Machine Learning based Land Use Identification of Aerial Images with Fusion of Thepade SBTC and Triangle Thresholding

Abstract

Discovery of Land Usage, also known as Land Usage Mining, essentially deals with determining the proper usage of a plot. For landform mapping to analyze locations, things, and features, aerial imagery is crucial. Remote sensing techniques are a fundamental and significant resource for detecting this. The aerial images may get captured with Modern means like satellites or drones. Additionally, the remote sensing technique gathers crucial data that may be applied to various planning tasks, including urban planning, conservation, forestry, land use, and many more. Urban regions may extract valuable information about land use and land cover from Very High Resolution (VHR) satellite pictures. Before practical usage is feasible, further research needs to be done on the many approaches that are now accessible. A machine learning technique called the ensemble technique combines many base models to create a single, ideal prediction model. In this paper, firstly, the extraction of features using TSBTC and then triangle thresholding is applied to the land usage dataset. After extracting features, a different ML algorithm is applied to compare the accuracy.Further, the triangle thresholding-based local features are ensembled with TSBTC II-ary global feature extraction technique to enhance performance. Additionally, in addition to the nine distinct ML algorithms, the recommended land use detection method uses three different ensembles of ML algorithms to evaluate performance. Results from the examination of this ensemble approach are based on the Land Use Dataset provided by UC Merced to demonstrate the benefit of using this ensemble method for classifying land use and significantly enhancing stand-alone methods.