An effective approach of CNN based hybrid Meta- heuristic optimization classifier for retrieving satellite images

Abstract

W ith the ever-increasing platform of satellite images, image retrieval is one of the interesting areato be analyzed by the researchers. In order to concern about the low-level semantics, the effective classifiers are used to retrieve the satellite images withmore accurate rate. Due to the processing of test and trained samples, the satellite images are grouped into several categories: preprocessing, segmentation, feature extraction, and classification. The raw data are processed to eliminate the noise content that present in an image. The segmented procedure is deployed to set a threshold broad rate to split the image. The intent that occurs in the feature extraction has the ability to extract the informative parameter and then CNN based hybrid meta-heuristic optimization (CNN-HMO) classifieris proposed to classify the image for better performance. The proposed work is demonstrated with the tool of MATLAB. However, different parameters such as precision, sensitity, specificity and accuracyare considered to estimate the proposed classifier. Several existing classifiers like NN, CNN-WOC, CNN-EWOC are used in this observation to analyze the accuracy as a reduced percentage of7.83 %, 4.21 % and 1.81 % respectively compared to the proposed classifier. Additionally, the dataset with different classes is constructed to analyze the performance for variety of classifiers.