Land Use / Land Cover Classification Optimization Using Swarm Algorithms

Tracking changes in urban land use and land cover is essential due to its significant implications for the environment and human society and due to the rapid expansion of cities across the globe. This expansion frequently disrupts natural habitats, but it can also boost economic activity and quality of life. Remote sensing and mapping are the most commonly applied tools to monitor such changes. However, It is a fundamental understanding that machine models are not perfect and In the case of high-resolution imagery, modelling errors are evident because of the data spectral variability, the similarities across the same class, along with the imbalanced data distribution across different classes.

Numerous studies aim to improve mapping accuracy by tackling several remote sensing modelling variables and factors, either individually or combined, leading to improvements in overall mapping accuracy. This includes the choice of source and auxiliary data, data class balancing, feature extraction methods, feature selection methods, machine learning tools and many others. This research primarily studies the impact of machine learning model configuration on overall accuracy, exploring ways to optimize a classification model's performance using basic spectral information and simple feature extraction techniques.