Optimized threshold-based convolutional neural network for plant leaf classification: a challenge towards untrained data

Abstract

The problem of identifying the plant type seems to be tough due to the altering leaf color, and the variations in leaf shape overage. The plant leaf classification is very challenging and important issue to solve. The main idea of this paper is to introduce a novel deep learning-based plant leaf classification model. Initially, the pre-processing is done by RGB to gray scale conversion, histogram equalization, and median filtering for improving the image quality necessary for additional processing. In CNN, the activation function is optimized by the hybrid Shark Smell-based Whale Optimization Algorithm (SS-WOA) in a manner that the classification accuracy is attained maximum. The classification of untrained images is very challenging task, so the optimized threshold-based CNN classification is introduced. From the analysis, the accuracy of the proposed SS-WOA-CNN is 0.86%, 0.78%, 1.28%, and 1.53% advanced than PSO-CNN, GWO-CNN, WOA-CNN, and SSO-CNN, respectively. The accuracy of the proposed SS-WOA-CNN is 4.02%, 3.23%, 1.95%, 2.12%, and 0.57% progressed than NB, SVM, DNN, NN, and CNN. The hybrid SS-WOA optimizes the threshold value that can attain maximum classification accuracy for untrained data. The performance of the developed method is validated by differentiating the diverse traditional machine learning.