Evaluation of Compaction Uniformity of Asphalt Pavement Based on Construction Depth Variation

Abstract

The continuous monitoring and real-time feedback on the uniformity of the mixture after paving and compaction is important to guide the construction and improve the life of the asphalt pavement. Thus, in this study, the latest digital image technology was used to collect images of asphalt pavement after paving and compaction according to the set parameters. The images collected were preprocessed in steps such as binary grayscale, filtering, histogram, and small particle filtering, which established the method for calculating the construction depth $H_P$ of the digital images of pavement. Moreover, a method was proposed to evaluate the uniformity of asphalt pavement using the construction depth coefficient of variation. The coefficients of variation were computed based on 100 images of the dependent project and compared with the actual images. The results indicated that the construction depth coefficient of variation of the digital images obtained by the developed calculation procedure was in agreement with the actual situation and combined the correlation degree of the calculation results. Small areas of 4² are recommended as the division of the asphalt pavement compaction uniformity evaluation method based on construction depth variation, and the use of $C_{v4}\ge 8.0$ as the evaluation standard for the segregation of asphalt pavement is recommended. The method proposed is a new and rapid test to evaluate and detect the asphalt pavement uniformity.

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

The segregation of asphalt mixtures is a common issue to asphalt concrete pavement surfaces. This segregation is accelerated by the dynamic load of vehicles, resulting in the development of potholes and pileups, which affect both the aesthetics and performance of the pavement in serious cases [1, 2]. Thus, the segregation of asphalt mixtures after paving is one of the root causes of local damage of asphalt concrete pavement [3].

Many studies have been conducted by domestic and international scholars to determine the segregation of asphalt mixture. The uniformity of asphalt mixtures mainly refers to the uniformity of distribution of coarse and fine aggregates [4]. Brock [5] stated that pavement segregation can lead to changes in mixture gradation and asphalt content, resulting in poor pavement structural performance. Kennedy et al. [6] analysed the phenomenon and mechanism of asphalt mixture segregation and proposed that it is related to the nonuniform dispersion of coarse and fine aggregates. Nondestructive technologies have been developed for the identification and evaluation of asphalt mixture homogeneity, such as thermal infrared imager, ground penetrating radar, and laser surface profiler [7–9]. Gilbert [10] used a thermal infrared imager to test the temperature distribution of the asphalt mixture during the construction process to reflect the degree of mixture segregation. The author found that the temperature of the asphalt mixture did not necessarily lead to aggregate segregation. Schmitt et al. [11] and Liu [12] used pavement ground-penetrating radar, which can determine the distribution of density of asphalt concrete pavement.

As a new interdisciplinary technology, the birth of digital image processing can be dated back to the 1960s and has become increasingly mature along with the development of computer technology and digital signal processing technology. In recent years, digital image processing technology has developed rapidly in the field of road engineering and has been applied to the quantitative evaluation of asphalt mixture uniformity research [13, 14]. Azari [15] verified the characteristics of aggregate distribution in computed tomography scan images of asphalt mixture specimens using the normal distribution method in statistics. If the distribution characteristics satisfied the standard normal distribution, the distribution was found to be uniform. Bruno et al. [16] analysed the gradation of aggregates by digital image acquisition of core samples after drilling and core samples of on-site asphalt pavement and extracted the aggregate particles in the core samples using a digital image method.

Considering the abovementioned studies, the application of digital image technology in asphalt mixtures has been mainly used to evaluate and analyse aggregate particle uniformity and asphalt mixture paving uniformity, whereas few studies have focused on the uniformity of asphalt mixture after paving and compaction and its correlation analysis with traditional methods. Therefore, in this study, for the asphalt mixture compaction stage, digital image processing technology was used to construct the relationship between the construction depth in the traditional sand paving method and the construction depth using digital image processing. The evaluation method of asphalt pavement compaction uniformity based on the construction depth variation was proposed to evaluate the compaction uniformity of asphalt pavement with the variation coefficient as the evaluation index.

2. Image Acquisition and Processing Technology

2.1. Image Acquisition

Acquisition of images is the first step in evaluating the uniformity of aggregate distribution in asphalt mixtures. The sketch of asphalt mixture image acquisition and calculation is shown in Figure 1. In this study, the image acquisition was tested on a sunny day, the metal oxide semiconductor COMS industrial camera in Figure 2 height from the asphalt mixture paving surface was 60 cm, and the front is vertical downward. The area of paving asphalt mixture captured by each image was measured to be $72\text{cm}\times 72\text{cm}$. The image pixel resolution was $2448\times 2448$ according to the performance of the COMS industrial camera used for image acquisition at the construction site, shooting height, etc.

[figure(s) omitted; refer to PDF]

2.2. Image Preprocessing

To calculate the statistical analysis of the collected asphalt mixture images, the photos need to be processed. We collected photos of the middle pavement surface layer AC-20 asphalt mixture (The pavement asphalt layer has three layers, the lower layer is ATB-30, the middle layer is AC-20, and the upper layer is AC-16.) at Shiyan highway after paving and compaction. A total of 586 photos (shooting without overlap) were preprocessed. Figure 3 shows an example of the preprocessing, taking photo SY-001 as an example, where SY denotes the name of Shiyan highway, and 001 denotes the photo number.

[figure(s) omitted; refer to PDF]

The acquired raw images were preprocessed by MATLAB software, and the processing steps were as follows: (1) greying the colour images using the rgb2gray function; (2) Use function medfilt2 for spatial filtering and denoising removal to improve the image quality; (3) histogram equalization to enhance the image contrast and improve the image clarity; (4) morphological processing and watershed segmentation of the histogram equalized images to identify and segment the adherent particles. After this process, the final binary image was obtained (Figure 3(b)).

3. Calculation Method of Construction Depth of Asphalt Pavement

3.1. Computational Model

As the surface of asphalt pavement is rough and uneven, light irradiation on the pavement will be diffusely reflected, and the light intensity of diffuse reflection will vary with the roughness of the middle pavement surface. Therefore, when the camera is used to capture the image of asphalt mixture, different light intensities are displayed differently in the image.

In this study, asphalt mixture compaction was the research object and the shooting occurred in natural light on a clear day. Studies reported that there is a linear relationship between the grayscale value and the construction depth (i.e., the greater the pavement construction depth, the smaller the grayscale value of the image) [17]. Therefore, digital image processing technology was used to calculate the depth of asphalt pavement construction. The average construction depth of the pavement image was calculated by reflecting the convexity of the asphalt pavement surface by the brightness and darkness of the image pixels, and the difference in height between the raised and concave pavement was reflected by the grayscale difference to construct a three-dimensional (3D) surface [18].

The pavement images acquired through the camera were red-green-blue images, which could be represented after preprocessing by MATLAB toolbox functions as an $m\times n$ of a two-dimensional (2D) matrix. If $Z_{ij}=X_i,Y_j$, then a 3D surface could be constructed (i.e., the surface texture construction map of the asphalt pavement). Figures 4 and 5 show the 2D matrix of the asphalt pavement image and the 3D surface of the asphalt pavement surface image, respectively.

[figure(s) omitted; refer to PDF]

Based on the data from the 2D matrix, the pavement construction depth was calculated as $$\begin{array}{}\text{(1)}& Z=Fx,y,\text{(2)}& V={\iint }_D{F}_{\max }-Fx,y\text{dxdy},\end{array}$$where $Z$ is the pixel value; $x,y$ are the horizontal and vertical coordinates corresponding to the pixel, respectively; $V$ is the volume of space enclosed by the pixel value of the surface construction; $F_{\max }$ denotes the largest grey value in the grey matrix; $Fx,y$ is the grey value of the pixel point whose coordinates are $x,y$; the integration region $D$ is equivalent to the sand spreading range in the sand spreading method; $V_p$ is the volume enclosed by the spatial surface and plane $F_{\max }$ of the pavement surface construction, which is equivalent to the sand spreading volume in the sand spreading method.

Digital image processing technology was used to calculate the digital construction depth $H_p$ of the pavement to characterize the roughness. The higher the $H_p$ value, the rougher the pavement. The formula for calculating the digital construction depth $H_p$ of the pavement is $$\begin{array}{}\text{(3)}& H_p=\frac{V}{D},\end{array}$$where $D$ is the integral region, calculated as the product of the number of rows and columns of the matrix.

The pavement surface construction depth $H_p$ needs to be corrected by the scale correction factor $\mu$ to obtain the final digital construction depth $H^{\prime }$ of the asphalt pavement surface. The calculation formulas are $$\begin{array}{}\text{(4)}& {\text{H}}^{\text{’}}=\frac{H_P}{\mu }=\frac{H_P}{Z_{\max }-Z_{\min }/\alpha },\text{(5)}& \mu =\frac{Z_{\max }-Z_{\min }}{\alpha },\end{array}$$where $Z_{\max }$ and $Z_{\min }$ are the maximum and minimum values of all elements in the matrix $Z$, respectively; $\alpha$ is the empirical coefficient and takes the value of 1.36.

3.2. Calculation Procedure for the Coefficient of Variation of the Construction Depth

3.2.1. Calculation Method

The asphalt pavement images can be divided into equal parts by 2², 3², 4², 5², and 10² as shown in Figure 6, and the mean depth of construction, variance between regions ($S$) and coefficient of variation ($C_v$), is calculated for each equal part of the area. The calculation formulas are $$\begin{array}{}\text{(6)}& {\overline{H}}_p=\frac{{\sum }_{i=1}^n{H}_{\text{Pi}}}{n},\text{(7)}& S=\sqrt{\frac{{H_{\text{Pi}}-{\overline{H}}_P}^2}{n-1}},\text{(8)}& C_v=\frac{S}{H_P\times 100},\end{array}$$where $H_{\text{Pi}}$ is the average digital construction depth of each aliquot (mm); ${\overline{H}}_P$ is the average digital construction depth of the image (mm); $n$ is the number of small regions divided, $n=4$, 9, 16, 25, 100.

[figure(s) omitted; refer to PDF]

3.2.2. Procedure for Calculating Coefficients of Variation for Digital Image Construction Depth

In this study, MATLAB software was used to write a fast procedure for calculating the coefficient of variation of the construction depth from asphalt pavement.

MATLAB software was run, and the pavement digital construction depth discrepancy coefficient of variation calculation program, whose main interface is shown in Figure 7, was loaded. The main functions of this calculation program include create task, load image, automate, draw pavement texture construction map, and calculate and export, which are controlled by five buttons on the main interface.

[figure(s) omitted; refer to PDF]

4. Correlation Analysis of the Coefficient of Variation ($C_v$) and Aggregate Segregation

4.1. Correlation Analysis of Coefficient of Variation of Construction Depth from Digital Images

The acquired images of the medium surface AC-20 asphalt mixture compacted pavement were imported into the MATLAB calculation program. The images were segmented according to five different aliquot methods of 2², 3², 4², 5² and 10², and the coefficient of variation $C_v$ of the pavement digital image construction were calculated separately. The calculation results are shown in Table 1.

Table 1

Depth of departure coefficient of variation for digital image construction of AC-20 asphalt mixture pavement.

Number	$C_{v2}$	$C_{v3}$	$C_{v4}$	$C_{v5}$	$C_{v10}$
1	4.68	5.04	5.18	5.39	5.60
2	3.17	4.44	5.66	5.65	6.48
3	4.38	5.92	5.65	5.76	6.02
4	4.29	4.81	4.69	5.35	5.43
5	7.86	7.38	7.12	7.09	7.29
6	3.15	3.87	3.55	4.24	4.88
7	2.15	2.77	2.41	2.59	2.80
8	3.58	3.29	4.49	4.60	5.47
9	5.49	6.02	6.04	6.46	6.98
10	5.35	5.78	5.09	5.51	5.51
11	2.12	2.76	3.57	3.37	3.84
......	......	......	......	......	......
93	3.16	3.29	3.36	2.97	3.80
94	6.46	6.71	6.65	6.69	7.02
95	10.52	9.95	9.98	10.32	10.46
96	5.88	6.35	6.31	6.42	6.42
97	3.46	3.77	3.37	3.81	4.10
98	5.33	5.42	5.33	5.21	5.18
99	6.31	5.80	5.80	5.92	6.20
100	7.35	7.16	7.12	6.79	7.30

The calculated results of the $C_v$ values were compared to the corresponding asphalt pavement images by visual observation and were overall consistent with the expected assumption that the higher the $C_v$ value for the digital image construction of the asphalt pavement, the higher the likelihood of aggregate segregation. This tendency can be observed in Figures 8(a) and 8(b) for uniformity of mix compaction and inhomogeneity, respectively. The $C_v$ results of the computation of the construction depth coefficient of variation in Table 1 were as follows: Figure 8(a), $C_{v2}=2.12$, $C_{v3}=2.76$, $C_{v4}=3.57$, $C_{v5}=3.37$, $C_{v10}=3.84$; Figure 8(b), $C_{v2}=12.75$, $C_{v3}=12.94$, $C_{v4}=13.35$, $C_{v5}=13.16$, $C_{v10}=14.08$. Thus, the values were corroborated with the visually observed uniformity of the asphalt pavement aggregate distribution.

[figure(s) omitted; refer to PDF]

4.2. Correlation Analysis of Depth Coefficient of Variation for Digital Image Construction with Different Equal Area Division Methods

The digital image construction depth coefficient of variation $C_v$ can be calculated according to five image segmentation methods of 2², 3², 4², 5², and 10². From Table 1, the $C_v$ values of the five segmentation methods of each image were different, but from the overall analysis of the 100 acquired asphalt pavement images, they showed a certain correlation. Therefore, a correlation analysis between two $C_v$ values was conducted, and a more stable and representative partitioning method was selected as the standard for calculating the coefficient of variation for asphalt pavements based on the results of the correlation analysis.

Taking the image divided by 2² and 3² as an example, the distribution of the digital image construction depth coefficients of variation $C_{v2}$ and $C_{v3}$ is shown in Figure 9. As observed, the same image divided by 2² and 3² division methods correspond to different construction depth coefficients of variation $C_{v2}$ and $C_{v3}$, respectively. However, from the overall analysis of 100 asphalt pavement images, the tendency of the construction depth coefficients of variation $C_{v2}$ and $C_{v3}$ is consistent. A linear regression with $C_{v2}$ as the independent variable $X$ and $C_{v3}$ as the variable $Y$ is shown in Figure 10. The linear regression equation is $Y=0.8509X+0.9335$, and the squared correlation coefficient is $R^2=0.91$. Thus, there is a good linear correlation between $C_{v2}$ and $C_{v3}$.

[figure(s) omitted; refer to PDF]

Using the same method, the linear regression equations and square of the correlation coefficients were obtained by constructing a correlation analysis of the coefficients of variation $C_{v2}$, $C_{v3}$, $C_{v4}$, $C_{v5}$, and $C_{v10}$ for the digital images, as shown in Table 2.

Table 2

Correlation relationships of digital image construction depth coefficients of variation $C_{v2}$, $C_{v3}$, $C_{v4}$, $C_{v5}$, and $C_{v10}$.

Independent variable	Variable	Linear regression equation (Math)	Square of correlation coefficient $R^2$
$$C_{v2}$$	$$C_{v3}$$	$$C_{v3}=0.8509C_{v2}+0.9335$$	0.9064
	$$C_{v4}$$	$$C_{v4}=0.8336C_{v2}+1.1385$$	0.8851
	$$C_{v5}$$	$$C_{v5}=0.8086C_{v2}+1.3984$$	0.8750
	$$C_{v10}$$	$$C_{v10}=0.782C_{v2}+1.8718$$	0.8297
$$C_{v3}$$	$$C_{v4}$$	$$C_{v3}=0.9735C_{v4}+0.2566$$	0.9642
	$$C_{v5}$$	$$C_{v3}=0.9467C_{v5}+0.5305$$	0.9580
	$$C_{v10}$$	$$C_{v3}=0.212C_{v10}+1.0028$$	0.9195
$$C_{v4}$$	$$C_{v5}$$	$$C_{v4}=0.9663C_{v5}+0.3141$$	0.9810
	$$C_{v10}$$	$$C_{v4}=0.9479C_{v10}+0.7509$$	0.9570
$$C_{v5}$$	$$C_{v10}$$	$$C_{v5}=0.981C_{v10}+0.4427$$	0.9755

From Table 2, the square of the correlation coefficient $R^2$ between $C_{v2}$ and $C_{v3}$, $C_{v4}$, $C_{v5}$, $C_{v10}$ ranged between 0.83 and 0.91, whereas the squares of the correlation coefficients $R^2$ between $C_{v3}$, $C_{v4}$, $C_{v5}$, $C_{v10}$ and the other four division methods ranged between 0.91–0.96, 0.89–0.98, 0.88–0.98, and 0.83–0.98, respectively. Thus, the correlations between the division methods of 2² and 10² and the other four division methods were relatively weak, whereas there was a strong correlation between the coefficients of variation from the construction depth of digital image calculated by aliquoting images by 3², 4², and 5² division methods. Considering that the more regions are divided, the slower their operation speed, and combined with the correlation degree of the calculation results, we determined that the best evaluation method of asphalt pavement compaction uniformity based on the construction depth variation consists of dividing the image into 4² small areas.

4.3. Evaluation Criteria Based on the Coefficient of Variation of the Construction Depth

According to the results of the coefficient of variation $C_{v4}$ of the construction depth based on asphalt pavement images calculated using MATLAB, combined with the visual observation of the asphalt pavement aggregate distribution uniformity, it was found that when $C_{v4}$ took a value close, equal, or greater than 8, there was a relatively evident segregation of the asphalt pavement aggregate distribution based on visual observation. As a result, the preliminary evaluation criterion of asphalt pavement segregation was set to $C_{v4}=8.0$.

For $C_{v4}=8.0$, the corresponding $C_{v2}$, $C_{v3}$, $C_{v5}$, and $C_{v10}$ values were calculated according to the correlation relationship equation in Table 2 as 7.875, 7.907, 8.071, and 8.334, respectively. From the calculated 100 asphalt pavement images, the images with values of $C_{v2}$, $C_{v3}$, $C_{v5}$, and $C_{v10}$ greater than or equal to 8 were selected. After screening, those with a coefficient of variation of $C_{v4}\ge 8.0$ are presented in Table 3.

Table 3

Images of AC-20 asphalt mixture pavement with coefficient of variation $C_{v4}\ge 8.0$.

Project name	$C_{v2}$	$C_{v3}$	$C_{v4}$	$C_{v5}$	$C_{v10}$
Image number	TXYS-70			TXYS-22
	TXYS-53	TXYS-22	TXYS-53	TXYS-53	TXYS-68
	TXYS-21	TXYS-53	TXYS-95	TXYS-21	TXYS-53
	TXYS-51	TXYS-95	TXYS-15	TXYS-51	TXYS-95
	TXYS-20	TXYS-15	TXYS-30	TXYS-15	TXYS-15
	TXYS-95	TXYS-30	TXYS-20	TXYS-30	TXYS-30
	TXYS-15	TXYS-20	TXYS-39	TXYS-95	TXYS-20
	TXYS-30	TXYS-39		TXYS-20	TXYS-39
	TXYS-39			TXYS-39
Number of images	9	7	6	9	7

Using $C_{v4}=8.0$ as the boundary value of aggregate segregation for the following layer of AC-20 asphalt mixture pavement, six asphalt pavement images were screened, and the asphalt pavement corresponding to these six images was considered to exhibit aggregate segregation, as shown in Figure 11. By observation, these asphalt pavements exhibited a more evident coarse and fine aggregate separation, indicating that the use of $C_{v4}=8.0$ for aggregate inhomogeneity determination was in accordance with the actual situation.

[figure(s) omitted; refer to PDF]

5. Conclusions

In this study, the compaction uniformity of asphalt pavement was evaluated based on the construction depth segregation for the compacted pavement of AC-20 asphalt mixture using digital image processing technology and MATLAB software programming. The main findings include

(1) A calculation method for calculating the construction depth $H_p$ of digital images of pavements was established by characterizing the height difference between raised and depressed pavements in terms of the image grayscale difference, and a method for evaluating the uniformity of asphalt pavements using the construction depth coefficient of variation was proposed

(2) A program was written in MATLAB to quickly calculate the digital construction depth coefficients of variation for pavements, and the coefficient of variation was also calculated for 100 images of the dependent project and compared with the actual images; it was found that the coefficients of variation calculated by the developed program were in accordance with the actual situation

(3) By analysing the correlation between different division regions two by two, it was found that the correlation between 2² and 10² division methods was weak, whereas the division methods of 3², 4², and 5² exhibited strong correlations. Combined with the operation speed, a small area of 4² is recommended as the division method of asphalt pavement compaction uniformity evaluation based on the construction depth variation

(4) By visually comparing and analysing the observed and calculated values, it is recommended to use $C_{v4}\ge 8.0$ as the evaluation criterion for asphalt pavement segregation

Acknowledgments

This work was supported by the Chongqing Natural Science Foundation Project (Grant no. cstc2020jcyj-msxmX0320) and the Science and Technology Research Program of Chongqing Municipal Education Commission (Grant No. KJQN202104001).

References

[1] W. Qin, Study on Segregation of Pavement Materials Based on Granular Characteristics, 2011.

[2] H. Zhu, Research on Application of New Method for Quick Evaluation of Segregation of Asphalt Mixture, vol. 32 no. 1, 2016.

[3] S. L. Li, N. X. Liang, S. Zeng, "Influence of acquisition methods on segregation evaluation of digital image of asphalt mixture," Journal of Chongqing Jiaotong University (Natural Science), vol. 40 no. 3, pp. 103-107, 2021.

[4] Y. F. Luo, J. Q. Zhang, C. Zhang, "Sensitivity analysis of factors influencing shear stress in asphalt pavements under high temperature conditions," Journal of Wuhan University (Engineering Edition), vol. 51 no. 10, pp. 895-900, 2018.

[5] J. D. Brock, Segregation of Asphaltic Mixtures (with Discussion), 1986.

[6] T. W. Kennedy, R. B. Mcgennis, R. J. Holmgreen, Asphalt Mixture Segregation: Diagnostics and Remedies (with Discussion), 1987.

[7] M. Lesturgie, H. Griffiths, "Guest editorial special issue: “catching the invisible”," IET Radar, Sonar & Navigation, vol. 10 no. 1,DOI: 10.1049/iet-rsn.2015.0603, 2016.

[8] B. Hui, M. Guo, X. F. Liu, "Quality assessment of milling pavement surface using 3D line laser technology," Journal of South China University of Technology (Natural Science Edition), vol. 2018 no. 11,DOI: 10.1155/2018/1343752, 2018.

[9] L. Q. Hu, Y. B. Tuan, T. L. Wang, "Analysis of temperature uniformity of asphalt mixture paving based on thermal imaging," Journal of Jiangsu University (Natural Science Edition), vol. 37 no. 6, pp. 735-739, 2016.

[10] K. Gilbert, Thermal Segregation, 2005.

[11] R. L. Schmitt, A. Faheem, I. L. Al-qadi, "Selection of non-destructive testing technologies for asphalt pavement construction," Airfield & Highway Pavement Conference,DOI: 10.1061/9780784413005.046, .

[12] T. Liu, Research on the Evaluation of Dielectric Properties and Construction Quality of Asphalt Pavement Based on Non-Destructive Testing Methods, 2016.

[13] W. L. Wu, Z. Li, X. N. Zhang, "Evaluation of asphalt mixture uniformity by digital image processing technique," Journal of Jilin University (Engineering Edition), vol. 39 no. 4, pp. 921-925, 2009.

[14] J. Ma, X. M. Zhao, S. H. He, "A review of road surface inspection techniques," Journal of Transportation Engineering, vol. 17 no. 5, pp. 121-137, 2017.

[15] H. Azari, Effect of Aggregate Inhomogeneity on Mechanical Properties of Asphalt Mixtures, vol. 66-12, 2005.

[16] L. Bruno, G. Parla, C. Celauro, "Image analysis for detecting aggregate gradation in asphalt mixture from planar images," Construction & Building Materials, vol. 28 no. 1, pp. 21-30, DOI: 10.1016/j.conbuildmat.2011.08.007, 2012.

[17] Y. M. Liu, S. Han, O. M. Xu, "Acquisition system for three-dimensional surface texture of asphalt pavement based on digital image processing," Journal of Southwest Jiaotong University, vol. 49 no. 2, pp. 351-357, 2014.

[18] D. Cheng, Study on Image-Based Texture Analysis Method and Prediction of Skid-Resistance & Tire/Pavement Noise Reduction of HMA, 2015.