Implementation of Image Processing and Machine Learning in High Resolution Aerial Image Datasets for Lake Resource Usage, Aquaculture, and Coastal Community
Last May 2019, fish farms in Taal Lake suffer from fish kill resulting in an estimated loss of 405 tons of fish. According to the report of BFAR, the measured water sample from the affected areas shows depletion of dissolved-oxygen level that caused the fish kill. DENR stated that the triggering fac...
Saved in:
| Main Author: | |
|---|---|
| Format: | text |
| Published: |
Archīum Ateneo
2020
|
| Subjects: | |
| Online Access: | https://archium.ateneo.edu/theses-dissertations/413 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Ateneo De Manila University |
| Summary: | Last May 2019, fish farms in Taal Lake suffer from fish kill resulting in an estimated loss of 405 tons of fish. According to the report of BFAR, the measured water sample from the affected areas shows depletion of dissolved-oxygen level that caused the fish kill. DENR stated that the triggering factor of the oxygen level deterioration is the over-crowding of fish cages. Recent studies utilize satellite remote-sensors to map and monitor the aquaculture inside the lake. The maps are being used as reference material for progress monitoring, as decision-support and lake management tool by the local government and regulatory agencies. With the advent of Unmanned Aerial Vehicle (UAV) technology, aerial images can be captured with higher resolution and much lower cost compared to satellite imagery. This study makes use of Ateneo Innovation Center high resolution aerial image datasets to create segmentation model for aquaculture structures and coastal settlements. The image dataset was sliced and annotated then fed into the training process to generate a detection model. This study implemented Mask Regional Convolutional Neural Network (Mask RCNN) as machine learning framework to detect and segment the desired artificial geographic objects (aquaculture and roof). After training and validation processes, the method resulted to detection and segmentation of aquaculture structures and coastal settlements. Finally, an analytical software was developed to utilize segmented maps for zone management plan implementation, lake resource usage calculation, and gauge the population of settlers along the coastline. This provides meaningful visual and statistical data regarding aquaculture population, lake resource usage, local settlement population and zone development plan status. |
|---|