Construction and Building Materials 166 (2018) 227–240

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier.com/locate/conbuildmat

Relating surface texture parameters from close range photogrammetry

to Grip-Tester pavement friction measurements
Reginald B. Kogbara a,⇑, Eyad A. Masad a,b, David Woodward c, Phillip Millar c
a
Mechanical Engineering Program, Texas A&M University at Qatar, P.O. Box 23874, Education City, Doha, Qatar
b
Zachry Department of Civil Engineering, Texas A&M University, College Station, TX 77843, USA
c
University of Ulster, Shore Road, Newtownabbey, County Antrim, BT37 0QB, UK

h i g h l i g h t s

Asphalt pavement surface texture is measured using close range photogrammetry (CRP).

CRP data processing methods and their impact on measured texture is investigated.

The top 2 mm of the pavement surface gives the best texture-friction correlations.

Identifies texture parameters that best correlate with GripTester-measured friction.

The density of peaks is the most influential indicator of pavement surface friction.

a r t i c l e i n f o a b s t r a c t

Article history: This study utilizes close range photogrammetry (CRP) to measure the texture of asphalt pavement sur-
Received 31 October 2017 faces with a view to explaining the variation of friction measurements obtained using a GripTester. A
Accepted 19 January 2018 handheld camera was employed to capture images at different locations on both lanes of pavement sec-
Available online 22 February 2018
tions using a scale rule for identification of control points. Proprietary software were employed for cre-
ation and analysis of 3D models from the images to determine pavement surface texture parameters.
Keywords: Different scenarios were considered including pavement surface analysis before and after filtering for
Areal texture parameters
micro- and macro-texture separation. Thresholding with respect to height to analyze the top 1–2 mm
Asphalt pavement
Close range photogrammetry
of the surface was also considered. Texture parameters were then related to friction measured at the
Feature parameters image capture locations. The lane with higher friction values generally showed higher individual texture
Pavement friction parameters across different scenarios. However, meaningful texture-friction correlations along the lanes
Skid resistance were only obtained with the top 2 mm of the surface. Stepwise regression indicated that the density of
peaks (Spd) and the peak material volume (Vmp) best correlate (R2 = 0.75–0.76) with friction, but the
Spd is more influential. These parameters can be used as indicators of pavement surface friction during
its service life.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction roughness that is related primarily to the fine-scale angularity

and texture of aggregate particles. It helps cut through the water
Pavement surface texture is defined as the deviations of the film between the aggregate particle and the tire rubber and needs
pavement surface from a true planar surface [1]. It is typically to be present at any speed. The macrotexture is determined by the
divided to four different scales with different wavelengths, namely, size, shape, spacing and arrangement of aggregate particles in the
microtexture (<0.5 mm), macrotexture (0.5–50 mm), megatexture pavement surface. It governs pavement friction at speeds above 90
(50–500 mm) and unevenness (500 mm–50 m) [2]. Pavement fric- km/h on wet pavements [1,3,5].
tion or skid resistance typically increases with an improvement in Conventional measurement of the surface texture of asphalt
the microtexture and macrotexture scales, while megatexture and pavements have focused on the macrotexture as it is a lot easier
unevenness are undesirable [3,4]. The microtexture is the surface to measure than the microtexture. The most common techniques
for macrotexture measurements include the volumetric sand patch
method, circular texture meter and the outflow meter. More
⇑ Corresponding author.
recently, laser-based, stereo photogrammetry, and microscopy
E-mail address: [email protected] (R.B. Kogbara).

https://doi.org/10.1016/j.conbuildmat.2018.01.102
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
228 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

methods have also been employed for macrotexture measurement height and the root mean square slope [6]. Similarly, the density
[3]. There is also a shift to 3D surface texture measurements, which and sharpness of the peaks obtained after empirical decomposition
provide a wide range of information and better represent the char- of the texture profile from a circular texture meter showed good
acteristics of the surface texture [6]. Some techniques for creation correlation with DFT-measured friction [12]. Another study also
of 3D models of pavement surface texture include laser devices and identified the peak density and core material volume from 3D laser
close range photogrammetry (CRP). Nevertheless, laser devices are scanner measurements as the most influential macro- and micro-
expensive, while an ordinary camera can be used for CRP [7]. CRP texture parameters that exhibit fairly good correlation with DFT
entails estimating 3D co-ordinates of points on an object using friction at high- and low-speed in wet conditions [13].
measurements from multiple images captured from different posi- Only a few studies have considered the relationship between
tions with an ordinary camera. From the images captured, 3D mod- pavement texture and friction from continuous friction measuring
els of surface texture are created and analyzed using proprietary equipment (CFME). CFMEs provide better detail about spatial vari-
software such as 3D Flow Zephyr Pro, Digital Surf MountainsMap, ability of pavement friction and have gained increased attention in
etc [7,8]. The tools and resources required for the CRP method are recent times [5]. A study by Kanafi et al. [14] evaluated the rela-
minimal and easily available. tionship between pavement texture measured using a 3D pro-
A range of parameters can be used for surface texture character- filometer and friction measured using a portable CFME. It was
ization. The mean texture depth (MTD), the mean profile depth observed that the full surface profile is subjected to change in
(MPD) and estimated texture depth (ETD) from laser devices are macro- and micro-scales over time. Hence, spectral analysis cannot
the most common for 2D surface texture characterization [9]. Areal separately characterize aggregate surface texture polishing at
roughness parameters are used for 3D surface texture characteriza- actual road conditions. Moreover, the surface texture profile
tion. These include height (root mean square height, arithmetic (macro- and micro-scales) did not correlate well with friction
mean height, skewness, etc), volume (material volume, void vol- [14]. Another study investigated the relationship between macro-
ume) and feature (density of peaks, arithmetic mean peak curva- texture of an airfield runway from a photometric stereo system
ture, etc) parameters. Details of the above surface texture and friction from a CFME, the GripTester [15]. The photometric sys-
characterization parameters have been recently reviewed [3]. A tem uses a minimum of three pavement surface images illumi-
number of studies have attempted to deduce correlations between nated from different directions and isolated from ambient
some surface texture parameters and friction measured using dif- lighting. The variation in intensity of lighting is then used for pave-
ferent equipment. For instance, a study by Torbruegge and Wies ment surface recovery. The study showed reasonable correlation
found no correlation between the ETD and friction measured by (R2  0.5) between texture indicators (MPD, root mean square
the British pendulum tester [10]. roughness and power spectrum energy) and friction [15].
Majority of the studies that investigated pavement texture- In contrast to the above works, the present study investigates
friction relationship used laser devices and the dynamic friction the relationship between texture measurements from a simple
tester (DFT). A significant quadratic polynomial relation was shown and readily available technique, CRP, and a portable CFME – the
to exist between macrotexture from a laser profile tester and DFT- GripTester. The GripTester is an economical management device
measured skid resistance [11]. A number of parameters including for pavement friction evaluation. An ordinary camera can be used
the MPD were used as macrotexture indicators [11]. However, no to easily collect texture information in the CRP technique com-
clear positive relationship was observed between the MPD and pared to more expensive laser devices used in previous studies.
DFT friction in another study [6]. The peak density and the arith- Hence, knowledge of the particular texture parameters determined
metic mean peak curvature of 3D macrotexture images from a 3D from CRP measurements that influence friction would assist in
laser scanner were observed to have significant positive influence design of asphalt pavements to meet skid resistance targets. The
on friction. There was little influence from the arithmetic mean objectives of this study were to:

Fig. 1. Pavement test sections (a) schematic and (b) mix types. Note: PRD – Percentage refusal density, Pen – Penetration grade, QCS – Qatar Construction Specifications
(essentially the Marshall method), PMB: Polymer modified bitumen.
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 229

Fig. 2. Example of a typical (a) scaled dense point cloud in 3DF Zephyr Pro, (b) imported point cloud in MountainsMap, (c) levelled point cloud, (d) extracted and re-levelled
area of interest, (e) general surface form, and (f) surface after form removal.

Investigate CRP data processing methods and their impact on 2. Methodology

measured texture.

Evaluate the texture parameters from CRP measurements that 2.1. Pavement test section(s)
best correlate with friction.

Evaluate the contribution of the micro- and macro-texture The texture and friction measurements were conducted on a
scales to the measured friction. 900-m long pavement section. It has two lanes (North-South (NS)
230 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

Fig. 3. Example of a typical (a) roughness surface, (b) waviness surface, (c) surface texture parameters table, (d) 3D view of roughness surface, and (e) 3D view of waviness
surface.

and South-North (SN) directions.) This section was constructed in GripTester-towing vehicle had an integrated global positioning
2010 as part of an access road to a sand processing plant (Fig. 1a). system (GPS) for the geolocation of data points. Two runs were
The pavement consists of six sections each 150 m long. The made across each lane and the average of the grip numbers was
wearing course layer (50-mm in thickness) was constructed using used.
dense graded asphalt concrete with gabbro aggregate (Fig. 1b).
Gabbro is an intrusive igneous rock that is equivalent in composi- 2.3. CRP measurement of pavement surface texture
tion to basalt. The difference in the six sections is in the asphalt
binder type and mix design. Penetration grade (pen) 40–50, pen A handheld CanonTM 6D EOS digital full frame single lens reflex
60–70 and a polymer-modified bitumen (PMB) based on Pen 60– camera was used to capture images. The camera also has GPS func-
70 with styrene-butadienestyrene (SBS) as a modifier were used tionality for geolocation determination. The images were captured
in the wearing course of the various sections (Fig. 1b). A detailed at twelve different locations (2 for each section) in the outer wheel
description of the pavement test sections including information path in each of the lanes. Each image was captured with at least
on the base, subbase, and subgrade layers is provided elsewhere 60% forward and 30% side overlap with successive images across
[16]. the area being investigated [19]. About twelve images were cap-
tured at a given location under natural lighting conditions. A
2.2. Pavement friction measurements 300-mm nominal length scale rule placed on the pavement surface
during image capture was used for identification of control points.
The GripTester was used to measure pavement friction in early The images were captured in the winter season a couple of weeks
winter in October 2016. The device uses a braked wheel, fixed slip after GripTester measurements of skid resistance on the pavement
(15%) and has transducers that directly measure the drag and load sections.
on the test wheel skidding at just under 15% of the rolling speed. A The images captured for a given location were imported into
gearing ratio of 27:32 between the test and drive wheels is used to 3DF Zephyr Pro v3.142 proprietary software (3DFlow, Verona, Italy)
achieve the 15% fixed slip. The GripTester uses a measuring tire and a dense point cloud was reconstructed from them. The control
with 25.4 cm (10-in) diameter slick tread [17]. The permissible points from the scale rule was used for scaling the dense point
speed range for the device is 5–130 km/h. However, the most usual cloud, which represents a 3D model of the pavement surface. The
speed (50 km/h) for highway testing [18] was used in this study. A dense point cloud was re-oriented and converted to xyz format
water flow rate of approximately 10 L/min was used, which yields in an open source software, Meshlab, and then imported into a pro-
a water film thickness of 0.5 mm for surface wetting. The pave- prietary software, MountainsMap Premium v7.3.7690 (Digital Surf,
ment friction output – grip number – was collected at 5-m inter- Besançon, France), for analysis. The imported xyz file was then sub-
vals in the outer wheel path of each lane of the pavement. The jected to a series of operations before surface texture calculations
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 231

Fig. 4. Screenshots showing the threshold operator alongside the surfaces before and after thresholding with respect to (a) 0.5% material bearing ratio, and (b) top 2 mm
height.
232 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

Table 1
Typical surfaces (point clouds) and their texture parameters in the NS lane.

S Representative surfaces for scenarios in the NS lane Surface texture parameters

First third Middle third Final third TP First Middle Final
1. Surface after form removal Sq 0.572 0.926 0.642
Sa 0.431 0.652 0.462
Vmp 0.044 0.090 0.026
Vmc 0.449 0.641 0.479
Spd 0.007 0.007 0.006
Spc 6.27 22.6 5.24

2a. Roughness surface Sq 0.077 0.163 0.096

Sa 0.041 0.081 0.049
Vmp 0.007 0.014 0.008
Vmc 0.036 0.065 0.036
Spd 0.014 0.009 0.032
Spc 10 43.6 11

2b. Waviness surface Sq 0.56 0.899 0.634

Sa 0.422 0.63 0.456
Vmp 0.043 0.089 0.026
Vmc 0.438 0.616 0.471
Spd 0.004 0.003 0.003
Spc 1.5 3.3 1.28

3. Top 1 mm Sq 0.111 8  105 2  104

0.5% MBR TH Sa 0.037 2  105 1  104
Vmp 0.020 0.008 0.026
Vmc 4  105 8  106 0.037
Spd 0.008 0.003 0.023
Spc 5.08 17.6 3.96

4. Top 2 mm Sq 0.364 0.194 0.464

0.5% MBR TH Sa 0.259 0.051 0.38
Vmp 0.042 0.032 0.043
Vmc 0.161 4  105 0.481
Spd 0.028 0.007 0.032
Spc 5.04 19.5 4.86

Note: S: Scenario, NS: North-South, MBR: Material bearing ratio, TH: Thresholding, Sq: Root mean square height (mm), TP: Texture parameter, Sa: Arithmetic mean height
(mm), Vmp: Peak material volume at 10% MBR (mm3/mm2), Vmc: Core material volume (mm3/mm2), Spd: Density of peaks (1/mm2), Spc: Arithmetic mean peak curvature
(1/mm2).

in MountainsMap. These include leveling with respect to a least arithmetic mean peak curvature) parameters. The texture parame-
squares plane and extraction of the area of interest to remove ters are described as follows [24]:
redundant data. The leveling step(s) were carried out so that height
parameters were not affected by an artificial gradient. Further-
Root-mean-square height (Sq): The Sq is equivalent to the stan-
more, the general form, which approximates to any surface rip- dard deviation of heights within the defined area. It represents
pling, was removed from the area of interest using a second the root-mean-square of ordinate values in the area.
order polynomial. The Robust Gaussian filter operator was also
Arithmetic mean height (Sa): The Sa represents the average
used to separate the roughness (high frequency/microroughness) height evaluated over the defined area. The Sa and Sq represent
and waviness (low frequency/macroroughness) of the surface an overall measure of the surface texture. However, they are
[6,20]. The cut-off length of the filter was classically chosen at insensitive in differentiating peaks, valleys and the spacing of
0.8 mm based on ISO 4288 [21], to be representative of the surface the various texture features.
considering the 0.5 mm wavelength border between micro- and
Peak material volume (Vmp): The Vmp is the volume of material
macro-texture [22]. Surface texture parameters were then calcu- comprising the surface from the height corresponding to a 10%
lated following ISO 25178-2 [23]. Examples of some of the opera- material-bearing ratio (MBR) level to the highest peak. It repre-
tions carried out from post processing to surface texture sents the volume of material that may be worn away and char-
calculations are shown in Figs. 2 and 3. acterizes the contact zone.
Six parameters were considered for surface texture characteri-
Core material volume (Vmc): The Vmc is the volume of material
zation. Two parameters for each of height (root-mean-square comprising the texture between heights corresponding to 10%
height and arithmetic mean height), volume (peak material vol- and 80% MBR. It represents the volume of material available
ume and core material volume) and feature (density of peaks and for load support after the top levels of the surface wears away.
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 233

Table 2
Typical surfaces (point clouds) and their texture parameters in the SN lane.

S Representative surfaces for scenarios in the SN lane Surface texture parameters

First third Middle third Final third TP First Middle Final
1. Surface after form removal Sq 0.575 0.818 0.681
Sa 0.417 0.564 0.485
Vmp 0.027 0.056 0.057
Vmc 0.432 0.561 0.492
Spd 0.005 0.002 0.004
Spc 4.42 7.64 5.09

2a. Roughness surface Sq 0.066 0.111 0.079

Sa 0.038 0.060 0.041
Vmp 0.005 0.008 0.008
Vmc 0.033 0.053 0.032
Spd 0.024 0.015 0.032
Spc 8.58 13.7 10

2b. Waviness surface Sq 0.567 0.805 0.673

Sa 0.409 0.553 0.478
Vmp 0.027 0.055 0.056
Vmc 0.423 0.55 0.484
Spd 0.003 0.002 0.002
Spc 1.14 1.65 1.3

3. Top 1 mm Sq 2  104 8  105 8  105

0.5% MBR TH Sa 1  104 1  105 2  105
Vmp 0.026 0.007 0.008
Vmc 0.043 6  105 9  106
Spd 0.021 0.0002 0.001
Spc 3.58 5.55 4.2

4. Top 2 mm Sq 0.449 0.172 0.209

0.5% MBR TH Sa 0.362 0.040 0.071
Vmp 0.026 0.020 0.036
Vmc 0.434 4  105 1  104
Spd 0.020 0.003 0.004
Spc 4.22 5.56 5.12

Note: S: Scenario, SN: South-North, MBR: Material bearing ratio, TH: Thresholding, Sq: Root mean square height (mm), TP: Texture parameter, Sa: Arithmetic mean height
(mm), Vmp: Peak material volume at 10% MBR (mm3/mm2), Vmc: Core material volume (mm3/mm2), Spd: Density of peaks (1/mm2), Spc: Arithmetic mean peak curvature
(1/mm2).

Peak density (Spd): The Spd is the number of peaks per unit area (Fig. 4a). Its application affected only 0.5% of the point cloud data.
making up the surface. It affects the manner in which the peaks The 1-mm or 2-mm height threshold was applied thereafter using
undergo elastic and plastic deformation under load. the corresponding ‘‘Slices threshold” to ascertain the projected

Arithmetic mean peak curvature (Spc): The Spc is the mean peak area involved (Fig. 4b). The choice of the top 1–2 mm of the surface
curvature for the various peak structures. It affects the degree of was based on the hypothesis that the GripTester tire envelops
elastic and plastic deformation of a surface under loading. pavement surface particles up to 2 mm in depth.

Surface texture parameters were determined for four scenarios,

namely, the: 2.4. Statistical analysis

(1) Surface (point cloud) after form removal. Simple descriptive statistics such as mean and standard devia-

(2a) Roughness and (2b) waviness surfaces mentioned previ- tion were employed for data presentation. Multiple linear regres-
ously from scenario 1. sion analysis was also carried out to determine surface texture

(3) Top 1 mm after 0.5% MBR removal from the top of the sur- parameters that significantly influence pavement friction using
face in scenario 1. the stepwise variable selection method [25]. The regression

(4) Top 2 mm after 0.5% MBR removal from the top of the sur- method adds variables with the largest contribution to the model
face in scenario 1. based on the Student’s t statistic and the 0.05 significance level.
It also uses the Fisher’s F test for the null hypothesis that there is
Scenarios (3) and (4) involved applying the thresholding opera- no effect of the explanatory variables for each model at the 0.05
tor with respect to material ratio and height to extract the thresh- significance level. In addition, the coefficient of determination
olded zone for surface texture analysis. The 0.5% MBR threshold (R2) was used to compare different models. The statistical analyses
based on depth from the highest point was first applied on a given were carried out using XLSTAT v2017.3 software (Addinsoft, New
surface after form removal to eliminate any extreme elevations York, USA).
234 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

Fig. 5. Texture parameters (a) Sq, (b) Sa, (c) Vmp, (d) Vmc, (e) Spd and (f) Spc of the pavement surface after form removal (Scenario 1).

3. Results and discussion Fig. 1) probably due to slight differences in the gabbro aggregates
used as was noticed from visual inspection of the aggregate parti-
3.1. Relationships between texture parameters and grip number cles. Based on visual observations, fine- and coarse-grained gabbro
were probably used for the SN and NS direction lanes, respectively.
Typical surfaces in the first, middle and final third of both lanes Differences in the traffic and loads of the trucks (e.g. half and fully
of the test pavement and their texture parameters are shown in loaded) in the NS and SN directions may also influence the polish-
Tables 1 and 2, respectively. Figs. 5–8 show the texture parameters ing level.
for all twelve points measured in each of both lanes and their rela- Tables 3 and 4 show the exact locations of the 12 texture and
tionship to friction (grip number) for scenarios 1–4, respectively. In friction measurements for the NS and SN lanes, respectively. The
the figures, the correlation coefficient (R) is shown for each lane average grip number for the 12 locations in the NS lane was 1.5
and for all data points from both lanes since both lanes have differ- times higher than in the SN lane (Tables 3 and 4). The NS lane with
ent textures. The SN lane is more polished than the NS lane (see higher grip numbers also generally showed higher individual tex-
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 235

Fig. 6. Selected texture parameters for pavement surface roughness, namely, (a) Sq, (c) Vmp and (e) Spd, and pavement surface waviness, namely, (b) Sq, (d) Vmp and (f) Spd
(Scenario 2).

ture parameters across different scenarios (Tables 1–4). This is values of Spd and Spc would cause increase in contact stresses,
ascribed to the aforementioned difference in the polishing level which leads to greater deformation and hysteretic stress [6]. High
of both lanes. The height and volume parameters were roughly pressures are set up on sharp edges in the pavement surface, which
1.1 times higher in the NS lane than the SN lane, while the feature assists in breaking through the lubricating water film between the
parameters were 1.6 times higher. The ratio of the feature param- tire and the pavement surface [26]. Thus, the presence of sharp
eters in the NS and SN lanes is very similar to the ratio of the aver- peaks on the pavement surface would largely influence the skid
age grip numbers for both lanes. This shows the importance of resistance. The only major exception to the above was in Scenario
these parameters (Spd and Spc) to GripTester-measured friction, 2a for the roughness surface (or microtexture). The height and vol-
which agrees with observations made using the DFT [6]. Higher ume parameters were 1.3 times higher in the NS lane than the SN
236 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

Fig. 7. Texture parameters (a) Sq, (b) Sa, (c) Vmp, (d) Vmc, (e) Spd and (f) Spc of the top 1 mm of the pavement surface after 0.5% MBR removal (Scenario 3).

lane, while the feature parameters were 1.1 times higher in that parameters indicated physically meaningful but weak positive cor-
scenario. This implies a better microtexture in the NS lane than relations for each of, and both lanes after application of the 0.5%
SN lane and similar macrotexture, in terms of height and volume MBR and (1–2 mm) height thresholds (Scenarios 3 and 4). With a
parameters. However, the macrotexture (waviness surface) was few exceptions, all six parameters also showed positive correla-
quite better in the NS lane than the SN lane in terms of feature tions with the grip number, especially with the top 2 mm of the
parameters. surface (Figs. 7 and 8). Strong correlations (R = 0.74–0.78) were
The surface after form removal (scenario 1, Fig. 5) showed neg- observed for the Vmp and Spd for the NS and SN lanes with the
ative or very weak positive correlations with grip number for the top 2 mm (Fig. 8). The Vmp showed a moderate positive (R =
height and volume parameters. While weak positive correlations 0.557) correlation for both lanes. This is similar to the findings of
were generally observed for the feature parameters. The correla- a previous study, which identified the Spd and Vmc as the most
tion trend in Scenario 2 is similar to that of Scenario 1. The height influential parameters for DFT friction [13]. The presence of strong
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 237

Fig. 8. Texture parameters (a) Sq, (b) Sa, (c) Vmp, (d) Vmc, (e) Spd and (f) Spc of the top 2 mm of the pavement surface after 0.5% MBR removal (Scenario 4).

correlations after surface thresholding suggests that further pro- rounded or polished projections, regardless of their size or the load
cessing of point cloud data to remove extreme/superfluous heights applied to them [26].
is required to obtain meaningful texture-friction relationships. The
absence of strong correlations for parameters other than the Vmp 3.2. Stepwise multiple linear regression analysis
and Spd is probably because the parameters are insensitive in dif-
ferentiating sharp peaks/projections on the pavement surface. The The stepwise regression for Scenarios 1–3 generally did not
friction coefficient under wet conditions is closely related to the generate meaningful models for the texture-friction relationship.
pressure over the contact area between tire rubber and individual For instance, models with texture parameters (e.g. Sq) that showed
coarse aggregate projections in the pavement surface. The neces- negative correlation with grip number are not physically meaning-
sary pressures for skid resistance are unlikely to be set up on ful (see Figs. 5–7 and Table 5). While the models with Spd for all
238 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

Table 3
Locations of the individual texture and friction measurements in the NS lane.

S TP Distance from start point in the NS direction (m)

43 74 162 208 302 388 476 544 612 687 778 835

1. Surface after form removal Sq 0.687 0.572 0.589 0.809 0.840 1.120 0.731 0.926 1.190 0.700 0.666 0.642
Sa 0.506 0.431 0.432 0.612 0.594 0.824 0.517 0.652 0.873 0.496 0.491 0.462
Vmp 0.031 0.044 0.051 0.028 0.028 0.080 0.051 0.090 0.095 0.035 0.030 0.026
Vmc 0.529 0.449 0.437 0.668 0.596 0.866 0.532 0.641 0.876 0.503 0.514 0.479
Spd 0.003 0.007 0.011 0.002 0.004 0.003 0.003 0.007 0.006 0.005 0.006 0.006
Spc 6.64 6.27 13.40 3.39 5.16 11.90 11.60 22.60 24.00 6.66 6.46 5.24
2a. Roughness surface Sq 0.106 0.077 0.079 0.098 0.089 0.181 0.126 0.163 0.228 0.101 0.109 0.096
Sa 0.062 0.041 0.044 0.051 0.047 0.106 0.057 0.081 0.111 0.054 0.058 0.049
Vmp 0.007 0.007 0.007 0.009 0.007 0.015 0.010 0.014 0.020 0.008 0.009 0.008
Vmc 0.056 0.036 0.035 0.038 0.038 0.093 0.042 0.065 0.088 0.045 0.045 0.036
Spd 0.011 0.014 0.022 0.019 0.011 0.012 0.005 0.009 0.011 0.024 0.022 0.032
Spc 13.50 10.00 18.90 9.43 15.40 23.10 20.50 43.60 44.20 12.90 13.20 11.00
2b. Waviness surface Sq 0.667 0.560 0.574 0.805 0.831 1.110 0.716 0.899 1.150 0.685 0.657 0.634
Sa 0.490 0.422 0.420 0.609 0.586 0.818 0.506 0.630 0.848 0.483 0.483 0.456
Vmp 0.030 0.043 0.051 0.028 0.027 0.081 0.049 0.089 0.093 0.035 0.030 0.026
Vmc 0.510 0.438 0.423 0.665 0.587 0.863 0.522 0.616 0.846 0.487 0.503 0.471
Spd 0.002 0.004 0.005 0.002 0.002 0.001 0.001 0.003 0.003 0.003 0.003 0.003
Spc 1.34 1.50 1.58 1.13 1.11 6.03 4.78 3.30 3.09 1.39 1.43 1.28
3. Top 1 mm Sq (x104) 1.450 1.110 0.950 1.620 1.570 0.900 1.790 0.842 0.842 1.210 1.530 1.680
0.5% MBR TH Sa (x104) 0.730 0.373 0.253 0.929 0.888 0.198 1.180 0.157 0.170 0.497 0.821 1.010
Vmp 0.030 0.020 0.013 0.028 0.027 0.010 0.023 0.008 0.009 0.027 0.029 0.026
Vmc (x104) 113 0.375 0.212 276 260 0.118 544 0.082 0.099 0.625 190 366
Spd 0.025 0.008 0.009 0.009 0.026 4  104 0.031 0.003 0.003 0.016 0.019 0.023
Spc 5.04 5.08 12.7 2.73 3.96 5.54 3.25 17.6 15.4 4.87 4.43 3.96
4. Top 2 mm Sq 0.456 0.364 0.296 0.492 0.475 0.224 0.471 0.194 0.209 0.410 0.459 0.464
0.5% MBR TH Sa 0.375 0.259 0.172 0.414 0.398 0.072 0.387 0.051 0.064 0.326 0.377 0.380
Vmp 0.030 0.042 0.048 0.042 0.027 0.038 0.047 0.032 0.032 0.033 0.029 0.043
Vmc 0.422 0.161 0.060 0.466 0.510 0.000 0.491 4  105 6  105 0.323 0.489 0.481
Spd 0.026 0.028 0.033 0.013 0.021 0.028 0.032 0.007 0.007 0.031 0.017 0.032
Spc 6.18 5.04 11.5 3.34 4.79 9.55 3.96 19.5 19.7 5.84 5.59 4.86
Grip number 0.34 0.38 0.38 0.34 0.29 0.32 0.34 0.29 0.27 0.34 0.28 0.36

Note: S: Scenario, NS: South-North, MBR: Material bearing ratio, TH: Threshold, Sq: Root mean square height (mm), TP: Texture parameter, Sa: Arithmetic mean height (mm),
Vmp: Peak material volume at 10% MBR (mm3/mm2), Vmc: Core material volume (mm3/mm2), Spd: Density of peaks (1/mm2), Spc: Arithmetic mean peak curvature (1/mm2).
The texture parameters are approximated values so may result in slightly different R-values from those in Figs. 5–8.

Table 4
Locations of the individual texture and friction measurements in the SN Lane.

S TP Distance from start point in the NS direction (m)

43 74 162 208 302 388 476 544 612 687 778 835

1. Surface after form removal Sq 0.737 0.729 0.575 0.606 0.763 0.616 0.818 0.890 0.876 0.664 0.862 0.681
Sa 0.515 0.509 0.417 0.450 0.545 0.460 0.564 0.629 0.641 0.499 0.626 0.485
Vmp 0.040 0.037 0.027 0.035 0.033 0.035 0.056 0.050 0.043 0.039 0.049 0.057
Vmc 0.511 0.504 0.432 0.466 0.554 0.483 0.561 0.636 0.657 0.522 0.633 0.492
Spd 0.003 0.002 0.005 0.003 0.003 0.005 0.002 0.002 0.004 0.003 0.003 0.003
Spc 3.84 3.79 4.42 3.99 5.33 5.94 7.64 6.88 15.10 5.02 8.03 5.09
2a. Roughness surface Sq 0.078 0.080 0.066 0.073 0.115 0.061 0.111 0.100 0.102 0.134 0.105 0.079
Sa 0.035 0.042 0.038 0.039 0.064 0.035 0.060 0.055 0.047 0.085 0.058 0.041
Vmp 0.005 0.006 0.005 0.005 0.008 0.005 0.008 0.007 0.007 0.010 0.007 0.007
Vmc 0.025 0.034 0.033 0.033 0.057 0.030 0.053 0.048 0.036 0.080 0.050 0.032
Spd 0.003 0.017 0.024 0.006 0.008 0.041 0.015 0.006 0.010 0.025 0.004 0.032
Spc 16.70 10.50 8.58 11.00 15.80 8.07 13.70 18.80 35.40 6.67 27.90 10.00
2b. Waviness surface Sq 0.728 0.722 0.567 0.599 0.752 0.607 0.805 0.878 0.831 0.652 0.847 0.673
Sa 0.508 0.502 0.409 0.444 0.536 0.452 0.553 0.619 0.597 0.489 0.614 0.478
Vmp 0.039 0.037 0.027 0.035 0.033 0.034 0.055 0.050 0.045 0.039 0.049 0.056
Vmc 0.503 0.495 0.423 0.459 0.542 0.473 0.550 0.624 0.607 0.512 0.620 0.484
Spd 0.002 0.001 0.003 0.002 0.002 0.003 0.002 0.002 0.003 0.001 0.002 0.002
Spc 0.83 0.74 1.14 0.97 1.26 1.03 1.65 1.50 1.30 1.78 1.37 1.30
3. Top 1 mm Sq (x104) 1.440 1.430 1.740 1.420 1.530 1.470 0.772 1.140 1.340 1.390 1.190 0.825
0.5% MBR TH Sa (x104) 0.662 0.683 1.090 0.678 0.789 0.718 0.130 0.389 0.552 0.615 0.433 0.160
Vmp 0.036 0.034 0.026 0.034 0.032 0.034 0.007 0.020 0.030 0.034 0.023 0.008
Vmc (x104) 0.728 44.60 426 39.50 133 62.80 0.057 0.352 0.547 0.692 0.411 0.091
Spd 0.006 0.012 0.021 0.008 0.015 0.017 2  104 0.007 0.010 0.005 0.008 0.001
Spc 2.99 3.50 3.58 3.11 4.20 4.69 5.55 4.92 7.91 3.38 6.40 4.20
4. Top 2 mm Sq 0.437 0.434 0.449 0.437 0.465 0.445 0.172 0.357 0.413 0.433 0.374 0.209
0.5% MBR TH Sa 0.338 0.347 0.362 0.350 0.380 0.357 0.040 0.233 0.301 0.339 0.251 0.071
Vmp 0.039 0.035 0.026 0.034 0.032 0.034 0.020 0.035 0.044 0.038 0.028 0.036
Vmc 0.296 0.334 0.434 0.398 0.406 0.405 4  105 0.107 0.187 0.289 0.122 1  104
Spd 0.014 0.019 0.020 0.018 0.011 0.013 0.003 0.020 0.024 0.014 0.005 0.004
Spc 3.41 4.06 4.22 3.75 4.98 5.45 5.56 5.90 9.73 3.92 7.85 5.12
Grip number 0.23 0.21 0.21 0.23 0.21 0.22 0.2 0.22 0.25 0.23 0.19 0.2

Note: S: Scenario, SN: South-North, MBR: Material bearing ratio, TH: Threshold, Sq: Root mean square height (mm), TP: Texture parameter, Sa: Arithmetic mean height (mm),
Vmp: Peak material volume at 10% MBR (mm3/mm2), Vmc: Core material volume (mm3/mm2), Spd: Density of peaks (1/mm2), Spc: Arithmetic mean peak curvature (1/mm2).
The texture parameters are approximated values so may result in slightly different R-values from those in Figs. 5–8. The locations in the SN lane are in terms of distance from
the start of the pavement sections in the NS direction.
 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240 239

Table 5
Stepwise linear regression analysis.

Scenario Lane Variable retained/Source Value Pr > |t| R2

1. Surface after form removal NS Intercept 0.429 <0.0001 0.463
Sq 0.129 0.015
SN – – – –
All Intercept 0.202 <0.0001 0.311
Spd 16.61 0.005
2a. Roughness surface NS Intercept 0.391 <0.0001 0.419
Sq 0.528 0.023
SN and All – – – –
2b. Waviness surface NS Intercept 0.429 <0.0001 0.457
Sq 0.131 0.016
SN – – – –
All Intercept 0.194 <0.0001 0.534
Vmp 2.023 0.012
Spd 38.93 0.001
Spc 0.042 0.001
3. Top 1 mm of surface after 0.5% MBR removal SN Intercept 0.191 <0.0001 0.341
Vmp 0.959 0.046
NS and All – – –
4. Top 2 mm of surface after 0.5% MBR removal NS Intercept 0.187 0.000 0.759
Vmp 2.656 0.025
Spd 1.834 0.036
SN Intercept 0.155 <0.0001 0.750
Vmp 1.404 0.025
Spd 1.092 0.047
All Intercept 0.148 <0.0001 0.615
Spd 5.212 <0.0001
Spc 2.101 0.048

Note: NS: South-North, SN: South-North, MBR: Material bearing ratio, Sq: Root mean square height (mm), Vmp: Peak material volume at 10% MBR), Spd: Density of peaks,
Spc: Arithmetic mean peak curvature.
Pr > |t|: Significance probability, R2: Coefficient of determination.

data points for the surface without form, and Vmp for the SN lane pavement friction measured using a GripTester. The results show
for the top 1 mm of the surface had low R2 (0.34) values. Simi- that the lane with higher grip numbers generally had higher indi-
larly, the variables, Vmp, Spd and Spc, in the model for all data vidual texture (especially feature) parameters. Nevertheless, no
points for the waviness surface negatively or weakly correlated meaningful correlations were obtained using CRP 3D models of
with the grip number; resulting in a relatively low R2 (0.53) the pavement surface after form removal, and its separation into
(Fig. 6d, f and Table 5). Hence, Scenarios 1–3 did not yield any rea- microtexture and macrotexture. The same applies to using the
sonable model. top 1 mm of the surface after 0.5% MBR thresholding. Therefore,
In contrast, the top 2 mm of the surface (Scenario 4) had reason- the entire region described by the CRP model of the surface after
able models incorporating the Spd and Vmp for each of both lanes form removal and its separation into microtexture and macrotex-
(Table 5). However, the Vmp is not significant when all data points ture does not directly explain tire-pavement interaction. However,
from both lanes are considered. The Spc is retained in the model for the use of the top 2 mm of the pavement surface after 0.5% MBR
all data points (alongside the Spd, Table 5) due to its p-value. Nev- thresholding led to meaningful texture-friction correlations. It is
ertheless, it weakly correlated (R2 = 0.02, Fig. 8f) with the grip hypothesized that this is because the area described by the top 2
number and was not applicable for each of both lanes. As men- mm of the surface actually interacts with the GripTester-
tioned previously, both lanes of the pavement had slightly different measuring tire. In contrast, the region covered by the top 1 mm
textures and levels of polishing. Hence, the Vmp at 10% MBR (i.e. insufficiently characterizes tire-pavement contact. Stepwise
the top 10%) would be a lot more sensitive to small changes on regression indicated that the texture parameters, Spd and Vmp,
the pavement surface than the Spd. This probably accounts for show the best correlation (R2 = 0.75–0.76) with the grip number,
the Spd having a higher correlation for all data points from both but the Spd has more influence than the Vmp. Thus, the density
lanes and significantly explaining the variability of the grip num- of peaks on the pavement surface and the volume of peak material
ber than the Vmp. The foregoing implies that the Spd and the should influence asphalt mixture design for pavement
Vmp show the best correlations with the grip number, but the construction.
Spd is more influential. Therefore, from the parameters in Table 5, It is recommended that future studies investigate different
the following are the acceptable regression equations: pavement sections with different types of aggregates and asphalt
mixtures. Further work may also investigate the texture-friction
GN ¼ 0:187 þ 2:656Vmp þ 1:834SPd ð1Þ
relationship(s) studied here using different grip tester speeds, since
the test speed affects the grip number.
GN ¼ 0:155 þ 1:404Vmp þ 1:092SPd ð2Þ

Where, GN. the grip number, while Vmp. d Spd. e as defined previ-
ously. Eqs. (1) and (2) are applicable for the NS and SN lanes, Acknowledgements
respectively.
This publication was made possible by an NPRP award (NPRP
4. Conclusions and recommendations No. 7-482-2-184: Thermo-Mechanical Tire-Pavement Interaction:
Computational Modelling and Field Measurements) from the Qatar
This study evaluated the relationship between texture parame- National Research Fund (QNRF – a member of The Qatar Founda-
ters derived from CRP measurements using a handheld camera and tion). The statements made herein are solely the responsibility of
240 R.B. Kogbara et al. / Construction and Building Materials 166 (2018) 227–240

the authors. The authors are grateful to Applus Velosi Qatar for [13] Q.J. Li, G. Yang, K.C.P. Wang, Y.J. Zhan, C. Wang, Three-dimensional-based areal
texture parameters for pavement friction, Transportation Research Board 96th
their assistance with the GripTester measurements.
Annual Meeting, Paper No. 17-05098, 2017.
[14] M. Mahboob Kanafi, A. Kuosmanen, T.K. Pellinen, A.J. Tuononen, Macro-and
References micro-texture evolution of road pavements and correlation with friction, Int. J.
Pavement Eng. 16 (2) (2015) 168–179.
[1] J.W. Hall, K.L. Smith, L. Titus-Glover, L.D. Evans, J.C. Wambold, T.J. Yager, et al., [15] A. Shalaby, A. El Gendy, 3D pavement surface macrotexture: Measurements
Guide for pavement friction. Contractor’s Final Report for National Cooperative and friction relationships, 7th Symposium on Pavement Surface
Highway Research Program (NCHRP) Project 01-43, Transportation Research Characteristics (SURF 2012), Norfolk, VA, 2012.
Board of the National Academies, Washington, D.C. Available: http:// [16] H. Sadek, E. Masad, O. Sirin, H. Al-Khalid, K. Hassan, Performance evaluation of
onlinepubs.trb.org/onlinepubs/nchrp/nchrp_w108.pdf (acesssed June 2015), full-scale sections of asphalt pavements in the State of Qatar, J. Perform.
2009. Constr. Facil 29 (5) (2015) 04014123.
[2] PIARC. Report of the committee on surface characteristics. Permanent [17] ASTM E1844-08. Standard specification for A Size 10 x 4-5 smooth-tread
International Association of Road Congress (PIARC) XVIII World Road friction test tire. West Conshohocken, PA: American Society for Testing and
Congress, Brussels, 1987. Materials (ASTM) International, 2015.
[3] R.B. Kogbara, E.A. Masad, E. Kassem, A. Scarpas, K. Anupam, A state-of-the-art [18] C. Walsh, The UK Pavement Management System (UKPMS) user manual,
review of parameters influencing measurement and modeling of skid Volume 3: Machine data collection for UKPMS, Chapter 11: GripTester.
resistance of asphalt pavements, Constr. Build. Mater. 114 (2016) 602–617. Available: http://www.ukroadsliaisongroup.org/download.cfm/docid/
[4] G. Descornet, Criterion for optimizing surface characteristics, Transp. Res. Rec. 6FC2D12A-93EE-4DE6-B2C3879F57EF918F (accessed September 2017), 2005.
1215 (1989) 173–177. [19] G. McQuaid, P. Millar, D. Woodward, Use of 3D modeling to assess pothole
[5] G.W. Flintsch, K.K. McGhee, E. de León Izeppi, S. Najafi. The little book of tire growth, in: A.F. Nikolaides (Ed.), 6th International Conference on Bituminous
pavement friction. Surface Properties Consortium. Available from: https:// Mixtures and Pavements, Thessaloniki, Greece, Taylor & Francis Group,
secure.hosting.vt.edu/www.apps.vtti.vt.edu/1-pagers/CSTI_Flintsch/The%20Little London, 2015, pp. 161–166.
%20Book%20of%20Tire%20Pavement%20Friction.pdf [Accessed October 2015]. [20] A. Dunford, Road surfaces, in: R. Leach (Ed.), Characterisation of areal surface
2012. texture Berlin, Heidelberg, Springer, Berlin Heidelberg, 2013, pp. 337–348.
[6] L. Hu, D. Yun, Z. Liu, S. Du, Z. Zhang, Y. Bao, Effect of three-dimensional [21] ISO 4288. Geometrical Product Specification (GPS) – Surface texture: Profile
macrotexture characteristics on dynamic frictional coefficient of asphalt method – Rules and procedures for the assessment of surface texture. Geneva,
pavement surface, Constr. Build. Mater. 126 (2016) 720–729. Switzerland: International Organization for Standardization, 1996.
[7] G. McQuaid, P. Millar, D. Woodward, S. Friel, Use of close range [22] L. Courard, A. Garbacz, B. Bissonnette, Topography evaluation methods for
photogrammetry to assess the micro-texture of asphalt surfacing aggregate. concrete substrates: parametric study, in: Third International Conference on
São Paulo, Brazil, Int. J. Pavements Conf. (2013). Concrete Repair, Rehabilitation and Retrofitting, ICCRRR-3, 3–5 September,
[8] T. Luhmann, S. Robson, S. Kyle, J. Boehm, Close-range photogrammetry and 3D Cape Town, South Africa, CRC Press, 2012, p. 379.
imaging, Second ed., De Gruyter, Berlin, Germany, 2013. [23] ISO. ISO 25178-2: Geometrical product specifications (GPS) – Surface texture:
[9] F.G. Praticò, R. Vaiana, A study on the relationship between mean texture Areal – Part 2: Terms, definitions and surface texture parameters. Geneva,
depth and mean profile depth of asphalt pavements, Constr. Build. Mater 101 Switzerland: International Organization for Standardization, 2012.
(Part 1) (2015) 72–79. [24] Michigan Metrology. Surface texture parameters glossary. Michigan
[10] S. Torbruegge, B. Wies, Characterization of pavement texture by means of Metrology, LLC, Livonia, MI. Available: https://www.michmet.com/
height difference correlation and relation to wet skid resistance, J. Traffic Texture_parameters.htm, 2014.
Transp. Eng. (English Ed.). 2 (2) (2015) 59–67. [25] F.E. Harrell, Regression Modelling Strategies: With Applications to Linear
[11] Y.H. Miao, D.W. Cao, Q.Q. Liu, Relationship between surface macrotexture and Models, Logistic Regression, and Survival Analysis, Springer-Verlag, New York,
skid resistance of asphalt pavement, J. Beijing Univ. Technol. 4 (37) (2011) 2001.
547–553. [26] B.E. Sabey, Pressure distributions beneath spherical and conical shapes
[12] M. Kane, Z. Rado, A. Timmons, Exploring the texture–friction relationship: pressed into a rubber plane, and their bearing on coefficients of friction
from texture empirical decomposition to pavement friction, Int. J. Pavement under wet conditions, Proc. Phys. Soc. 71 (6) (1958) 979–988.
Eng. 16 (10) (2015) 919–928.