مدل طیفی مورد استفاده برای شناسایی شاخص مصالح سطوح آسفالته و تجزیه و تحلیل حساسیت

This article focuses on the spectral variability of asphalt compared with its surface characteristics, in particular, with the amount of exposed aggregate. Any uncertainties in the chemical or physical composition of the asphalt or in its alteration status may cause erroneous results. Improved understanding of the spectral properties of asphalt can be used to optimise road network management policies. Pavement aging and degradation detection is one of the primary issues in infrastructure management faced by local authorities in the area of safety standards. In this study, various asphalt samples from Central Italy are characterised by digital RGB photos combined with their spectral signatures. Because the spectral response is influenced by the presence of exposed aggregate and bitumen, it is crucial to define an objective index that could indicate either their presence or absence. Using the Exposed Aggregates Index (EAI) method on each photo, the aggregates surface occupation is determined by the supervised classification of the RGB photo dataset using the parallelepiped method. The result is then compared with a spectral response of the target. Using this process, a series of new spectral indices is identified in a range of wavelengths from 400 to 900 nm that show statistical correlation and physical significance to changes in bitumen and exposed aggregates. In particular, the first derivative of the spectrum at 400 nm and the reflectance values at 460, 490, 740 and 830 nm are very sensitive to changes in the EAI. An empirical relation for the exposed aggregates is found during the calibration step for any of the relations between the spectral index and the EAI. This relation is linked to the degradation of the targets with an RMSE of 0.09. The final phase of the work focuses on uncertainty and sensitivity analyses of the model, demonstrating the robustness of the equation identified for the relation.

Because of recent advancements in imaging spectrometry, physical and chemical properties of materials at very detailed levels can now be detected. There is evidence that road properties such as aging and material composition can affect spectral characteristics; however, the relation between the reflectance and the specific road surface quality has yet to be determined. The use of spectroradiometric data to characterise aggregates and bituminous mixtures can help improve the analysis to discriminate different types of man-made surfaces, such as paved areas. Generally, in the range of wavelengths between 350 and 2500 nm, the radiometric response of currently used asphalts is dominated by bitumen, which absorbs most of the incident solar radiation. The composition and dimensions of the aggregates have only a marginal influence on the spectral behaviour; because of the degradation processes over time, the asphalt surfaces lose bitumen, thus increase in reflectance [1] and [2]. Oxidation processes and exposure of the rocky components modify the spectral signature of the new asphalt, as exhibited by the appearance of the iron oxide absorption peaks at 520, 670 and 870 nm, while the loss of the oily compounds is seen in the disappearance of the characteristic peaks of the hydrocarbons [3] and [4]. The absorption of hydrocarbons, particularly evident in new asphalts, is seen at 1750 nm and after 2100 nm, with a significant doublet at 2310 and 2350 nm [5] and [6]. In addition, in old asphalts, the spectral signature presents a significant change of slope between 2100–2200 nm and 2250–2300 nm, from the influence of silicate minerals outcropping and hydrocarbons [7]. Simultaneously, in the VIS region, bitumen loss causes a slope change [8]. Currently, most spectroradiometric studies are focused on new road pavement management systems, to identify relations between spectral data and paved surface quality conditions. The analyses of multi- and hyper-spectral remotely sensed imagery enable the discrimination of road networks and the evaluation of the weathering of asphalt surfaces using spectral indices [9], [4] and [10]. The spectral difference between 490 and 830 nm was used by [5], who correlated the pavement condition index (PCI) and the spectral data with an R-squared of 0.63. Additionally, [11] used this difference to evaluate paved area conditions. Wavelengths at 490 nm and at 830 nm are sensitive to iron oxide absorption. The spectral difference between both bands emphasises the spectral contrast between new and aged roads, and increases the change of spectral shape in the VIS. The first-order derivative was used by [2] to distinguish 4 types of asphalt surfaces. Based on the slope change in the VIS region for new to oldest paved areas, 400 nm and 700 nm were used to categorise asphalt during laboratory measurements [11]. Viewing the effects of the presence of hydrocarbons on the spectral signatures of asphalts, there is a need to develop a method to evaluate the bitumen content in asphalt surfaces. In [8], the percentage of bitumen covering the aggregates was evaluated in the field using the geologic classification of Shvetsov [12]. Digital photos were used for descriptive data validation. Pictures taken using digital cameras are treated in different ways for different purposes. Statistical analysis of RGB values for snow cover monitoring [13], supervised classifications for automatic agriculture inspection [14], and statistics computation of achromatic and chromatic spatial patterns of colour images for indexing and content-based retrieval [15] are some examples. Elaboration techniques performed using cameras could represent the real added value to environmental monitoring when combined with remotely sensed data. Digital photos then become a useful tool at the local scale providing statistical analysis of RGB values for component quantification in the pictures. In this study, an image classification of calibrated RGB digital photos was initially performed to evaluate the Exposed Aggregates Index (EAI). To develop an equation for the relation with spectroradiometric data, different spectral indices were computed. Based on the results, a sensitivity analysis of the equation was performed. The purpose of this research is to increase the knowledge of how the variation of the surface characteristics of asphalt may affect spectral measurements. This paper is organised as follows. First an overview of the technique to identify the physical index is presented, followed by a discussion of the experimental results. Then, the cal/val analysis and the sensitivity analysis is presented.

This study demonstrated that the process chain of empirical experiences in asphalt measurements is repeatable. Moreover, it is possible to find a relation between the extraction of information of picture colourimetry and spectral in situ measurements. The results emphasised how asphalt reflectance is not constant and varies as a function of surface conditions, specifically of exposure of aggregate on the surface. One of the primary future research directions is the application of the empirical evidence to remote sensing applications. Using this procedure, an index was developed that represents the level of aggregate exposure on the surface, improving the in situ assessment of asphalt conditions with an overall accuracy of 91.4%. Statistical and simulation data support the idea of a link among classified data and spectral indices, achieving a Bravais–Pearson correlation of 0.83. The model output presented a 95%-confident normal distribution with a standard deviation of 0.095, which is comparable to the classification of aggregate by picture. This index is fundamental in the spectral characterisation of targets with different asphalt surfaces and was used to determine a regression law based on an exponential relation of reflectance at wavelengths where asphalt is very sensitive to changes (460, 490, 740, and 830 nm). Validation of the law resulted in an RMSE of 0.09; thus these values are coherent with the overall accuracy of the classification images adopted to define the EAI. A sensitivity analysis was applied to assess the weight of a single wavelength in the modelling. The results showed that the primary effects were from the near infrared (NIR) reflectance. This agrees with the spectral results of the aggregate exposed on the surface, which are rocky, and primarily composed of carbonate and basalt in the secondary phase. Usually, their reflectances are higher than bitumen both in the blue and the NIR wavelengths, providing a good separation of phases. The index used in this study is a step forward towards the comprehension of the surface characteristics of asphalt. The variation in time of the exposure aggregate index may be relevant in multi-temporal studies and may be linked to the surface decay of asphalt quality. Future research could focus on EAI extraction by satellite sensor bands whose spectral ranges are close to the studied wavelengths, thus providing information for asphalt road quality assessment in wide areas.