sets of ground verification points obtained by applying PAI also demonstrate its practicality. Therefore, the combination of PAI and GF-2 high-resolution remote sensing images can be used to evaluate the aging status of asphalt pavements.


Introduction
In the process of operating highway pavement, the strength of the pavement structure gradually decreases as a result of temperature, moisture, weathering, and load factors.As a result, the road surface eventually ages and develops various diseases (such as cracks, potholes, ruts, etc.). 1 In addition to reducing the performance of the road surface, this results in huge economic losses and increased accident risks. 2 Over the last few years, the number of highways in China has grown rapidly, and the number of roads that must be maintained has also increased steadily (Fig. 1).The total length of China's roads is expected to reach 5.2807 million km by 2021, an increase of 82,600 km from the previous year.99.4% of highway mileage was covered by highway maintenance, covering 5.2516 million km. 3 There will be considerable inconvenience to people's daily lives and to the transportation industry when the road surface has problems and cannot be maintained in a timely manner.In addition, it will place substantial pressure on the transportation department in terms of monitoring and maintaining the road. 4As a result, it is essential to investigate and detect road conditions rapidly in order to ensure the safety and stability of road traffic. 2 Although different road materials, different road diseases, and different road aging degrees will present considerable management challenges for road supervision, management, maintenance, and renovation. 4,5The need for comprehensive, large-scale monitoring, and management of the health of roads has therefore become increasingly urgent.
During the early stages of highway development, road maintenance personnel conduct manual measurements of road surface conditions. 2 Manual detection, however, is less efficient and destructive.1][12] The use of these technologies has disadvantages, such as the obstruction of traffic and the delay in obtaining information regarding the condition of the entire section, as well as the time required to conduct the sampling survey.
9][20][21] They can also extract and calculate parameters, such as length, width, and area of the defects, 22,23 with recognition accuracy reaching centimeter 24 or millimeter 25 levels.However, the pavement research using drone images is limited to a small area.Although it has a significant accuracy advantage, it is still insufficient in terms of research scope and rapid monitoring, which needs to be effectively supplemented by other means.Due to its characteristics of macroscopic nature, present situation, and objectivity, [26][27][28][29] satellite remote sensing technology has shown great application potential in the detection of road surface conditions due to the rapid development of satellite earth observation technologies, particularly high-resolution remote sensing satellites.
At home and abroad, successive studies have been conducted on the spectral characteristics of asphalt pavement and the status of road health.According to Herold et al., 30,31 there is a variable correlation between remote sensing image information and road conditions.By comparing the relationship between index and pavement surface condition index (PCI) established by hyperspectral remote sensing, they verified the feasibility of monitoring road conditions using hyperspectral remote sensing.According to Levinson,32 the spectral curve of aging asphalt roads shows significant slope changes in the wavelength range of 2100 to 2200 nm, due to the absorption of silicate compounds, and in the wavelength range of 2250 to 2300 nm, due to the absorption of hydrocarbons.In analyzing the conversion of airborne visible infrared imaging spectrometer (AVIRIS) and Hyper Spec TIR (HST) imaging spectral data to PCI based on the different spectral characteristics of different road disaster units, Liu et al. 33 explored the possibility of converting the absolute reflectance of these units to PCI as well.The correlation between pavement reflectance values and texture features of various health conditions was studied by Fig. 1 Changes of national highways in China in the past 5 years. 3mery et al. 34 The feasibility of evaluating pavement condition using high-resolution satellite images was verified.The scatter plot was drawn by Mei et al. 35,36 based on the reflectivity values of different roads at 460 and 740 nm.Interestingly, these points were on the same asphalt line, and the same level of asphalt road showed high polymerization, demonstrating the feasibility of multispectral and hyperspectral imaging for monitoring road conditions.Mei and Manzo 37 combined with road field spectral measurements, conducted a preliminary study on the application of hyperspectral remote sensing based on airborne hyperspectral images acquired by MIVIS and AVIRIS and concluded that the road spectral features change with changes in road conditions.Based on the difference in asphalt pavement with different aging degrees over the wavelength range of 400 to 900 nm, Jin et al. 38 proposed the difference health index (DHI), ratio health index (RHI), normalized DHI, and logarithmic health index.The research progress in remote sensing monitoring of highway pavement quality has been summarized by Pan. 2 He noted that pavement condition monitoring technology based on airborne and airborne sensors has limitations, such as low precision and poor robustness.Cheng's 4 aging assessment results using the multiplication aging index and WorldView-2 remote sensing images are in significant agreement with the results of a large-scale aging assessment conducted on the road.Xiao et al. 39 constructed spectral index and gray co-occurrence matrix based on GF-2 remote sensing image data and used SVM classification method to realize the identification of road surface material types.Chen et al. 40 designed a new method based on an improved recurrent neural network for identifying pavement with similar spectral features from Worldview-2 satellite images and realized automatic monitoring of asphalt pavement aging phenomenon.
][43][44] Although the study of asphalt pavement spectral response mechanism is still in its infancy, the use of spatial remote sensing for road condition monitoring has not yet developed mature methods and effective quantitative evaluation models. 4,45,46rom the experience of testing and maintaining aged asphalt roads, it is apparent that there is a clear correlation between the aging and damage of asphalt roads and changes in the chemical composition of the road surface, as well as the level of exposure to rock aggregates.Accordingly, this study combines previous understandings of the relationship between asphalt pavement PCI, spectral reflectance changes, and road aging indexes. 30,31,33,38An experimental study was conducted to determine the aging status of some asphalt roads in Shenyang by analyzing the spectral characteristics of asphalt pavement with different degrees of aging.As a result of verifying others' road aging indexes, this study developed an innovative asphalt pavement aging monitoring index suitable for local conditions and used GF-2 remote sensing images to validate and evaluate the proposed index.The experimental results demonstrated the applicability of highresolution remote sensing data in assessing the aging condition of asphalt pavement, providing a technological means for the transportation department to achieve large-scale macroscopic monitoring of road conditions. 47

Study Area
The research area is selected near the eastern section of Hunnan Middle Road in Hunnan District, Shenyang City, starting from Changqing Street in the west and Guogongzhai Street in the east, encompassing Changqing Bridge, Xinlipu Bridge, Zhuke Street, and portions of Hunnan Expressway (Fig. 2).The roads are wide.Among the pavement streets, Hunnan Middle Road is a two-way eight-lane road, whereas other pavement streets are two-way six-lane roads.Both sides of the road have buildings that are far away, less sheltered, and have good visibility.These factors are conducive to the extraction of road aging information from remote sensing images. 48In spite of the fact that the aging condition of the research objects differs, the majority of the road surface is still intact.There are only a few sections with individual damage phenomena, such as cracks, potholes, and other conditions.In the course of the field investigation, six sections of asphalt pavement with varying aging degrees were selected for the actual measurement of pavement technology (Fig. 2), which was mainly used to study the spectral characteristics of asphalt pavement with varying aging levels.

Introduction to GF-2 High-Resolution Remote Sensing Data
The GF-2 Satellite is the first civilian optical remote sensing satellite independently developed by China with a spatial resolution >1 m.It is equipped with two cameras with panchromatic resolutions of 1 m and multispectral resolutions of 4 m.GF-2 imagery was adopted for the experiment data on November 5, 2021, and its data number is GF2_PMS2_E123.5_N41.8_20211105_L1A0006016599.Table 1 provides detailed information.The Research and Application Center of Liaoning Town and the Traffic Environment of the High-Resolution Earth Observation System of Shenyang Jianzhu University is the data source.ArcGIS 10.2 is used to capture vector data of the roads in the area to prepare for further analysis.

Assessment of Pavement Technical Condition
Currently, it is difficult to distinguish the aging status of asphalt pavement using a set of recognized accurate numbers or standards, although a corresponding index of highway technical conditions can be used to approximate the aging status of different roads to a certain extent.In order to facilitate the analysis of the spectral change characteristics of these aged asphalt pavements, this paper calculates the PCI according to the highway technical condition assessment method stipulated in the Highway Technical Condition Assessment Standard (JTG 5210-2018), thus indirectly indicating the aging condition of different asphalt pavements.PCI is determined by the type and severity of road damage.The severity of the damage can be determined by counting the geometric characteristics of the disease.Among these characteristics are the length and density of cracks, the number and size of potholes, the length and width of ruts, etc. [49][50][51] The PCI equations are Eqs.( 1) and ( 2): E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 2 ; 1 1 7 ; 5 5 6 In the equations, DR is the pavement damage rate (pavement distress ratio), which is the sum of various diseases and that of the pavement survey area (%).The A i is the area (m 2 ).A is the pavement area (the product of survey length and effective pavement width, m 2 ).The w i is the weight of class I pavement disease damage, and the value varies according to different pavement types.For details, refer to the Highway Technical Conditions Assessment Standard (JTG 5210-2018).Model parameter a 0 is 15.00 for asphalt pavement.Model parameter a 1 is 0.421 for asphalt pavement.The i is the pavement damage type of item i considering the degree of damage (light, medium, and heavy).The i 0 is the total number of damage types including the degree of damage (light, medium, and heavy), and the value for asphalt pavement is 21. 2

Spectral Characteristic Analysis of Asphalt Pavement
During the aging process of asphalt pavement, spectral changes are primarily caused by changes in the pavement chemical composition and the degree of exposure of rock aggregate (Fig. 3). 38ased on the spectral data obtained from the ground measurement using a FieldSpec ProFR portable earth spectral radiation meter, Jin found significant differences between the existing and old roads between 400 and 900 nm (Fig. 4).It has been found that the more serious the pavement aging, the faster the reflectivity increases, and this is not affected by the value of the reflectivity. 38heng measured the spectral reflectance of road surfaces using the ASD FieldSpec spectrometer.After analyzing the spectral curve by guide and envelope removal, the reflectivity of the asphalt pavement within 400 to 680 nm increases; the reflectivity of the spectral curve of the aged asphalt pavement decreases between 860 and 970 nm, whereas the new asphalt pavement increases (Fig. 5). 4 According to the above characteristics of the asphalt pavement spectral curve, combined with the wavelength coverage of GF-2, the GF-2 blue band (450 to 520 nm), red band (630 to 690 nm), and near-infrared band (770 to 890 nm) in the range of 450 to 890 nm are selected as parameters for studying the aging index of asphalt pavement.The results of Jin et al., 38 Pan et al., 2 Cheng et al., 4 and Xu's 45 research have led to the construction of a spectral index model that reflects the aging condition of asphalt pavement and is suitable for use with GF-2 data. 38) DHI is a spectral index of road health based on the difference in reflectance between the near-infrared band and the blue light band.In general, the newer the road surface, the smaller the reflectance gap between the NIR and blue light bands; the older the road surface, the larger the reflectance gap: 38   ( The RHI is a spectral index used to measure the health of roads based on the reflectivity of the near-infrared and blue light bands: 38 E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 4 ; 1 1 7 ; 7 1 3 (3) The normalized difference health index (NDHI) is a health index based on the normalized vegetation index: 38 E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 5 ; 1 1 7 ; 6 6 3 B 4 and B 1 are the reflectivities of bands 4 and 1 of GF-2 spectral data, respectively.
4 Experimental Results

Test Results of PCI
According to the highway technical condition evaluation method specified in the "Highway Performance Assessment Standards" (JTG 5210-2018), the length of a PCI calculation unit is 100 m, and the calculated width is the actual road width.So, the PCI computing units for Changqing Bridge and Xinlipu Bridge are 100 × 3.6 × 5 × 2 ¼ 3600 m 2 , the PCI calculation units for Hunnan Middle Road, Hunnan Expressway, Zhuke Street, and Guogongzhai Street are 100 × 3.6 × 4 × 2 ¼ 2880 m 2 .According to the specific values calculated by PCI, the aging condition of asphalt pavement is divided into five levels: excellent (≥90), acceptable (≥80, <90), average (≥70, <80), unacceptable (≥60, <70), and extremely poor (<60).After on-site measurements and calculations based on Eqs. ( 1) and ( 2), the PCI and aging degree of six experimental objects (18 PCI calculation units) were obtained as shown in Table 2.

Verification Results of the Aging Index of Asphalt Pavement
To conduct the asphalt pavement aging condition extraction experiment, the corresponding sample road sections (Fig. 6) were selected from the GF-2 high-resolution remote sensing image according to the three pavement health monitoring index Eqs.( 3)-( 5).To extract the aging conditions of the asphalt pavement of the experimental objects, three health monitoring indexes were used in turn using the Update Range tool of eCognition.As shown in Fig. 7, the experimental results of Guogongzhai Street and Xinlipu Bridge have been selected for display.
Based on the comparison of the above experimental data, we can conclude that the three indexes can reflect asphalt pavement aging to varying degrees but that there are some differences in extraction results and applicability (Fig. 8).Upon comparison and analysis of the RHI and NDHI extraction results with the actual measurement results of pavement aging, it can be concluded that the three indexes can accurately reflect the aging condition of asphalt pavement.As an asphalt pavement ages, it is strongly affected by external factors, and there are some differences between DHI and the actual aging process.In the RHI extraction, the shadow area can be clearly seen, but the presence of the vehicle cannot be determined.NDHI can be used to extract the shadow area, but it cannot be used to extract the vehicle information.Despite the fact that DHI extraction is capable of clearly displaying the positions of vehicles, it is highly insensitive to the presence of shaded areas.
Using the data obtained from the tested samples, univariate linear regression models were established using IBM SPSS Statistics software for PCI and DHI, RHI, and NDHI (Fig. 9).Accordingly, the simple linear correlation coefficients R between PCI and DHI, RHI, and NDHI are −0.41,0.58, and 0.61, respectively (Fig. 9), with corresponding root mean square errors of 15.80, 13.94, and 13.65, respectively.As can be seen, PCI has a low positive correlation with RHI and NDHI while having a low negative correlation with DHI.Correlation coefficient values of −0.41, 0.58, and 0.61 represent relatively low correlations.
Among the three aging condition monitoring indexes, RHI and NDHI are relatively effective.To a certain extent, they can be used to extract information regarding asphalt pavement aging conditions without interfering with the data.In the case of common shadows and vehicle interference, however, the two indices cannot be effectively distinguished.RHI and NDHI are therefore unable to combine GF-2 high-resolution image data for extensive analysis of road aging conditions.Note: The selected road section for the experiment is located in Shenyang city, with a short road renovation cycle and generally not severe aging of asphalt roads.Moreover, due to the low number of heavy vehicles in the urban area, other distresses, such as ruts, potholes, and looseness, except for cracks, are rarely found in the experimental area.In the experimental measurement section, only three types of distresses were found: cracks, map cracks, and ruts, with cracks being the main ones.Fig. 9 A uni-linear regression model for PCI and DHI, RHI, and NDHI.

Innovative Results of the Aging Index of Asphalt Pavement
After an analysis of the aging spectral curve of asphalt pavement and summarizing a large number of experimental results, this paper proposes the pavement aging index, PAI, a new aging index for asphalt pavement.In order to make RHI and NDHI more practical for asphalt pavement aging monitoring, PAI combines the calculation characteristics of the two indexes.As the value range of the membership function in the eCognition software is continuously updated, the correlation between each band of the two indexes and the aging condition of asphalt pavement is compared and observed.The correlation between the B 4 band and road aging is relatively low, and the NDHI extraction results clearly illustrate the shadow area.Therefore, the new equation is based on the NDHI formula, where B 4 is replaced by the NDHI value and then multiplied by the NDHI value to account for asphalt aging.PAI [Eq.( 6)] is as follows [NDHI in the equation is the normalized difference health index in Eq. ( 5)].By using this method, Fig. 10 shows the asphalt pavement aging conditions of Guogongzhai Street and Xinlipu Bridge: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 6 ; 1 1 4 ; 3 8 2 After comparing and analyzing the actual field survey data with the extraction results of DHI, RHI, and NDHI, it has been determined that PAI can not only be used to indicate asphalt pavement aging but also accurately identify shadow areas and vehicles on the road surface (Fig. 11).
As a result of collecting the sample points from the high-resolution remote sensing images, the PAI of each experimental section is presented in Fig. 12.In accordance with the simple linear correlation analysis formula, the simple linear correlation coefficient R between PCI and PAI is 0.894 (R 2 ¼ 0.799), and the corresponding root mean square error is 7.77.Therefore, PAI and PCI are strongly correlated (Fig. 13).We tested the significance of the regression results, with statistics R 2 F ¼ 63.8 and significance α ≤ 0.000.Statistics consider the linear relationship between the two relationships to be significant.From Table 3, it can be seen that the characterization effect of PAI is significantly better than previous indices, such as DHI, RHI, and NDHI.The PAI can therefore be used to characterize the aging condition of asphalt pavements.According to the above comprehensive verification, the PAI can be used not only to extract the aging status of asphalt pavements but also to distinguish pavement interference information, such as shadows and vehicles.

Assessment Results of the Aging Condition of the Asphalt Pavement
Figure 14 shows the results of the road aging conditions in the study area based on the PAI.According to the figure, Zhuke Street has the most serious aging condition, followed by Hunnan Middle Road.Guogongzhai Street has also developed significantly in terms of aging.In general, the roads on the Hunnan Expressway and Changqing Bridge are in good condition.In order to verify the accuracy of the evaluation results, this experiment referred to the practices of Pan, 2 Jin, 38 and Cheng 4 and randomly selected on-site ground verification points to compare the aging conditions with the evaluation results.
Three sets of ground verification points were selected for comparison, and the comparison results are shown in Fig. 15.As shown in Fig. 15, the asphalt on the road surface of set ( 1) is completely intact, and the stones are not exposed.It is a very flat and dark-colored road.It is not uncommon for longitudinal and transverse cracks to appear at the edge or center of the road, along with shallow and narrow cracks.As a result, Changqing Bridge pavement is a healthy road.In Fig. 15, the asphalt on the road surface of set ( 2) is partially missing, and the stones are not exposed, resulting in a pitted surface.The color of the road changes to a light gray.There are often transverse cracks running through the road surface, spaced about 10 m apart, and they are deep and wide.Accordingly, Guogongzhai Street is a moderately aged street.Asphalt pavement in set (3) in Fig. 15 lacks an asphalt oil film, and white stones are clearly visible.On the road surface, large areas of map cracks and long extending transverse and longitudinal cracks are evident, usually filled with fine-grained sand.As a result, it can be concluded that Zhuke Street is an aged road that needs to be improved.Results of the actual investigation are entirely consistent with the results of the PAI evaluation, which shows that PAI is capable of identifying different aging conditions of pavements and can be used for a broad range of pavement health assessment and monitoring.
Compared to the traditional PCI index of road physical parameters, such as cracks, ruts, and roughness, a quantitative model of road spectral characteristics can be achieved through remote sensing, rapid monitoring, and evaluation, which does not require a large amount of human resources for field measurement evaluation, allowing for more efficient road maintenance.

Conclusion
As a result of analyzing the spectral characteristics of asphalt pavements with various aging conditions, this paper utilized high-resolution remote sensing images from the GF-2 satellite to verify and compare the asphalt pavement health monitoring index proposed by predecessors, using the asphalt pavement of the Hunnan Middle Road in Shenyang as an example.Using GF-2 data, the extraction index of asphalt pavement aging degree is proposed based on the results of the experiments and analysis.The PAI has been demonstrated to have a significant impact on the aging of asphalt pavements.Following is a summary of the main conclusions: (1) Based on previous achievements, DHI, RHI, and NDHI were validated for GF-2 data.
Based on the experimental verification, the three indices can be used to estimate asphalt pavement aging to varying degrees.In spite of this, there are some differences in their applicability.PCI and RHI show a low positive correlation, with correlation coefficients R values of 0.58 and 0.61, respectively, which can be used to extract asphalt pavement aging conditions without interference information.(2) This paper proposes a new asphalt PAI, namely the PAI.As a result of experimental verification, the method not only accurately extracts the aging status of asphalt pavement but also has a good discriminative effect on interferences, such as shadows and vehicles on the road.Meanwhile, the correlation coefficient R between PAI and PCI was calculated to be 0.894, indicating a significant linear relationship.As a result, PAI can be used to determine the aging status of asphalt pavements.(3) The PAI was used to assess the aging condition of some road sections in the study area.
Furthermore, three sets of ground verification points were selected for comparative verification, and the results confirmed the accuracy of the PAI evaluation.While there have been some successes in this experimental study, there are still some shortcomings: ① The division of road aging degrees corresponding to the indexes is mainly based on empirical values, which have some subjectivity.② This model is only suitable for identifying and detecting lightly aged roads.However, it is not suitable for detecting road diseases, such as upheavals, ruts, and cracks.③ This paper proposes a research model that is applicable only to asphalt pavements. 4

Fig. 2
Fig. 2 Location map of the study area.The positions in the red boxes are the study area.(a) Changqing Bridge section, (b) Hunnan Expressway section, (c) Zhuke Street section, (d) Hunnan Middle Road section, (e) Guogongzhai Street section, and (f) Xinlipu Bridge section.The entire area represented by the figure is ∼6560 m long from east to west and 5645 m long from north to south.

E
Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 3 ; 1 1 4 ; 9 6 DHI ¼ B 4 − B 1 :

Fig. 6
Fig. 6 The corresponding position of the actually measured working road section in the remote sensing image.(a) Changqing Bridge section, (b) Hunnan Expressway section, (c) Zhuke Street section, (d) Hunnan Middle Road section, (e) Guogongzhai Street section, and (f) Xinlipu Bridge section.

Fig. 7
Fig. 7 Extraction results of the asphalt pavement aging condition in the corresponding road sections.The upper row displays the extraction results of DHI (left), RHI (middle), and NDHI (right) on Guogongzhai Street.The lower row displays the DHI (left), RHI (middle), and NDHI (right) extraction results of the Xinlipu Bridge.The different shades of colors in the picture represent different aging conditions of the pavement.

Fig. 8
Fig. 8 The extraction results of different indexes on the shaded area and vehicles on the road of Hunnan Middle Road.The first figure is the remote sensing image of a section of Hunnan Middle Road, which clearly shows the road under the shadow of buildings and some vehicles on the left side of the road.The second, third, and fourth figures are the results of DHI, RHI, and NDHI extraction.In the third and fourth figures, the shaded parts are divided in the middle because of the influence of urban light rail.

Fig. 10
Fig. 10 PAI extraction results of (a) Guogongzhai Street and (b) Xinlipu Bridge.The different shades of colors in the picture represent the different aging conditions of the pavement.

Fig. 11
Fig. 11 The extraction result of PAI on the shaded area and vehicles on the road of Hunan Middle Road.(a) The remote sensing image of a section of Hunnan Middle Road, which clearly shows the road under the shadow of buildings and some vehicles on the left side of the road.(b) The result of PAI extraction.In (b), the shaded area and vehicles can be clearly displayed.The shaded parts are divided in the middle because of the influence of urban light rail.

Fig. 12
Fig.12The PAI in the high-resolution remote sensing images of each experimental section.

Fig. 14
Fig. 14 Assessment results of road aging conditions in the study area.The two endpoint positions of the indicator line are the starting and ending points of the study road sections, respectively.

Fig. 15
Fig.15The aging condition and existing cracks of the pavement at three sets of ground verification points.The first, second, and third lines display the aging condition and existing cracks of the asphalt pavement in Changqing Bridge, Guogongzhai Street, and Zhuke Street, respectively.

Table 1
Detailed information of experimental image data.

Table 2
PCI and aging determination of the subjects.

Table 3
Comparison of characterization effects between health monitoring indexes and PCI.