Ex-RSFO: Exponential Rider-based Sun Flower Optimization enabled Deep Convolution Neural Network for Image Classification
Main Article Content
Abstract
The collection of large set images is easily obtainable from web pages from huge video datasets. The progression of automatic approaches for handling large amount of images is the most significant one in daily life. However, retrieving and indexing the data is a major challenge in web pages. In addition, automatic organization and indexing of images is also a most important problem. In this paper, Exponential Rider Sun Flower Optimization (exp RSFO)-based Deep Convolution Neural Network (Deep CNN) is developed for image classification. Here, a weighted shape size pattern spectra is employed for extracting the significant features and analysing the patterns. The weighted shape size pattern spectra is considered by adapting weight shape decomposition and gray scale decomposition. Furthermore, Fuzzy Local Information C-Means Clustering (FLICM) technique is applied for clustering process. Besides, Deep Convolution Neural Network (Deep CNN) is utilized for image classification, and the classifier is trained by developed exp RSFO technique. The proposed exp RSFO algorithm is devised by incorporating Rider-based Sun Flower Optimization (RSFO) method and Exponential Weighted Moving Average (EWMA) scheme, whereas RSFO model is the combination of Rider Optimization Algorithm (ROA) and Sun Flower Optimization (SFO) technique. Conversely, the test image is acquired and it is subjected to weighted shape size spectra for feature extraction in testing phase. Here, matching is performed by centroid to identify cluster information, and new cluster vector is generated. Along with this, the performance of the developed approach is evaluated using three metrics, like accuracy, specificity and sensitivity. Hence, the developed technique achieved better accuracy, sensitivity and specificity of 96.21%, 95.22% and 96.20% for image size.
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.