Spatial Attention Mechanism Integrated Ensemble Machine Learning for Real Time Prediction and Classification in Precision Agriculture

Generating spatial and temporal features for real time prediction and classification in precision agriculture is challenging due to inaccurate estimation of features from the remote sensing images. Using Geographical Information System (GIS) and systematic data regarding spatial and temporal data for monitoring and management of agricultural activities. The proposed research integrated GIS and used satellite image processing with Ensemble Machine Learning (EML) algorithms. Initially, the remote sensing data is pre-processed with normalization technique to remove noise. The pre-processed data is given as input to the Improved Speeded Up Robust Features (ISURF) for fast computation and identification of relevant features. The EML algorithms such as Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) are used as a prediction and classification models considering the ISURF features as input. Additionally, the Spatial Attention Mechanism (SAM) is integrated with the EML algorithm to particularly focus on the spatial features of the remote sensing data. The experimental results showed that the proposed SAM-EML achieved precise prediction and classification with 98.30% accuracy, and 97.19% precision when compared to the existing methods Random Forest Regression (RFR) and RF-SVM.