Content based Image Retrieval using Reptile Search Algorithm with Deep Learning for Agricultural Crops

Abstract

Content Based Image Retrieval (CBIR) remains a hot research domain due to its versatile applicability in healthcare, e-commerce, remote sensing, agriculture, etc. It is commonly utilized for searching a huge image database and retrieves the related images which are highly resembling the query image (QI). The CBIR model involves the major process of feature extraction, which aims to extract both low level and high level features. The recent advances in deep learning (DL) models enable to effectively design of CBIR models for agricultural sector. In this aspect, this paper presents a new reptile search algorithm with deep learning driven CBIR (RSA-DLCBIR) model for crops. The proposed RSA-DLCBIR model focuses on the retrieval of similar images in the agricultural imaging database. At the initial stage, the RSA-DLCBIR model employs EfficientNetB0 model for feature extraction process. In addition, the RSA has been exploited to fine tune the hyperparameters related to the EfficientNetB0 model. Finally, Minkowski distance is used as a similarity measurement tool for restoring the related images based on QI. The experimental validation of the RSA-DLCBIR model is tested by benchmark dataset and the outcomes are inspected under several factors. The comparative analysis reported the betterment of the RSA-DLCBIR model over recent approaches.