Mountain Pine Beetle Detection using IoT
| dc.contributor.advisor | Ghaderi, Majid | |
| dc.contributor.author | Rochester, Edward | |
| dc.contributor.committeemember | Reid, Mary L. | |
| dc.contributor.committeemember | Jacob, Christian J. | |
| dc.date | 2020-06 | |
| dc.date.accessioned | 2020-01-07T23:06:11Z | |
| dc.date.available | 2020-01-07T23:06:11Z | |
| dc.date.issued | 2020-01-06 | |
| dc.description.abstract | The large-scale damage caused by outbreaks of forest pests has significant impact on both the ecosystem and forestry industry. Currently, pest outbreaks are monitored using field, aerial and remote sensing surveys. These methods, however, only provide partial spatial coverage and can detect outbreaks only after they have substantially progressed across wide geographic areas. The goal of this thesis is to build and evaluate an IoT system for real-time Mountain Pine Beetle (MPB) infestation detection using bioacoustic recognition via deep learning techniques. First, we present the design of the IoT system and describe its various hardware and software components. We built our IoT device using LoRa communication technology and DIY components. We use a modified version of an adaptive differential pulse-code modulation for sample encoding to reduce the size of the recorded samples and prepare them for transmission. As a result, 15 packets are required to transmit a single, compressed half-a-second sample. The transmitted samples are then received at a gateway and forwarded to a client-server application for decoding, storing, analysis and visualization. Second, we analyze the MPB bioacoustic characteristics in indoor and outdoor scenarios with the use of live beetles. The samples collected using the designed IoT system hardware were manually analyzed and labeled for the convolutional neural network (CNN) training and testing. To predict the presence of MPB with trees, we evaluate a set of state-of-the-art (SOTA) convolutional neural networks (CNN) architectures on recorded samples and live-measurements. To use CNN we transform the recorded bioacoustic samples into spectrograms. Our evaluation results show that with low-resolution samples provided by the low-power, low data rate IoT devices, the SOTA CNN (specifically, ResNet152) can achieve up to 77% accuracy. We observe that even in the outdoor environment the model still achieves as high as 68% accuracy. We conclude by proposing several approaches for improving the proposed system's performance. | en_US |
| dc.identifier.citation | Rochester, E. (2020). Mountain Pine Beetle Detection using IoT (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. | en_US |
| dc.identifier.doi | http://dx.doi.org/10.11575/PRISM/37418 | |
| dc.identifier.uri | http://hdl.handle.net/1880/111445 | |
| dc.language.iso | eng | en_US |
| dc.publisher.faculty | Science | en_US |
| dc.publisher.institution | University of Calgary | en |
| dc.rights | University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. | en_US |
| dc.subject | Internet of things | en_US |
| dc.subject | Mountain Pine Beetles | en_US |
| dc.subject | Convolutional Neural Networks | en_US |
| dc.subject.classification | Computer Science | en_US |
| dc.title | Mountain Pine Beetle Detection using IoT | en_US |
| dc.type | master thesis | en_US |
| thesis.degree.discipline | Computer Science | en_US |
| thesis.degree.grantor | University of Calgary | en_US |
| thesis.degree.name | Master of Science (MSc) | en_US |
| ucalgary.item.requestcopy | true | en_US |