Road surface condition evaluation is a critical component for the operation and maintenance of road infrastructure. The situation is more heightened for unpaved roads as they lack protective hard surfaces such as concrete or asphalt. New methods are now emerging for the identification and extraction of road surface condition parameters from remotely sensed imagery. Accordingly, this study illustrates an approach to identify and extract road surface condition parameters from remotely sensed imageries obtained from unpaved roads with unique environmental characteristics. An Unmanned Aerial Vehicle (UAV) was used to capture images on unpaved roads where they were pre-processed, and analysed by building a digital elevation model (DEM) and orthomosaic. The processed imagery was validated with a conventional field measurement of road defects. The number of road defects identified and their severity was similar for both approaches. The defects counts were similar for all except one defect which vary for both approaches. The predictions of the extent of defect severity identified as percent count were exact – vegetation encroachment had the highest 100% and depression had the lowest of 25%. These findings illustrate the potential use of remotely sensed imagery from UAV to identify, extract, and measure road surface condition parameters for unpaved roads. The results can facilitate the monitoring operation of road agencies in the country towards low-cost and effective maintenance of roads.
Short Report
Mapping of Road Surface Condition features for Unpaved Roads Through the use of Remotely Sensed Imagery from Unmanned Aerial Vehicle
https://doi.org/10.21203/rs.3.rs-2623642/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
Road maintenance
Road surface condition monitoring
Unpaved roads
Remote sensing
UAV
Image processing.
Since the 1900s, road surface inspection and surveillance (road inventory) had relied on the visual inspection method [1], [2] to collect road surface data. This method necessitates riding at about 40km/h or walking while conducting the inventory. The exercise is tedious, time-consuming, ineffective and sometimes fatigue can cause individuals to misrepresent results from the survey [3]–[6]. The advancement of technology and software can help modernize the way through which road surface data could be assessed. Image processing and analysis may be applied to roads, just like any other surface, to aid in feature extraction. The extraction of road surface features is significant because it provides a wide range of data and information that is utilized to manage roads and their accessories including unpaved roads. Unpaved roads are those roads with surfaces made from native earth material or the addition of material-aggregate (for stability), and the road surfaces are neither cemented, asphalted, nor have any surface treatment [6]. Unpaved road maintenance is very essential in effectively assessing the conditions and states at which the road is deteriorating [7]. Careful and timely interventions of any pavement defects associated with the road will reduce the cost of repair, prevent vehicle damage and save lives in the long run [8]. Unpaved road defect inspection and surveillance is a critical aspect of effective road maintenance as this enhances the smooth vehicular flow and reduces traffic and frustration. So far, there is no specific agreed-on technique for extracting road surface parameters from images. The exploitation of how effectively one can identify, extract and measure unpaved road surface defects are still being prospected. This work seeks to add to the existing knowledge on how to extract unpaved road surface features from remotely sensed images through the use of Unmanned Aerial Vehicle (UAV).
A road condition survey, similar to any geographic feature extraction, can be accessed through one of the many ways of accessing any geographic information. For understanding and evaluating these data, engineers from around the world had developed an intrusive and non-intrusive way of assessing the condition of roads. Using images for unpaved road condition surveys is one of the non-intrusive ways [6] of assessment. There are some other methods of gathering information from the road surface and these involve the use of sophisticated gadgets, equipment, devices, machine learning programs and algorithms. Examples of this can be found in Sattar et al. (2018) and Zang et al. (2018) when they used smartphones for road condition surveys and inventory. Some of these commercial smartphone applications use an automated threshold-based method – found by statistical values such as standard deviation, variance, and zero-crossing rate [9], to gather information from the road surface, and they usually measure road roughness [6]. There are also vision-based applications for road surface condition monitoring and inventories. These systems use camera(s) [11] mounted on vehicles and other devices ([11]; [6]; [7], [12] to capture videos that are used to assess the condition of roads. The vision-based approach, rely solely on video processing analysis such as texture extraction, algorithms [13], comparison of captured images, photogrammetry analysing programs [14], and other machine learning approach (Saeed et al., 2020) to depict road surface distresses. In Canada, an Automated Road Analyser (ARAN), that records and analyzes road roughness, geometric profile and other parameters are used for road surface condition inventories and monitoring. This ARAN is an automated vehicle with both software and hardware and does robust data analysis while using the road. Many such gadgets are available and used in different jurisdictions.
The approach to identifying and extracting unpaved road surface parameters was scheduled under a five-phased approach. These approaches are reconnaissance study, planning, data collection, data processing and data analysis. The reconnaissance study dealt with a desk study on the area of interest, considering soil records, online literature, the socio-economic demography of the area, and views of commuters on the road network of the selected study area (from road experts at the municipal office). The planning phase helped with the decision on which software, equipment and camera extrinsic and intrinsic parameters to be used. Data collection dealt with mobilising tools and equipment to the site to access the raw data. This was done for the imagery approach and conventional field approach. A UAV was used to collect data for the imagery approach in conjunction with other measuring instruments. Data processing focused on the use of Pix4D and Argisoft Photoscan to process the data into 3D and Global Mapper for road surface feature extraction. The processing steps include adding images, aligning images, building a dense point cloud, building mesh, building a tiled model, building a digital elevation model (DEM), and orthomosaic. Data analysis deals with extracting the road surface parameters from the modelled image space and translating its meaning from the extracted modelled data. To extract the defects, the point cloud, DEM and orthomosaic files were exported into the Global mapper software. Profile, sectional analysis, volume models and dimensional analysis were used to extract desired parameters. The tools and equipment needed for the data collection and processing include but are not limited to a UAV, meter wheel, spray paint, measuring devices, and safety clothing among others.
The imagery approach work required flying the unmanned aerial vehicle at varying altitudes of 15 to 35 meters above the ground on four unpaved roads. The images captured from individual roads were processed into 3D with the Pix4D mapper and Argisoft Photoscan software. The modelled 3D images were exported into the Global mapper for road surface feature extraction and hence, measurement. The conventional field method approach was conducted concurrently with the imagery method for validation. This required walking the entire stretch of the road while conducting inventory through physical measurements of various road defects. This was done on one out of the four roads for validation purposes.
2.1. Study area
The study area encompasses four selected roads in the Oforikrom Municipality in the Ashanti region of Ghana. This area was selected to facilitate ease in mobilization and also most of the unpaved roads are inhabited by fewer activities, therefore, facilitating easy and fast data collection [15]. Road 1 is an engineered road and starts from the Deduako-Bonsuom junction to Teacher Quarters and is under construction and supervised by the Oforikrom Municipal Urban Road Department. The considered length is 1km for Road 1 and 100m for the others. The other roads – Road 2 (Allied and Health Science Village, KNUST), Road 3 (Kumasi High School wall road), and Road 4 (Laud K junction road) are mixed with engineered and unenginnered roads. Only Road 1 is subject to both fieldwork and image acquisition process for validation, the remaining roads were assessed through images acquired from Unmanned Aerial Vehicle.
The Phantom 4 UAV was hired for the data collection. Although cost of hiring is expensive, it is less when compared to hiring a professional for road condition inventory. The study area was not under any UAV zonal restrictions. Although zonal restriction were not implied, high tension lines were some of the area challenges faced. This influenced an altitude of 35m for road 1 while Road 2 to 4 were bellow 18m. To extract road surface parameters from images, a secondary software - Global Mapper was employed. In this perspective, point cloud, Digital Elevation Model (DEM) and Orthomosaic from selected study areas (Road 1–4) modelled in 3D were imported for the road surface feature extraction and analysis. A point cloud is a large collection of points that are placed in a three-dimensional coordinate system [15]. DEM are the graphical 3D elevational representation of terrain and Orthomosaic represent orthorectified image where geometric distortion has been corrected and the images are colour balanced [16]. The exported files represent the X, Y, and Z geometric coordinates of a point on an underlying sampled surface. To measure and analyze the various road defects obtainable on each road, the modelled image files were imported into Global Mapper and the profile was selected. This made it possible to view the nature, profile and cross-section of the road and understand some of these defects available on the road. The profile and volume from the Global Mapper give the X, Y and Z dimensions of the defect if any apply. Figure 1 presents an example of a profile selected for Road 1, ready for parameter extraction and other analysis. The severity level from various roads analyzed gave a better understanding of the extent of deformation on the road [15]. In the Global Mapper, the imported point cloud, DEM or orthomosaic were analyzed in the software by dividing the road into sections.
The sectioning was done at 50m intervals for Road 1 and 10m for Roads 2, 3 and 4. This was done for easy defect identification, extraction, measurement and classification. The severity level demonstrates the measure and extent of defects on each road. This can be derived from the Pavement Maintenance Management Programme (PMMP) from the Ghana Highway Manual. Table 1 is a list of the defects and severity levels and Table 2 summarizes the defect types identified on each road. It is clear from Table 1 and Table 2 that unpaved road defects and severity at varying stages were identified in this work, iterating the possibilities of extracting defect parameters from images.
To identify the severity and number of defects available on each road from the images, the defects were scheduled as count and number of defects found provided in Fig. 2. This illustrates the power of identifying each defect per sample section and improves the cluster of defects identified by conventional fieldwork and imagery approaches. The findings from the work are presented in two-fold, the conventional field approach and the UAV imagery approach. The results from the analysed fieldwork indicate that the imagery approach could mimic the conventional fieldwork extracted from surface defects taken in situ [17]. The number of defects counted was the same for both conventional and imagery approaches. The per cent count represents the proportion of road length occupied by that defect.
Corrugation, rut, pothole, erosion, vegetation encroachment and depression had 10, 19, 6, 16, 20 and 5 counts detected respectively for both UAV imagery and field approaches. Inappropriate cross-sections identified differences in the count of 16 for conventional fieldwork and 17 for UAV imagery approach. This shows the slight variation in both approaches for defect count. The percentage of road length being occupied by a particular defect saw vegetation encroachment as highest at 100% and depression as lowest at 25%. Thus, at every 50m section, vegetation encroachment was identified and counted for the entire 1km road. The percentage count for depression also spanned about 250m of the road. The counts for each approach relate to Roberts et al. (2020) when distress results of the pavement model were sectioned and rated to check severity. Individually, these defect count similarities presented above indicate that imagery collected via a UAV can help identify the exact defect an unpaved road is subject to.
This study presents the use of an unmanned aerial vehicle for surface condition survey. The altitudes above which images were collected varied between 15m to 35m. The acquired images were imported and modelled (using Pix4Dmapper and Agisoft Photoscan) into a 3D reconstructed image. Point cloud, Orthomosaic and DEM were exported onto a secondary software (Global mapper) for analysis and road surface feature extraction. The extraction was done by manually measuring defects from the exported files in the model space (of Global mapper) for all roads. This manual extraction was done by dividing the 3D modelled images into sections depending on the length of the road and taking measurements as such. Regarding measuring, Road 1 was sectioned into 50m for the entire 1km and Roads 2, 3 and 4 were sectioned 10m apart for the 100m road length. All defects found during the conventional field approach could also be identified using the manual UAV imagery extraction method. All levels of severity were obtained from the work signifying the ability to extract unpaved road defects from images. From the results obtained, it could be deducted that unpaved road surface parameters such as length, width and depth of defects could be extracted from images by manually measuring and extracting the defects.
Further research should be channelled onto automatically identifying road defects from modelled 3D images.
Conflict of Interest
On behalf of all authors, the corresponding author states that there is no conflict of interest
- T. H. E. Efficiency, E. Of, C. P. To, M. Potholes, and I. N. Asphalt, “(Received: October 12, 2020; Accepted for Publication: November 22, 2020),” vol. 32, no. 2, pp. 179–197, 2020.
- E. Romero-chambi, S. Villarroel-quezada, and E. Atencio, “applied sciences Analysis of Optimal Flight Parameters of Unmanned Aerial Vehicles ( UAVs ) for Detecting Potholes in Pavements,” 2020.
- A. Alhasan, D. J. White, and K. De Brabanter, “Quantifying Roughness of Unpaved Roads by Terrestrial Laser Scanning,” Transp. Res. Rec. J. Transp. Res. Board, vol. 2523, no. 1, pp. 105–114, 2015, doi: 10.3141/2523-12.
- A. C. Chukwuocha, “Modern Trends in Topographic Data Collation for Environmental Studies and Engineering,” no. December 2012, 2018.
- S. K. Ryu, T. Kim, and Y. R. Kim, “Feature-based pothole detection in two-dimensional images,” Transp. Res. Rec., vol. 2528, pp. 9–17, 2015, doi: 10.3141/2528-02.
- N. Saeed, M. Dougherty, R. G. Nyberg, P. Rebreyend, and D. Jomaa, “A Review of Intelligent Methods for Unpaved Roads Condition Assessment,” Proc. 15th IEEE Conf. Ind. Electron. Appl. ICIEA 2020, no. May, pp. 79–84, 2020, doi: 10.1109/ICIEA48937.2020.9248317.
- C. Zhang, “An UAV-based photogrammetric mapping system for road condition assessment,” Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci., pp. 627–632, 2008, [Online]. Available: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.150.8490&rep=rep1&type=pdf
- X. Yu and E. Salari, “Pavement pothole detection and severity measurement using laser imaging,” IEEE Int. Conf. Electro Inf. Technol., 2011, doi: 10.1109/EIT.2011.5978573.
- S. Sattar, S. Li, and M. Chapman, “Road surface monitoring using smartphone sensors: A review,” Sensors, vol. 18, no. 11, p. 3845, 2018.
- K. Zang, J. Shen, H. Huang, M. Wan, and J. Shi, “Assessing and mapping of road surface roughness based on GPS and accelerometer sensors on bicycle-mounted smartphones,” Sensors, vol. 18, no. 3, p. 914, 2018.
- M. Jalayer, H. Zhou, J. Gong, and S. Hu, “A Comprehensive Assessment of Highway Inventory Data Collection Methods,” no. January 2015, 2014, doi: 10.5399/osu/jtrf.53.2.4219.
- Y. Tan and Y. Li, “UAV photogrammetry-based 3D road distress detection,” ISPRS Int. J. Geo-Information, vol. 8, no. 9, 2019, doi: 10.3390/ijgi8090409.
- C. Zhang and A. Elaksher, “An unmanned aerial vehicle-based imaging system for 3D measurement of unpaved road surface distresses,” Comput. Civ. Infrastruct. Eng., vol. 27, no. 2, pp. 118–129, 2012, doi: 10.1111/j.1467-8667.2011.00727.x.
- V. A. Knyaz and A. G. Chibunichev, “PHOTOGRAMMETRIC TECHNIQUES FOR ROAD SURFACE ANALYSIS,” vol. XLI, no. July, pp. 515–520, 2016, doi: 10.5194/isprsarchives-XLI-B5-515-2016.
- A. M. Saad and K. N. Tahar, “Identification of rut and pothole by using multirotor unmanned aerial vehicle (UAV),” Meas. J. Int. Meas. Confed., vol. 137, pp. 647–654, 2019, doi: 10.1016/j.measurement.2019.01.093.
- D. Jeong et al., “Efficient 3D Volume Reconstruction from a Point Cloud Using a Phase-Field Method,” vol. 2018, 2018.
- R. K. Rhodes, “UAS as an Inventory Tool: A Photogrammetric Approach to Volume Estimation,” Theses Diss. Sch., p. 115, 2017, [Online]. Available: http://scholarworks.uark.edu/etd/2424/