abdulquawiyy-owolabi/Sentinel-2-Image-Classification-Using-Random-Forest-and-SVM

This project involves analyzing Sentinel-2 satellite imagery for Iwo Local Government, Osun State, Nigeria. The study focuses on understanding spatial patterns, extracting key insights using machine learning models, and visualizing vegetation and land cover features. # Sentinel 2 Image Classification Using Random Forest and SVM ## **Project Overview** This project analyses Sentinel-2 satellite imagery for Iwo Local Government, Osun State, Nigeria. The study focuses on understanding spatial patterns, extracting key insights using machine learning models, and visualizing vegetation and land cover features. The workflow includes: 1. Data exploration and preprocessing. 2. Visualization of spectral bands and derived indices (e.g., NDVI). 3. Implementation of machine learning models for land classification. --- ## **Objectives** - Conduct an exploratory analysis to understand the spectral and spatial characteristics of Sentinel-2 data. - Apply machine learning models to classify land cover types and evaluate their performance. --- ## **Data Sources** - **Sentinel-2 Imagery:** Multispectral satellite data providing high-resolution information across 13 spectral bands. - **Geospatial Shapefiles:** Boundary data of Iwo Local Government and class files and the feature classes used for the project area ['Water Bodies', 'Trees', 'Crops', 'Built up Areas', 'Bareland', 'Rangeland'] --- ## **Methodology** ### **1. Exploratory Data Analysis (EDA)** - **Data Loading and Preparation:** Loaded and preprocessed Sentinel-2 bands using `rasterio`, aligning and clipping the data to the study area. - Calculation of NDVI and NDMI - **Data Visualization:** - True-color and false-color composites were created to visually understand the landscape. * The top left image shows the `Natural Color` view, which presents the landscape as it would appear to the human eye from above. In this image, we can see what appears to be an urban centre (appearing as a bright white area in the centre) surrounded by darker regions that likely represent less developed or natural areas. * The "False Color Composite" (top right) uses specific wavelength bands to emphasize vegetation and urban features. In this visualization, the red areas typically indicate healthy veg …