Enhanced Method of Detection of Skin Defects in Citrus Fruits Using Convolutional Neural Network Algorithm in Comparison with Principal Component Analysis Method to Improve Accuracy

Aim: To identify key characteristics and evaluate accuracy of identifying skin diseases in citrus fruits on their surface using a novel convolutional neural network compared with principal component analysis. Materials and method: To investigate the accuracy for the dataset containing sample size (N=10), two techniques namely a Novel Convolutional Neural Network algorithm and Principal Component Analysis algorithm were incorporated with G-power=0.8. Result: It was discovered that the categorization accuracy of CNN was found to be (91.82%) whilst the principal component analysis algorithm's accuracy was (62%) using citrus diseases dataset. The T-test shows that the convolutional neural network classifiers appear to be of superior significance than principal component analysis with p=0.03 (p<0.05). Conclusion: In this article, types of citrus diseases are classified using the two models, PCA and CNN. The models helped in extracting discriminative characteristics from the input images. From results, it can be concluded that in comparison to other methodologies, the deep learning methodology is a pragmatic approach.