Fusion of Color-Texture Features based Classification of Fruits using Digital and Thermal Images: A Step towards Improvement

Now a day’s people are very much concerned about quality and safety of food products. So this concern leads to the opportunities in the research to classify infected or noninfected fruits. The research aims to implement an approach to improve the classification accuracy for RGB and thermal image fruit database. The large scope of the field demands a proper choice of data collection and feature representation approach. Here, for the purpose of classification eleven categories of fruits have been considered. In this paper, different feature extraction techniques like color (Color Moments and Color Coherence Vector) and texture (GLCM) have been executed. To achieve better classification accuracy of different fruit images, the fusion of features like color-texture has been carried to classify the images using RF and KNN classifier (algorithm). The results revealed that color features give 100% accuracy with RF and KNN classifier for RGB images whereas fusion of GLCM, Color Moments and Color Coherence Vector gives highest accuracy of 93.4% with RF classifier for thermal images.