Evolutionary Algorithm with Machine Learning Enabled Color Texture Image Segmentation and Classification Model

Abstract

Abstract Many image processing and computer vision algorithms rely on image segmentation. Typically, images are too complex to segment using standard image processing techniques. It consists of a wide range of colour, texture, and shape from correlation. A different labelling of the boundary or region develops a difficult task to segment techniques. Color, shape, and texture descriptors are used in the problem assignment for advanced image segmentation techniques. The purpose of this article is to describe the Evolutionary Algorithm with Machine Learning Enabled Color Texture Image Segmentation and Classification (EAML-CTISC) model. The EAML-CTISC model's main goal is to accurately segment and classify images. The EAML-CTISC model primarily performs image pre-processing to improve image quality. The EAML-CTISC model then combines biogeography-based optimization (BBO) with Shannon entropy-based image segmentation to determine the region of interest. A colour co-occurrence matrix (CCM) model-based feature extraction technique is also used. Finally, for image classification, a gravitational search algorithm (GSA) with an autoencoder (AE) model is used. Accuracy, Precision, Recall, and Jaccard Index all have average values of 99.01%, 95.66%, and 95.46%, respectively. This result demonstrated that the EAML-CTISC proposed approach effectively performs orchard images better than the existing approach. The experimental result analysis of the EAML-CTISC algorithm can be tested utilizing a set of benchmark images. The comparison study stated the enhanced performance of the EAML-CTISC approach over recent approaches.