An Extreme Learning Machine and Modified Local Binary Pattern Based Framework for Image Classification

The robustness of machine learning classifiers depends directly on the quality of feature extraction. Image classifiers have demonstrated strong results in a variety of applications due to their ability to resolve nonlinear problems; however, their performance is typically determined by the underlying application. For example, the accuracy of image classifiers remains low for plant leaf image classification problems. This work presents a novel Extreme Learning Machine-Local Binary Pattern (ELM-LBP) system with different distance measures for image classification. Experimental results across diverse datasets demonstrate that replacing the traditional Euclidean distance with the Bhattacharrya measure enhances both accuracy and precision in complex tasks.