Graffiti Detection using a variable temperature MOx sensor and Machine Learning
(2021) In Master’s Theses in Mathematical Sciences FMSM01 20202Mathematical Statistics
- Abstract
- MOx sensors are widely used for sensing gases. In this thesis a variable temperature MOx sensor was used to detect graffiti spray paint and differentiate it from other substances using the sensors sensitivity to different gases at different temperatures using machine learning. The machine learning methods Support Vector Machines, k-Nearest Neighbour and neural networks were evaluated and compared. Different choices of features, including temperature sensitivity and derivatives, as well as different sensors and operating temperatures were tested. Data preprocessing, in particular standardization, was performed and its results evaluated. Data collection was done by spraying graffiti spray paint and applying other substances to the sensor... (More)
- MOx sensors are widely used for sensing gases. In this thesis a variable temperature MOx sensor was used to detect graffiti spray paint and differentiate it from other substances using the sensors sensitivity to different gases at different temperatures using machine learning. The machine learning methods Support Vector Machines, k-Nearest Neighbour and neural networks were evaluated and compared. Different choices of features, including temperature sensitivity and derivatives, as well as different sensors and operating temperatures were tested. Data preprocessing, in particular standardization, was performed and its results evaluated. Data collection was done by spraying graffiti spray paint and applying other substances to the sensor outdoors.
This resulted in a graffiti spray paint detector that worked very well in differentiating graffiti spray paint from other substances. All three used machine learning methods performed well on the data, reaching a very good performance when there was a significant amount of gas reacting with the sensor. The inclusion of derivatives as a feature resulted in higher performance but a larger amount of false alarms due to the added noise. (Less) - Popular Abstract
- Gas sensors, or gas detectors, have long been used for the purpose of sensing and detecting gases. Metal Oxide (MOx) gas sensors are one type of gas sensor that are widely used for this purpose.
In this thesis we have constructed a graffiti spray paint detector that can detect if graffiti spray paint is sprayed near an MOx gas sensor. We investigate the viability of using a single MOx gas sensor in combination with several different machine learning algorithms for the purpose of detecting graffiti spray paint. The particular sensor used is capable of quickly changing its temperature which makes identifying gases possible as the sensor will react in slightly different ways to different gases depending on its temperature.
To... (More) - Gas sensors, or gas detectors, have long been used for the purpose of sensing and detecting gases. Metal Oxide (MOx) gas sensors are one type of gas sensor that are widely used for this purpose.
In this thesis we have constructed a graffiti spray paint detector that can detect if graffiti spray paint is sprayed near an MOx gas sensor. We investigate the viability of using a single MOx gas sensor in combination with several different machine learning algorithms for the purpose of detecting graffiti spray paint. The particular sensor used is capable of quickly changing its temperature which makes identifying gases possible as the sensor will react in slightly different ways to different gases depending on its temperature.
To differentiate gases from each other machine learning was used. Machine learning is a tool that learns from data and identifies patterns that can be used for the differentiation. Three different machine learning methods were evaluated and compared. These methods are called Support Vector Machine, k-Neareast Neighbour and a neural network.
This resulted in a graffiti spray paint detector that worked well for detecting graffiti spray paint and differentiating it from other gases. All evaluated machine learning methods achieved good performances. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9043078
- author
- Kristensson, Mikael LU and Clemedtson, Christian LU
- supervisor
- organization
- alternative title
- Graffitidetektion med en temperaturvariabel MOx sensor och maskininlärning
- course
- FMSM01 20202
- year
- 2021
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- MOX sensor, variable temperature, machine learning, graffiti, spray paint, Support Vector Machine, k-Nearest Neighbour, neural network
- publication/series
- Master’s Theses in Mathematical Sciences
- report number
- LUTMS-3412-2021
- ISSN
- 1404-6342
- other publication id
- 2021:E14
- language
- English
- id
- 9043078
- date added to LUP
- 2021-05-12 09:39:31
- date last changed
- 2021-05-18 13:01:31
@misc{9043078, abstract = {{MOx sensors are widely used for sensing gases. In this thesis a variable temperature MOx sensor was used to detect graffiti spray paint and differentiate it from other substances using the sensors sensitivity to different gases at different temperatures using machine learning. The machine learning methods Support Vector Machines, k-Nearest Neighbour and neural networks were evaluated and compared. Different choices of features, including temperature sensitivity and derivatives, as well as different sensors and operating temperatures were tested. Data preprocessing, in particular standardization, was performed and its results evaluated. Data collection was done by spraying graffiti spray paint and applying other substances to the sensor outdoors. This resulted in a graffiti spray paint detector that worked very well in differentiating graffiti spray paint from other substances. All three used machine learning methods performed well on the data, reaching a very good performance when there was a significant amount of gas reacting with the sensor. The inclusion of derivatives as a feature resulted in higher performance but a larger amount of false alarms due to the added noise.}}, author = {{Kristensson, Mikael and Clemedtson, Christian}}, issn = {{1404-6342}}, language = {{eng}}, note = {{Student Paper}}, series = {{Master’s Theses in Mathematical Sciences}}, title = {{Graffiti Detection using a variable temperature MOx sensor and Machine Learning}}, year = {{2021}}, }