Horticulture image based weed detection in feature extraction with dimensionality reduction using deep learning architecture

Smart farming has become essential, and the use of Unmanned Aerial Vehicle (UAV) photography is becoming an essential component of the process. To successfully acquire UAV footage of different spectral bands and identify plant infestations, machine learning algorithms have been deployed. This study suggests a unique method for weed detection in horticulture images utilising feature extraction and dimensionality reduction approaches. Here, a horticultural image was used as the input, which was subsequently treated for noise reduction, smoothing, and normalisation. Kernel extreme component radial learning machine is used for dimensionality reduction and feature extraction. Experimental analysis is performed in terms of RMSE, NSE, F-1 score, recall, accuracy, and precision. The proposed method achieved 92% accuracy, 76% precision, 66% recall, 58% F-1 score, 59% RMSE, and 55% NSE.