Elite Firefly and Long Short Term Memory Based Model for Texture Classification
International journal of intelligent engineering and systems · 2021
The texture is an important feature in the visual content of the images and plays a significant role in the computer vision system.Many existing methods were involved in applying the feature selection and classifiers for texture classification.Existing methods have the limitation of poor convergence in the feature selection and overfitting problems in the classification.In this research, the Elite FireFly -Long Short Term Memory (EFF-LSTM) model is proposed to improve the efficiency of the texture classification.The EFF method has the advantage of replacing the low fitness firefly with a high fitness firefly in the search process and the LSTM model has the advantage of a store the important features in long term for the classification.This process provides good convergence in the feature selection and important features are applied to overcome the overfitting problem.Two datasets such as UIUC and KTHTIPS-2b were used to test the performance of the proposed EEF-LSTM model.The Min-Max normalization method is applied to enhance the quality of the images.The Local Directional Ternary Pattern (LDTP), Dual-Tree Complex Wavelet Transform (DTCWT), and Grey-Level Co-occurrence Matrix (GLCM) were the extracted features from the pre-processed images.The EFF method is applied to select the relevant features with good convergence.The LSTM method is applied to classify the images based on the EFF selected features.The proposed EFF-LSTM model has an accuracy of 96.5 % and the existing Convolutional Neural Network (CNN) has 94.3 % accuracy.