Data Fusion of PRISMA Satellite Imagery for Asbestos-containing

Materials: An Application on Balangero’s Mine Site (Italy)

Giuseppe Bonifazi

, Giuseppe Capobianco

, Riccardo Gasbarrone

, Silvia Serranti

Sergio Bellagamba

and Daniele Taddei

1,* f

DICMA, Department of Chemical Engineering, Materials and Environment, Sapienza - University of Rome,

via Eudossiana 18, 00184, Italy

INAIL - Italian Workers' Compensation Authority, Research Division, DIT - Department for Technological Innovations

and Security Equipment, Products and Human Settlements, via R. Ferruzzi 38/40, Rome, Italy

Keywords: PRISMA, Imaging Spectroscopy, Classification, Asbestos, Asbestos-containing Materials (ACMs).

Abstract: In the last few decades, the procedure for identifying, classifying and mapping the asbestos-containing

materials (ACMs), and contaminated areas, is considered one of the most important aspects for the purpose

of remediation. This task, carried out by skilled workers, can be very long and difficult to perform, and it can

also increase the risk of inhalation of asbestos fibers. The identification and characterization of areas

contaminated by asbestos using remote sensing techniques represent a valid alternative to census methods,

traditionally based on visual inspection of surfaces and in situ sampling to be analyzed later in the laboratory.

The aim of this work was to explore the possibilities of using machine learning techniques to identify possible

asbestos-contaminated areas and ACMs by using PRISMA satellite imagery in areas where chrysotile was

once extracted, processed and used in asbestos-containing products (ACPs). The study area is located in the

Balangero’s asbestos mine site. More in detail, Principal Component Analysis (PCA) was performed on a

Visible, Near-InfraRed and Short-Wave InfraRed (VNIR-SWIR) PRISMA image to reduce data

dimensionality and used as an exploratory analysis tool. Classification And Regression Trees (CART)

technique was finally utilized to test a classification of six predetermined classes on the panchromatic image.

1 INTRODUCTION

In many countries, asbestos contaminated areas are

still a relevant issue. Asbestos was widely used during

the 20th century thanks to its important physical and

mechanical characteristics.

There is not a group of minerals that, from a

mineralogical point of view, goes under the name

’asbestos’, but there are various mineral types that can

be distinguished based on their crystallographic and

chemical characteristics. According to the European

applicable legal references, the general term

‘asbestos’ is used to identify six naturally occurring

https://orcid.org/0000-0001-6935-2686

https://orcid.org/0000-0003-1914-7275

https://orcid.org/0000-0003-4739-1582

https://orcid.org/0000-0003-2435-5212

https://orcid.org/0000-0002-4295-0704

https://orcid.org/0000-0002-2814-8459

* Corresponding author: daniele.taddei@uniroma1.it, via

Eudossiana 18, 00184 Rome, Italy, Tel:+39-3471866206

silicate minerals belonging to the serpentine

(Chrysotile) and amphibole (Amosite, Crocidolite,

Tremolite, Anthophyllite and Actinolite). They can

be found in several different crystalline forms, but

only the ﬁbrous forms are classiﬁed as asbestos

(Council of the European Union, 2003). Based on

numerous epidemiological studies carried out since

the 1960s and proving the carcinogenic nature of

these ﬁbers, all the asbestos minerals have been

classiﬁed as carcinogens by the International Agency

for Research on Cancer (IARC) (Paglietti et al.,

2016). Many countries like Italy, have thus banned

the production, importation, processing and

distribution of Asbestos-Containing Products (ACPs)

150

Bonifazi, G., Capobianco, G., Gasbarrone, R., Serranti, S., Bellagamba, S. and Taddei, D.

Data Fusion of PRISMA Satellite Imagery for Asbestos-containing Materials: An Application on Balangero’s Mine Site (Italy).

DOI: 10.5220/0011059400003209

In Proceedings of the 2nd International Conference on Image Processing and Vision Engineering (IMPROVE 2022), pages 150-157

ISBN: 978-989-758-563-0; ISSN: 2795-4943

(i.e. roofing sealant, pipe lagging, duct tape, furnace

cement and glue for flooring, etc.) and Asbestos-

Containing Materials (ACMs) (i.e. corrugated cement

sheets, flat cement sheets, etc.) and have

recommended action plans for the mapping and safe

removal of asbestos from public and private buildings

and remediation of highly contaminated areas

(Paglietti et al., 2016). The procedure to identify,

classify and map ACMs and contaminated areas is

considered one of the most important aspects for the

purpose of remediation; this procedure performed by

skilled workers can be very long and difficult to

perform, and it can also increase the risk of inhalation

of asbestos fibers. Scientific literature reports many

studies on the utilization of remote sensing techniques

to map asbestos in anthropic and natural environment,

using different approaches (Amano et al., 2008; Chen

et al. 2012; Aplin et al. 2008; Blaschke, 2010) as

hyperspectral remote sensing (Frassy et al., 2014;

Marino et al. 2001; Massarelli et al. 2017). Several

studies have also explored the abilities to identify

ACMs by HyperSpectral Imaging (HSI) proximal

sensing in Short-Wave InfraRed (SWIR) (Bonifazi et

al., 2018; Bonifazi et al. 2019; Serranti et al. 2019).

Aim of this work was to explore the possibilities of

using machine learning techniques to identify possible

asbestos-contaminated areas and ACMs by using

PRISMA (Hyperspectral PRecursor of the Application

Mission) satellite imagery (ASI, 2009) in areas where

chrysotile was once extracted, processed, and used in

ACPs. Principal Component Analysis (PCA) was

performed on a Visible and Near-InfraRed - Short-

Wave InfraRed (VNIR-SWIR) PRISMA image to

reduce data dimensionality and used as an exploratory

analysis tool. Classification And Regression Trees

(CART) technique was used to test a classification of

six classes (i.e. ‘ACM’, ‘Urban Area’, ‘Anthropogenic

vegetation’, ‘Natural vegetation’, ‘Water’,

‘Balangero’s mine’), predetermined on the

panchromatic image. The classification was performed

on a novel dataset, where the panchromatic image was

fused with the PCA scores resulting from the

application of PCA on the VNIR-SWIR dataset. The

data fusion strategy selected was mid-level (Figure 1).

According to this procedure the features extracted from

the different blocks are concatenated to build a single

array which is then processed by the desired

chemometric technique (Biancolillo et al. 2014).

2 MATERIALS AND METHODS

The studied areas are located near Turin, in the

Piedmont region (Northern Italy), the former

Balangero’s and Corio’s asbestos mine site (Figure 2

and Figure 3).

Figure 1: Scheme of the data-fusion and classification

approach adopted.

Figure 2 shows an asbestos mining (2a) characterized

by the presence of building roofs containing asbestos

and a rural area (2b) with building roofs where

asbestos presence was not detected.

(a)

(b)

Figure 2: Balangero’s asbestos mining site with building

roofs containing asbestos (a) and rural areas with building

roofs without asbestos (b).

Data Fusion of PRISMA Satellite Imagery for Asbestos-containing Materials: An Application on Balangero’s Mine Site (Italy)

151

(a)

(b)

Figure 3: Portion of the underwater asbestos open-pit mine

with presence of natural vegetation (a) and urban area with

the presence of an abandoned industrial site where roofs

contain asbestos (b).

Figure 3a shows another portion of asbestos mine

with presence of natural vegetation. Figure 3b shows

and urban area with the presence of an abandoned

industrial site where building roofs contain asbestos.

The datasets corresponding to Figure 2 were used as

calibration set, while the datasets corresponding to

Figure 3 were utilized as validation set.

2.1 Data Handling and Processing

The PRISMA equipment is made up of an imaging

spectrometer, able of acquiring VNIR (Visible and

Near-InfraRed) and SWIR (Short-Wave InfraRed)

images (~ 250 bands), with a spatial resolution of 30

m on a swath of 30 km, and a panchromatic camera

with spatial resolution of 5 m. The spectral resolution

is about 12 nm in the spectral range of 400-2500 nm,

that are VNIR and SWIR regions (ASI, 2009).

The PRISMA VNIR-SWIR hyperspectral image

and the panchromatic image datasets were imported

into the MATLAB® environment (R2021a, Version

9.10, The Mathworks, Inc.). The details of utilized

dataset are shown in Table 1.

Table 1: Dataset sizes.

Area 1 Area 2

Raw data

VNIR-SWIR

Training

dataset

193*304*23

(Figure 1a)

213*224*234

(Figure 1 b)

Prediction

dataset

125*191*234

(Figure 2a)

240*190*234

(Figure 2b)

Raw data

panchromatic

image

Training

dataset

193*304*1

(Figure 1a)

213*224*1

(Figure 1 b)

Test

dataset

125*191*1

(Figure 2a)

240*190*1

(Figure 2b)

Figure 4: Region of Interests (ROIs) selected on the

panchromatic image.

The imported data were analyzed using the

PLS_Toolbox (Version 8.2 Eigenvector Research,

Inc.) (Wise et al., 2006) and Statistics and Machine

Learning Toolbox. The PLS_Toolbox was used to

pre-process data and for performing the Principal

Component Analysis (PCA), while Statistics and

Machine Learning Toolbox was utilized for setting up

the Classification And Regression Trees (CART).

2.2 Class Setting

From a small part of the panchromatic image,

depicting near a half of the open-pit mine and its

surrounding, six Region of Interests (ROIs) were

selected and the following classes were set: ACM,

Urban Area, Anthropogenic vegetation, Natural

vegetation, Water and Balangero’s mine (Figure 4).

IMPROVE 2022 - 2nd International Conference on Image Processing and Vision Engineering

152

The above-mentioned classes were then selected

according to surface truth to the ground. In ‘ACM’

was included Asbestos Containing Materials (i.e.

Eternit corrugated roofing). In ‘Urban Area’ were

included areas characterized by urban morphology. In

‘Anthropogenic vegetation’ were selected areas with

cultivated crops. In ‘Natu-ral vegetation’ were

included areas with woods and sparse trees patterns.

While, in ‘Water’ class was included the underwater

open-pit mine area. Finally, ‘Balangero’s mine’ class

was chosen selecting barren areas near the open-pit

mine. The class set of the panchromatic image was

then transferred to the VNIR-SWIR dataset.

2.3 Data Pre-processing

Spectral data preprocessing is a necessary step in

order to reduce detector noise, to eliminate the

spectral nonuniformity due to the illumination,

scattering phenomena and the influence of the

changing environmental conditions. Data pre-

treatment is also addressed to better solve the spectra

information by enhancing differences among the

classes. To choose the right pre-processing

algorithms, different pre-processing algorithms were

tested among those widely adopted (Rinnan &

Engelsen, 2009). The algorithms combination which

gave the best data decomposition of the class scores

were selected. The adopted pre-processing

combination was spatial Median Filter (MF),

Standard Normal Variate (SNV), Gap-Segment (GS)

Derivative and Multiway Center (MC).

2.4 Principal Component Analysis

The Principal Component Analysis (PCA) is a well-

known data exploratory method, widely adopted to

HSI datasets, that gives the possibility to have an

overview of complex multivariate data. PCA allows to

reduce the dimensionality of a data matrix containing

multiple interrelated variables, while retaining as much

as possible of the variation present in the data set (Bro

et al., 2014). In this case, PCA was carried out to detect

outliers and choose the data to built-up the classifiers

by reducing the dimensionality of the dataset. The PCA

was performed on the VNIR-SWIR dataset. Six PCs

were chosen. The scores of the performed PCA were

then concatenated with the panchromatic image.

2.5 Classification and Regression Trees

(CART)

Classification And Regression Trees (CART), a non-

parametric statistical technique, was used to classify

the six classes on the panchromatic image fused with

the PCA scores (Shao et al. 2012). Classification And

Regression Trees (CART), a non-parametric

statistical technique, was used to classify the six

classes on the panchromatic image fused with the

PCA scores. CART classification algorithm,

developed by Breiman et al. (1984), allows to build a

decision tree based on Gini’s impurity index as

splitting criterion. In classification and regression

problems, CART algorithm produces a decision tree

describing a response varying as a function of

multiple explanatory variables. A tree hierarchy is

produced by the subdivision process.

In the tree hierarchy, the observation subsets are

represented by the nodes, while the leaves are the

final nodes. A binary model, formulated in each node,

is responsible for the subdivision process. In each

node, all the samples satisfying the model are

clustered in a sub-group, while the remaining nodes

are assigned to another subgroup. The classification

process therefore follows a path along the tree from

the root towards a final leaf. This process can be

synthetized in three steps. In the 1

step of CART

analysis, the binary split procedure allows to build the

maximum tree by finding the best split which

maximizes the splitting criterion. Usually, overfitting

can occur when the maximum tree is overgrown

closely describing the used training set. To correct the

overfitted model a pruning process occurs in the 2

step. The pruned model results in multiple less

complex tree, that are derived from the maximum

tree. In the 3

step, finally, a cross-validation process

helps to select the optimal tree (Deconinck et al.

2006).

The main advantage of CART algorithm relies on

the fact of being nonparametric. Moreover, it can be

used in combination with other prediction algorithms

and by combining both testing with a test data and

cross-validation thus enabling to more precisely

measure the quality of the model fitting.

2.6 Classification Performance Metrics

The performance evaluation of the classification

methods was carried out in terms of prediction maps,

that is false colour images representing the

classification and in terms of the statistical

parameters: Sensitivity and Specificity (Ballabio &

Todeschini, 2009). In more detail, the Sensitivity

represents the ability of determined classifier to

correctly recognize samples belonging to a specified

class and is defined by Equation (1). On the other

hand, Specificity relates to the model ability to

Data Fusion of PRISMA Satellite Imagery for Asbestos-containing Materials: An Application on Balangero’s Mine Site (Italy)

153

correctly reject samples belonging to all the other

classes as defined by Equation (2).

 





(1)

 





(2)

TP are the total number of True Positive, FN the

total number of False Negative, TN the total number

of True Negative and FP the total number of False

Positive. Sensitivity and Specificity were calculated

according to the number of correctly or not correctly

assigned pixel to each defined class, with reference to

calibration (CAL), cross-validation (CV) and to the

prediction of the validation set (PRED).

Figure 5: Region of Interests (ROIs) transferred from the

panchromatic image to the VNIR-SWIR dataset.

Figure 6: Raw spectra of the six classes. (L

TOA

: Top of

Atmosphere radiance).

3 RESULTS AND DISCUSSION

The six Region of Interests (ROIs) selected on the

panchromatic image and transferred to the VNIR-

SWIR dataset (Figure 5) and corresponding mean

spectra are shown in Figure 6. The average spectra of

the six classes show significant spectral differences

but are also very noisy. For this reason, in order to

reduce the noise and emphasize the spectral variation

(as already explained in the Chapter 2.3), the data

were pre-processed as shown in Figure 7.

Figure 7: Pre-processed spectra averaged for the six classes.

The adopted pre-processing combination is spatial Median

Filter (MF), Standard Normal Variate (SNV), Gap-Segment

(GS) 1st Derivative and Multiway Center (MC).

After data preprocessing, a PCA model was

created (Figure 8). PCA model allows to capture

99.69% of the total variance with six principal

components. In detail, the PCA score plot of PC1 and

PC3 shows 6 separated clusters corresponding to the

6 different classes considered (i.e. ‘ACM’, ‘Urban

Area’, ‘Anthropogenic vegetation’, ‘Natural

vegetation’, ‘Water’ and ‘Balangero’s mine’). The

positive space of scores on PC3 is mainly influenced

by ‘Balangero’s Mines’ spectra. The negative space

of scores on PC1 is mainly influenced by water of

Balangero’s lake. The combination of the positive

value of scores PC1 and negative value of scores on

PC3 allows the separation of ‘Urban area’,

‘Anthropogenic’ and ‘Natural vegetation’. Finally,

the ACM class is separated from the other classes by

the combination of negative PC1 and positive PC3

values.

Pre-processed L

TOA

IMPROVE 2022 - 2nd International Conference on Image Processing and Vision Engineering

154

Figure 8: PCA scores plot (a) and loadings plot (b) of the

first and third PCs.

The negative space of scores on PC1 is mainly

influenced by water of Balangero’s lake. The

combination of the positive value of scores PC1 and

negative value of scores on PC3 allows the separation

of ‘Urban area’, ‘Anthropogenic’ and ‘Natural

vegetation’. Finally, the ACM class is separated from

the other classes by the combination of negative PC1

and positive PC3 values.

Starting from the good separation obtained, this

PCA model was used for the data reduction of

PRISMA hyperspectral datasets. The PCA-reduced

datasets were concatenated to the corresponding

panchromatic images as shown in the example

reported in Figure 9. In detail, the concatenated

procedure allows to combine chemical information

coming from VNIR-SWIR range with the shapes and

contours of the topographical details obtained from

the panchromatic image.

Figure 9: False color image of the novel dataset, resulting

from the fusion of VNIR-SWIR PCA scores and the

panchromatic of the studied area.

Starting from this new data set, the classes for the

calibration dataset were set as shown in Figure 10.

The calibration set was then created and utilized to

build the CART model.

Figure 10: Class map of the Training set.

The results in terms of Sensitivity and Specificity

(Table 2) confirm the good performance of the model,

with values ranging from 0.816 (‘ACM’ class) to

0.980 (‘Water’ class) and 0.979 (‘Anthropogenic

vegetation’ class) to 1.00 (‘Water class’),

respectively, both in calibration and cross-validation.

Data Fusion of PRISMA Satellite Imagery for Asbestos-containing Materials: An Application on Balangero’s Mine Site (Italy)

155

Table 2: results in terms of Sensitivity and Specificity.

Class

Sens.

(Cal)

Spec.

(Cal)

Sens.

(CV)

Spec.

(CV)

ACM 1,000 1,000 0.816 0.999

Anthropogenic

vegetation

1,000 1,000 0.981 0.979

Balangero's

Mines

1,000 1,000 0.946 0.997

Natural

vegetation

1,000 1,000 0.962 0.992

Urban area 1,000 1,000 0.963 0.988

Water 1,000 1,000 0.980 1.000

The CART prediction map of the validation set is

reported in Figure 11, whereas the performance

metrics of the classification model applied to the

validation set are shown in Table 3. The results in

terms of prediction images (Figure 11) are in

agreement with those achieved in the calibration

phase. By analyzing the performance metric

parameters reported in Table 3, despite a slight

decrease in sensitivity in the identification of ACM

roofs, is clear that the identification of the main

structures with asbestos is correct, confirming the

success of the proposed test.

4 CONCLUSIONS

In this paper, a novel approach was developed and

implemented to identify ACM from "Urban Area",

Anthropogenic and natural vegetation", "Water" and

“Balangero’s mine”.

PRISMA satellite hyperspectral images were

elaborated through multivariate statistical analysis in

order to extract the chemical features of the classes.

Subsequently, the PRISMA hyperspectral data,

reduced by PCA, were concatenated with the

panchromatic image in order to combine chemical

information with the shapes and contours of the

topographical details. Afterwards, starting from the

fused dataset, a CART classification model was

developed in order to recognize the roofing

containing asbestos from other objects on the images.

The adopted procedure proved to have a

significant discriminating capacity in terms of

sensitivity and specificity enabling the possibility to

use this approach for more extended areas.

Figure 11: CART prediction map.

Table 3: Performance metrics of CART classification in

prediction calculated on the test set.

Class

Sens.

(Pred)

Spec.

(Pred)

ACM 0.756 0.997

Anthropogenic vegetation 0.993 0.981

Balangero's Mines 0.999 0.993

Natural vegetation 0.937 0.997

Urban area 0.980 0.994

Water 1.000 1.000

The possibility of a systematic and integrated use

of PRISMA image combined with machine learning

tools for ACMs identification, proved to be a

complementary method for a faster identification and

mapping of contaminated areas, with less risk of

exposure for operators and the possibility to perform

a fast and reliable survey of ACPs by remote sensing.

The fulfilment of the previous mentioned goals

could produce positive environmental impacts, as

well as big economic benefits related to the lower

identifying and mapping costs.

IMPROVE 2022 - 2nd International Conference on Image Processing and Vision Engineering

156

ACKNOWLEDGEMENTS

The study was developed in the framework of INAIL

(italian National Institute for Insurance against

Accidents at Work) project: BRIC ID 60/2 -

Valutazione e comparazione dei livelli di

informazione ottenibili da remoto con sensoristica

ottica innovativa per l’identificazione di MCA in

diversi contesti del territorio nazionale; confronto dei

dati ottenuti con risultati analitici acquisiti in

laboratorio. Definizione di tecniche di

campionamento ed analisi per il monitoraggio della

presenza di erionite.

REFERENCES

Amano, Y, Takagi, N and Getz, AFH. (2008). A study on

the classification of urban region using hyper-spectrum

data at AVIRIS. In: Geoscience and Remote Sensing

Symposium, 2008. IGARSS 2008. IEEE International.,

Boston, MA, USA: IEEE, pp. 687-690.

Aplin, P. & Smith, G.M. (2008). Advances in Object Based

Image Classification. The International Archives of the

Photogrammetry, Remote Sensing and Spatial

Informatio Sciences, Beijing, Vol.37, 725-728.

http://www.isprs.org/proceedings/XXXVII/congress/7

_pdf/4_WG-VII-4/45.pdf

Ballabio, D. & Todeschini, R. (2009). Multivariate classifi-

cation for qualitative analysis. Infrared spectroscopy for

food quality analysis and control, 83, e102.

Biancolillo, A., Bucci, R., Magrì, A. L., Magrì, A. D., &

Marini, F. (2014). Data-fusion for multiplatform

characterization of an Italian craft beer aimed at its

authentication. Analytica chimica acta, 820, 23-31.

Blaschke, T. (2010). Object-based Image Analysis for

Remote Sensing. ISPRS Journal of Photogrammetry

and Remote Sensing, 65, 2-16.)

Bonifazi, G., Capobianco, G., & Serranti, S. (2018).

Asbestos containing materials detection and

classification by the use of hyperspectral imaging.

Journal of hazardous materials, 344, 981-993.

Bonifazi, G., Capobianco, G., & Serranti, S. (2018).

Asbestos containing materials detection and

classification by the use of hyperspectral imaging.

Journal of hazardous materials, 344, 981-993.

Bonifazi, G., Capobianco, G., & Serranti, S. (2019).

Hyperspectral imaging and hierarchical PLS-DA

applied to asbestos recognition in construction and

demolition waste. Applied Sciences, 9(21), 4587.

Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984).

Cart. Classification and Regression Trees; Wadsworth

and Brooks/Cole: Monterey, CA, USA.

Bro, R., & Smilde, A. K. (2014). Principal component

analysis. Analytical methods, 6(9), 2812-2831.

Chen, Z., Ning, X. and Zhang, J. (2012). Urban Land Cover

Classification Based on WorldView-2 Image Data.

IEEE International Symposium on Geomatics for

Integrated Water Resource Management, Lanzhou, 19-

21 October 2012, 1-5.

Deconinck, E., Zhang, M. H., Coomans, D., & Vander

Heyden, Y. (2006). Classification tree models for the

prediction of blood− brain barrier passage of drugs.

Journal of chemical information and modeling, 46(3),

1410-1419.

Directive 2003/18/EC of the European Parliament and of

the Council of 27 March 2003 amending Council

Directive 83/477/EEC on the protection of workers

from the risks related to exposure to asbestos at work.

Frassy, F., Marchesi, A., Candiani, G., Rusmini, Marco.,

Maianti, P., Marchesi, A., Rota Nodari, F., Della Via,

G., Albonico, C., Gianinetto, Marco. (2014) "Mapping

Asbestos Cement Rooﬁng with Hyperspectral Remote

Sensing over a Large Mountain Region of the Italian

Western Alps," Sensors, vol. 14, no. 9, pp. 1590015913,

doi: 10.3390/s140915900.

Marino, C. M., Panigada, C. and Busetto, L. (2001)

Airborne hyperspectral remote sensing applications in

urban areas: asbestos concrete sheeting identiﬁcation

and mapping, IEEE/ISPRS Joint Workshop on Remote

Sensing and Data Fusion over Urban Areas (Cat.

No.01EX482), pp. 212216, doi: 10.1109/DFUA.20

01.985882.

Massarelli, C., Matarrese, R., Uricchio, V. F., Muolo, M.

R., Laterza M. & Ernesto L. (2017) Detection of

asbestos containing materials in agroecosystem by the

use of airborne hyperspectral CASI1500 sensor

including the limited use of two UAVs equipped with

RGB cameras. International Journal of Remote

Sensing, 38:810, 21352149, DOI: 10.1080/01431161.

2016.1226528

Paglietti, F., Malinconico, S., della Staffa, B. C.,

Bellagamba, S., & De Simone, P. (2016). Classification

and management of asbestos-containing waste:

European legislation and the Italian experience. Waste

management, 50, 130-150.

Rinnan, Å., Van Den Berg, F., & Engelsen, S. B. (2009).

Review of the most common pre-processing techniques

for near-infrared spectra. TrAC Trends in Analytical

Chemistry, 28(10), 1201-1222.

Serranti, S., Bonifazi, G., Capobianco, G., Malinconico, S.,

& Paglietti, F. (2019, May). Hyperspectral imaging

applied to asbestos containing materials detection:

specimen preparation and handling. In Advanced

Environmental, Chemical, and Biological Sensing

Technologies XV (Vol. 11007, p. 110070S).

International Society for Optics and Photonics.

The PRISMA mission (PDF). ASI. 4 September 2009.

Archived from the original (PDF) on 4 October 2013.

Retrieved 23 August 2013.

Wise, B. M., Gallagher, N. B., Bro, R., Shaver, J. M.,

Windig, W., & Koch, R. S. (2006). Chemometrics

tutorial for PLS_Toolbox and Solo. Eigenvector

Research, Inc, 3905, 102-159.

Shao, Y., Lunetta, R. S. (2012) Comparison of support

vector machine, neural network, and CART algorithms

for the landcover classiﬁcation using limited training

data points. ISPRS Journal of Photogrammetry and

Remote Sensing, Volume 70, Pages 7887.

Data Fusion of PRISMA Satellite Imagery for Asbestos-containing Materials: An Application on Balangero’s Mine Site (Italy)

157