Vision-Based Detection of Water Hyacinth

Abstract

This research proposes a vision-based method for detecting the invasive aquatic weed water hyacinth, commonly known as Eichhornia crassipes. They flourish in moving water, including rivers, lakes, and streams. This plant can double in size and cover an entire body of water in a matter of weeks. By decreasing the amount of oxygen in the water, water hyacinth negatively impacts aquatic life. The surrounding water and soil are drained of nutrients as a result. Using computer vision and machine learning, this article presents a model for detecting water hyacinths. The research provides a method for extracting features from hyacinth images using the Gray Scale Co-Occurrence Matrix (GLCM), a statistical methodology of the second order. For feature vector compilation, the Haralicks characteristics contrast, energy, homogeneity, dissimilarity, and correlation are utilized. The LGBM (Light Gradient Boosting Machine) classifier accurately identifies hyacinths with 88% of accuracy.