Classification of granular features in urban built land using machine learning techniques in Google Earth Engine platform
Rapid urbanisation has resulted in uncontrollable growth in developing cities, thus threatening the environment‘s stability and quality of life. while expanding infrastructure development is intended to benefit city dwellers, rising traffic, health, and environmental problems are causes for concern....
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/102432/1/SarojiniDewiNagappanPRAZAK2022.pdf http://eprints.utm.my/id/eprint/102432/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:151689 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Rapid urbanisation has resulted in uncontrollable growth in developing cities, thus threatening the environment‘s stability and quality of life. while expanding infrastructure development is intended to benefit city dwellers, rising traffic, health, and environmental problems are causes for concern. to mitigate the negative effects of increasing urbanisation, urban planning is critical in improving city planning and advancing the goal of sustainable urban development. the urban planning sector uses land use land cover (LULC) change as a primary reference point for monitoring, where it is now primarily used to monitor environmental conditions with little emphasis on defining infrastructure in developed areas. a more accurate representation of urban development properties for a densely populated area allows urban planners to make better decisions about future development, hence mitigating the effects of uncontrollable growth caused by rapid urbanisation. as a consequence, this research aims to enhance urban planners' visualisation of urban development by developing an urban built land classification model using Google Earth Engine (GEE) and satellite data. the research was conducted in three phases: first, a literature review was conducted; second, the classification model was developed using satellite imaging data; and third, the classification model's performance was evaluated using designated assessment metrics. the first step in developing this model was to investigate the machine learning techniques and features used in existing LULC models, focusing on those built using the gee platform. random forest was chosen to develop the urban built model in this study due to its resilience and performance in creating classification models on the GEE platform. in addition, the features analysis resulted in the emergence of a new set of granular features for the classification of urban developed land, namely the automobile, construction land, transport lane, building, vegetation, and water bodies. in the LULC class system, these characteristics represent a finer scale for urban or built-up land and come closest to defining an area's urban development properties. the urban built land classification model was developed on gee using landsat 7 and landsat 8 imagery from the Google Earth Engine data catalog for Selangor from 2015 to 2020. the model was created using random forest with the optimal number of trees and the feature set of automobile, construction land, transport lane, building, vegetation, and water bodies after the hyperparameters were tuned. each feature's classification result was displayed on the map, clearly illustrating the distribution of pixels for each detected feature using a defined colour code to provide an accurate representation of the feature's concentration. the accuracy of the urban built land classification model was then determined using the Overall Accuracy (OA), Kappa coefficient, producer accuracy, and user accuracy, yielding 88% to 92%, 0.69 to 0.79, 53% to 98%, and 53% to 96%, respectively. the high overall accuracy showed that the urban built land classification model had successfully classified finer scale details such as automobile, construction land, transport lane, and building spread, thereby improving existing LULC models and providing a more complete picture of development. in conclusion, the findings of this study will help urban planners make informed decisions about highly urbanised cities, thereby achieving a safe, resilient, and sustainable city whilst limiting unsustainable development. |
|---|