"This site requires JavaScript to work correctly"
minimize Header

Michael Fernandes, Dipl.-Phys.

Centre for Applied Research

Technology Campus Grafenau

Academic Staff

Grafenau

08552/975620-44

Sortierung:
Zeitschriftenartikel

  • Alexander Pletl
  • Michael Fernandes
  • N. Thomas
  • A. Rossi
  • Benedikt Elser

Spectral clustering of CRISM datasets in Jezero crater using UMAP and k-Means

In: Remote Sensing vol. 15

  • (2023)

DOI: 10.3390/rs15040939

In this paper, we expand upon our previous research on unsupervised learning algorithms to map the spectral parameters of the Martian surface. Previously, we focused on the VIS-NIR range of hyperspectral data from the CRISM imaging spectrometer instrument onboard NASA’s Mars Reconnaissance Orbiter to relate to other correspondent imager data sources. In this study, we generate spectral cluster maps on a selected CRISM datacube in a NIR range of 1050–2550 nm. This range is suitable for identifying most dominate mineralogy formed in ancient wet environment such as phyllosilicates, pyroxene and smectites. In the machine learning community, the UMAP method for dimensionality reduction has recently gained attention because of its computing efficiency and speed. We apply this algorithm in combination with k-Means to data from Jezero Crater. Such studies of Jezero Crater are of priority to support the planning of the current NASA’s Perseversance rover mission. We compare our results with other methodologies based on a suitable metric and can identify an optimal cluster size of six for the selected datacube. Our proposed approach outperforms comparable methods in efficiency and speed. To show the geological relevance of the different clusters, the so-called “summary products” derived from the hyperspectral data are used to correlate each cluster with its mineralogical properties. We show that clustered regions relate to different mineralogical compositions (e.g., carbonates and pyroxene). Finally the generated spectral cluster map shows a qualitatively strong resemblance with a given manually compositional expert map. As a conclusion, the presented method can be implemented for automated region-based analysis to extend our understanding of Martian geological history.
  • TC Grafenau
  • DIGITAL
Vortrag

  • N. Thomas
  • Michael Fernandes
  • Benedikt Elser
  • A. Rossi
  • Alexander Pletl
  • G. Cremonese

Machine learning approaches to matching CaSSIS colour imaging with CRISM imaging spectroscopy (Abstract B4.2-0036-22)

In: 44th COSPAR Scientific Assembly

Committee on Space Research Online

  • 16.-24.07.2022 (2022)

The Colour and Stereo Surface Imaging System (CaSSIS) onboard ESA's ExoMars Trace Gas Orbiter (TGO) (Thomas et al., 2017) has been providing 4-band colour observations of the surface of Mars since entering prime mission on 21 April 2018. The filters were selected to provide discrimination between minerals at high signal to noise while retaining the relatively high spatial resolution of a broad-band visible imager. Tornabene et al. (2018) showed that CaSSIS should be effective at this task although the atmospheric contribution to the data had not been modelled at that time. The data set obtained in the prime mission has more than justified the careful selection of the filters and imaging approach. The data quality has prompted us to consider whether CaSSIS data can be used to "extend" the coverage of imaging spectroscopy data provided by the CRISM instrument onboard NASA's Mars Reconnaissance Orbiter. The idea is that by linking CRISM spectra to the CaSSIS colour data one can then use CaSSIS data to identify mineral relationships beyond the coverage of the CRISM data set. The link is provided by mapping the CaSSIS four-band images to spectral clusters using a machining learning classifier. Clearly, it is important to benchmark this approach with respect to traditional correlation based methods. Furthermore, quantitative analyses are needed to provide a statistically justifiable confidence level for the derived results. Currently the work suggests that this can be achieved at least in highly feature-rich landscapes. We also note that Gao et al. (2021) have been applying similar approaches for studies of Jezero Crater. In principle, three different statistical tasks have to be addressed to apply machine learning: Dimensionality reduction, clustering and classification. Tests are being carried out to establish the best algorithm for the classification and the optimum number of classes. We also seek to establish why one algorithm is better than another so that a robust approach can be adopted for many sites. Correlation between CRISM and CaSSIS data shows that the CaSSIS NIR band has a high discrimination power (as we intuitively assumed leading to its inclusion in the design) and certain spectral features in CRISM do correlate strongly. First preliminary results indicate that a Random Forest classifier slightly outperforms other standard machine learning methods. Splitting the spectral data into VIS, NIR- ranges is a relevant preprocessing step to determine suitable clusters. The progress will be described in the presentation with examples from the Vallis Marineris and Noachis Terra regions. Ref. Gao et al., (2021), Generalized Unsupervised Clustering of Hyperspectral Images of Geological Targets in the Near Infrared, arXiv:2106.13315v1 [Titel anhand dieser ArXiv-ID in Citavi-Projekt übernehmen] Thomas, N., and 60 colleagues,(2017),The Colour and Stereo Surface Imaging System (CaSSIS) for the ExoMars Trace Gas Orbiter,Space Science Reviews,212,1897 Tornabene, L.L. et al., (2018),Image Simulation and Assessment of the Colour and Spatial Capabilities of the Colour and Stereo Surface Imaging System (CaSSIS) on the ExoMars Trace Gas Orbiter, Space Science Reviews,214,18
  • TC Grafenau
  • DIGITAL
Vortrag

  • Michael Fernandes
  • N. Thomas
  • Benedikt Elser
  • A. Rossi
  • Alexander Pletl
  • G. Cremonese

Extrapolation of CRISM based spectral feature maps using CaSSIS four-band images with machine learning techniques

In: EGU General Assembly 2022

Vienna, Austria

  • 23.-27.05.2022 (2022)

DOI: 10.5194/egusphere-egu22-2765

Spectroscopy provides important information on the surface composition of Mars. Spectral data can support studies such as the evaluation of potential (manned) landing sites as well as supporting determination of past surface processes. The CRISM instrument on NASA’s Mars Reconnaissance Orbiter is a high spectral resolution visible infrared mapping spectrometer currently in orbit around Mars. It records 2D spatially resolved spectra over a wavelength range of 362 nm to 3920 nm. At present data collected covers less than 2% of the planet. Lifetime issues with the cryo-coolers prevents limits further data acquisition in the infrared band. In order to extend areal coverage for spectroscopic analysis in regions of major importance to the history of liquid water on Mars (e.g. Valles Marineris, Noachis Terra), we investigate whether data from other instruments can be fused to extrapolate spectral features in CRISMto these non-spectral imaged areas. The present work will use data from the CaSSIS instrument which is a high spatial resolution colour and stereo imager onboard the European Space Agency’s ExoMars Trace Gas Orbiter (TGO). CaSSIS returns images at 4.5 m/px from the nominal 400 km altitude orbit in four colours. Its filters were selected to provide mineral diagnostics in the visible wavelength range (400 – 1100 nm). It has so far imaged around 2% of the planet with an estimated overlap of ≲0.01% of CRISM data. This study introduces a two-step pixel based reconstruction approach using CaSSIS four band images. In the first step advanced unsupervised techniques are applied on CRISM hyperspectral datacubes to reduce dimensionality and establish clusters of spectral features. Given that these clusters contain reasonable information about the surface composition, in a second step, it is feasible to map CaSSIS four band images to the spectral clusters by training a machine learning classifier (for the cluster labels) using only CaSSIS datasets. In this way the system can extrapolate spectral features to areas unmapped by CRISM. To assess the performance of this proposed methodology we analyzed actual and artificially generated CaSSIS images and benchmarked results against traditional correlation based methods. Qualitative and quantitative analyses indicate that by this novel procedure spectral features of in non-spectral imaged areas can be predicted to an extent that can be evaluated quantitatively, especially in highly feature-rich landscapes.
  • TC Grafenau
  • DIGITAL
Zeitschriftenartikel

  • Michael Fernandes
  • Alexander Pletl
  • N. Thomas
  • A. Rossi
  • Benedikt Elser

Generation and Optimization of Spectral Cluster Maps to Enable Data Fusion of CaSSIS and CRISM Datasets

In: Remote Sensing vol. 14 pg. 2524.

  • (2022)

DOI: 10.3390/rs14112524

Four-band color imaging of the Martian surface using the Color and Stereo Surface Imaging System (CaSSIS) onboard the European Space Agency’s ExoMars Trace Gas Orbiter exhibits a high color diversity in specific regions. Not only is the correlation of color diversity maps with local morphological properties desirable, but mineralogical interpretation of the observations is also of great interest. The relatively high spatial resolution of CaSSIS data mitigates its low spectral resolution. In this paper, we combine the broad-band imaging of the surface of Mars, acquired by CaSSIS with hyperspectral data from the Compact Reconnaissance Imaging Spectrometer (CRISM) onboard NASA’s Mars Reconnaissance Orbiter to achieve a fusion of both datasets. We achieve this using dimensionality reduction and data clustering of the high dimensional datasets from CRISM. In the presented research, CRISM data from the Coprates Chasma region of Mars are tested with different machine learning methods and compared for robustness. With the help of a suitable metric, the best method is selected and, in a further step, an optimal cluster number is determined. To validate the methods, the so-called “summary products” derived from the hyperspectral data are used to correlate each cluster with its mineralogical properties. We restrict the analysis to the visible range in order to match the generated clusters to the CaSSIS band information in the range of 436–1100 nm. In the machine learning community, the so-called UMAP method for dimensionality reduction has recently gained attention because of its speed compared to the already established t-SNE. The results of this analysis also show that this method in combination with the simple K-Means outperforms comparable methods in its efficiency and speed. The cluster size obtained is between three and six clusters. Correlating the spectral cluster maps with the given summary products from CRISM shows that four bands, and especially the NIR bands and VIS albedo, are sufficient to discriminate most of these clusters. This demonstrates that features in the four-band CaSSIS images can provide robust mineralogical information, despite the limited spectral information using semi-automatic processing.
  • TC Grafenau
  • DIGITAL
Zeitschriftenartikel

  • S. Goisser
  • S. Wittmann
  • Michael Fernandes
  • H. Mempel
  • C. Ulrichs

Comparison of colorimeter and different portable food-scanners for non-destructive prediction of lycopene content in tomato fruit

In: Postharvest Biology and Technology vol. 167 pg. 111232.

  • (2020)

DOI: 10.1016/j.postharvbio.2020.111232

Lycopene, the red colored carotenoid in tomatoes, has various health benefits for humans due to its capability of scavenging free radicals. Traditionally, the quantification of lycopene requires an elaborate extraction process combined with HPLC analysis within the laboratory. Recent studies focused simpler methods for determining lycopene and utilized spectroscopic measurement methods. The aim of this study was to compare non-destructive methods for the prediction of lycopene by using color values from colorimeter measurements and Vis/NIR spectra recorded with three commercially available and portable Vis/NIR spectrometers, so called food-scanners. Tomatoes of five different ripening stages (green to red) as well as tomatoes stored up to 22 days after harvest were used for modeling. After measurement of color values and collection of Vis/NIR spectra the corresponding lycopene content was analyzed spectrophotometrically. Applying exponential regression models yielded very good prediction of lycopene for color values L*, a*, a*/b* and the tomato color index of 0.94, 0.90, 0.90 and 0.91, respectively. Color value b* was not a suitable predictor for lycopene content, whereas the (a*/b*)² value had the best linear fit of 0.87. In comparison to color measurements, the cross-validated prediction models developed for all three food-scanners had coefficients of determination (r²CV) ranging from 0.92 to 0.96. Food-scanners also can be used for additional measurements of internal fruit quality, and therefore have great potential for fruit quality assessment by measuring a multitude of important fruit traits in one single scan.
  • TC Grafenau
  • DIGITAL
  • NACHHALTIG
Beitrag in Sammelwerk/Tagungsband

  • S. Goisser
  • J. Krause
  • Michael Fernandes
  • H. Mempel

Determination of tomato quality attributes using portable NIR-sensors

pg. 1-12.

Karlsruhe

  • (2019)

DOI: 10.5445/KSP/1000087509

As part of a research project a multidisciplinary approach of different research institutes is followed to investigate the possibility of using a commercially available miniaturized NIR-sensor for the determination of tomato fruit quality parameters in postharvest. Correlation of spectra and tomato reference values of firmness, dry matter and total soluble solids showed good prediction accuracy. Additionally the decline of firmness over storage time with respect to storage temperature of tomatoes could be modelled. Therefore, the decline of firmness as an indicator for shelf-life can be predicted using this portable NIR-Sensor.
  • TC Grafenau
  • Angewandte Wirtschaftswissenschaften
  • NACHHALTIG
  • DIGITAL
Vortrag

  • Michael Fernandes
  • Florian Wahl

Tomatenkrimi - Der Foodscanner als Ermittler . Keynote

In: Tag der offenen Türe des Technologie Campus Grafenau

Grafenau

  • 12.07.2019 (2019)
  • Angewandte Informatik
  • TC Grafenau
  • NACHHALTIG
Beitrag in Sammelwerk/Tagungsband

  • S. Goisser
  • Michael Fernandes
  • C. Ulrichs
  • H. Mempel

Non-destructive measurement method for a fast quality evaluation of fruit and vegetables by using food-scanner

pg. 1-5.

  • (2018)

DOI: 10.5288/dgg-pr-sg-2018

Recent reports estimate the volume of food loss along the supply chain to 1,3 billion tons globally per year, which equals one-third of food produced for human consumption (FAO, 2011). Further studies conducted for the German food supply chain estimate the quantity of annual food loss between 11 million (Universität Stuttgart, 2012) and 18 million (WWF, 2015) tons. Fruits and vegetables, with a percentage of 44 of the total food loss, are commodities most frequently thrown away (BMEL, 2012). In recent years, a lot of attention is given to so-called food-scanners. Food-scanner are miniaturized near-infrared (NIR) spectrometers, which allow a fast and noninvasive determination of food quality. They can be used as a multidimensional predictor to determine the chemical and physical composition of agricultural and food products (e.g. soluble solids, dry matter, moisture, firmness). Due to their small size and portability these devices can be used for in-field application as well as for researchers andend-consumers (Santos et al.,2013). Studies of Flores et al. (2009) and Kim et al. (2013) indicate that NIRS is suitable for predicting quality attributes of various tomato varieties. The experiments described in this study are conducted on tomatoes and focus on the performance of a food-scanner compared to a benchtop NIR-spectrometer. Important quality parameters of tomato, such as sugar content and firmness, are evaluated with respect to their predictability in order to validate the performance of this new kind of non-destructive measurement method.
  • TC Grafenau
  • NACHHALTIG
  • DIGITAL
Vortrag

  • Michael Fernandes

Food-Scanner: Lebensmittelqualität einfach bestimmen . Posterpräsentation

In: 5. Tag der Forschung

Technische Hochschule Deggendorf Deggendorf

  • 08.03.2018 (2018)
  • Angewandte Informatik
  • TC Grafenau
  • NACHHALTIG
  • DIGITAL
Beitrag in Sammelwerk/Tagungsband

  • S. Goisser
  • Michael Fernandes
  • H. Mempel

Zerstörungsfreie Messmethode zur schnellen Qualitätsbewertung und Haltbarkeitsabschätzung von Lebensmitteln mit Hilfe von Food Scannern

In: 52. Gartenbauwissenschaftliche Jahrestagung „Klimafolgen und Herausforderungen für den Gartenbau“. null (BHGL-Schriftenreihe) pg. 34.

  • (2018)
  • TC Grafenau
  • NACHHALTIG
  • DIGITAL
Zeitschriftenartikel

  • Nari Arunraj
  • Robert Hable
  • Michael Fernandes
  • Karl Leidl
  • Michael Heigl

Comparison of Supervised, Semi-supervised and Unsupervised Learning Methods in Network Intrusion Detection Systems (NIDS) Application

In: Anwendungen und Konzepte in der Wirtschaftsinformatik (AKWI) pg. 10-19.

  • (2017)
With the emergence of the fourth industrial revolution (Industrie 4.0) of cyber physical systems, intrusion detection systems are highly necessary to detect industrial network attacks. Recently, the increase in application of specialized machine learning techniques is gaining critical attention in the intrusion detection community. A wide variety of learning techniques proposed for different network intrusion detection system (NIDS) problems can be roughly classified into three broad categories: supervised, semi-supervised and unsupervised. In this paper, a comparative study of selected learning methods from each of these three kinds is carried out. In order to assess these learning methods, they are subjected to investigate network traffic datasets from an Airplane Cabin Demonstrator. In addition to this, the imbalanced classes (normal and anomaly classes) that are present in the captured network traffic data is one of the most crucial issues to be taken into consideration. From this investigation, it has been identified that supervised learning methods (logistic and lasso logistic regression methods) perform better than other methodswhen historical data on former attacks are available. The results of this study have also showed that the performance of semi-supervised learning method (One class support vector machine) is comparatively better than unsupervised learning method (Isolation Forest) when historical data on former attacks are not available.
  • TC Grafenau
  • 30 S: TC Vilshofen S_EN: TC Vilshofen
  • TC Teisnach Sensorik
  • Institut ProtectIT
  • DIGITAL
Vortrag

  • Michael Fernandes

Das Auge analysiert mit - Datenanalyse und Visualisierung

In: 3. Tag der Forschung - Themenbereiche Wirtschaft und Gesundheit

Technische Hochschule Deggendorf Deggendorf

  • 25.02.2016 (2016)
  • TC Grafenau
  • Angewandte Informatik
Zeitschriftenartikel

  • Nari Arunraj
  • Diane Ahrens
  • Michael Fernandes

Application of SARIMAX model to forecast daily sales in retail industry

In: International Journal of Operations Research and Information Systems (IJORIS) vol. 7 pg. 1-20.

  • (2016)

DOI: 10.4018/IJORIS.2016040101

Abstract During retail stage of food supply chain (FSC), food waste and stock-outs occur mainly due to inaccurate sales forecasting which leads to inappropriate ordering of products. The daily demand for a fresh food product is affected by external factors, such as seasonality, price reductions and holidays. In order to overcome this complexity and inaccuracy, the sales forecasting should try to consider all the possible demand influencing factors. The objective of this study is to develop a Seasonal Autoregressive Integrated Moving Average with external variables (SARIMAX) model which tries to account all the effects due to the demand influencing factors, to forecast the daily sales of perishable foods in a retail store. With respect to performance measures, it is found that the proposed SARIMAX model improves the traditional Seasonal Autoregressive Integrated Moving Average (SARIMA) model. Article Preview 1. Introduction Discount retail stores have been a noticeable feature of German retail market since the 1980s. In particular, the growth in number of discount retail stores have significantly increased after reunification of Germany. Recently, there is a growing trend of increasing varieties of fruits and vegetables with year-around availability across all the German discount retail outlets rather than just in their traditional growing season. In order to attract customers and remain competitive in the market, the fruits and vegetables are exported from foreign countries and stocked for longer periods. Particularly, increase in number of retail stores, availability of varieties of fruits and vegetables (in stock) with short shelf-lives, frequent price variations, and different storage conditions increase the complexity and results in huge amount of food waste. In Germany, the retail sector produces the food waste of around 0.5 million tons per year (Kranert et al., 2012). Although the retail sector contributes only 5% of the total food waste in food supply chain, mostly they are avoidable food waste (wasting food which is fit for consumption). The quantity of food waste that occurs in the home (61%) is partially due to the management decisions in the retail sector (e.g. frequent promotions) that stimulate the consumer’s eagerness to purchase, and distract them to equate their demand with the purchase (Arunraj et al., 2014; Gooch et al., 2010). Hence, the proper decision making in the retail sector can help the suppliers and consumers to avoid the food waste. The role of sales forecasting in reducing the food waste in retail stores is a significant topic of discussion in the recent food waste related studies (Mena et al., 2011; Mena et al., 2014). According to Mena et al. (2011) and Stenmarck et al. (2011), the improvement of forecast accuracy is one of the essential remedial measures to reduce the food waste in the retail sector of food supply chain.
  • TC Grafenau
  • DIGITAL
Vortrag

  • Michael Fernandes

Intelligentes Prognose- und Dispositionsverfahren im Lebensmitteleinzelhandel

In: EssensWert Fachtagung

München/Fürstenfeldbruck

  • 07.10.2014 (2014)
  • Angewandte Informatik
  • TC Grafenau
Beitrag in Sammelwerk/Tagungsband

  • Nari Arunraj
  • Diane Ahrens
  • Michael Fernandes
  • M. Müller

Time series sales forecasting to reduce food waste in retail industry

  • (2014)
  • TC Grafenau
Vortrag

  • Nari Arunraj
  • Diane Ahrens
  • Michael Fernandes
  • Martin Müller

Time series sales forecasting to reduce food waste in retail industry

In: 34th International Symposium on Forecasting

Rotterdam

  • 29.06.-02.07.2014 (2014)
  • Angewandte Informatik
  • TC Grafenau