CROP CLASSIFIER : RNN VS KNN FOR AGRICULTURAL LAND IMAGE CLASSIFICATION

ABSTRACT The classification of agricultural land using satellite or drone imagery has become increasingly significant for land management, crop monitoring, and environmental analysis. Traditional image classification methods often rely on manual feature extraction, which can be time-consuming and may not capture intricate patterns present in high-resolution images. To address these limitations, this project presents an automated approach that utilizes deep learning, specifically a Regression-based Neural Network (RNN), to classify land images into categories such as agricultural or forest regions. The system is developed in Python using PyTorch, with preprocessing techniques including Principal Component Analysis (PCA) for feature reduction and noise removal. The model is trained on a labeled dataset of land images, where features are extracted and passed through multiple neural layers for accurate prediction. Additionally, a K-Nearest Neighbors (KNN) algorithm is implemented for comparison, demonstrating the advantages of deep learning in handling complex spatial data. Model accuracy is evaluated using metrics like accuracy score and confusion matrix. A Flask-based web application is also integrated, allowing users to upload images and view classification results in real-time. This combination of regression-based learning and comparative analysis with KNN offers a scalable and efficient solution for land image classification tasks in agricultural monitoring. Keywords: Agricultural Land Classification, Regression Neural Network, PyTorch, KNN, PCA, Deep Learning, Flask, Image Processing.