Exposure to high levels of radon can be very harmful and lead to serious health issues. Measuring radon in buildings and houses is an important measure to prevent this. One way of measuring radon is to place out plastic films to be exposed over a period and then analyze images of them. Image processing together with deep learning has become very useful for image recognition and analysis. Training artificial neural networks using huge amount of data to learn to classify and predict new data is a widely used approach. In this project, artificial neural networks were trained to be able to predict the radon measurement of exposed plastic films. The data used was microscopic images of these films that first was modified to fit the training better and then sorted into two datasets. The datasets were divided into 10 classes with measurement values in intervals of 100 up to 1000. Two main types of neural networks were used in different shapes and with different training parameters: Convolutional neural networks and Dense neural networks. The convolutional model was able to predict new data with a 70 percent accuracy and the performance increased with a bigger image size (more pixels) but not with a deeper network architecture. Over 90 percent of the wrong data predictions also belonged to a class in the interval just above or below the predicted result which shows that the network has potential for improvements. The dense model only had a 35 percent accuracy but had a training accuracy of over 90 percent. This is because the model was heavily overfitted. A way to get better results could be to increase the dataset that was used with more images.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:uu-441986 |
| Date | January 2021 |
| Creators | Löfgren, Max |
| Publisher | Uppsala universitet, Signaler och system |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | UPTEC F, 1401-5757 ; 21014 |
Page generated in 0.0172 seconds