Citrus disease detection using convolution neural network generated features and Softmax classifier on hyperspectral image data

Abstract

Identification and segregation of citrus fruit with diseases and peel blemishes are required to preserve market value. Previously developed machine vision approaches could only distinguish cankerous from non-cankerous citrus, while this research focused on detecting eight different peel conditions on citrus fruit using hyperspectral (HSI) imagery and an AI-based classification algorithm. The objectives of this paper were: (i) selecting the five most discriminating bands among 92 using PCA, (ii) training and testing a custom convolution neural network (CNN) model for classification with the selected bands, and (iii) comparing the CNN’s performance using 5 PCA bands compared to five randomly selected bands. A hyperspectral imaging system from earlier work was used to acquire reflectance images in the spectral region from 450 to 930 nm (92 spectral bands). Ruby Red grapefruits with normal, cankerous, and 5 other common peel diseases including greasy spot, insect damage, melanose, scab, and wind scar were tested. A novel CNN based on the VGG-16 architecture was developed for feature extraction, and SoftMax for classification. The PCA-based bands were found to be 666.15, 697.54, 702.77, 849.24 and 917.25 nm, which resulted in an average accuracy, sensitivity, and specificity of 99.84%, 99.84% and 99.98% respectively. However, 10 trials of five randomly selected bands resulted in only a slightly lower performance, with accuracy, sensitivity, and specificity of 98.87%, 98.43% and 99.88%, respectively. These results demonstrate that an AI-based algorithm can successfully classify eight different peel conditions. The findings reported herein can be used as a precursor to develop a machine vision-based, real-time peel condition classification system for citrus processing.

Citrus disease classification with convolution neural network generated features and machine learning classifiers on hyperspectral image data

“…In addition to accuracy, areas under the ROC curve (i.e., AUC) were also used to measure the performance of the custom shallow CNN with SoftMax and the custom VGG16 with SoftMax. This was done because it considers the entire range of threshold values between 0 and 1 and is not affected by class distribution and misclassification cost (Bradley, 1997;Wang et al, 2009;Yadav et al, 2022a). The AUC can be treated as a measure of separability and the lines belonging to a class that reaches close to the top-left corner is the most separable one.…”

Classifying E.coli concentration levels on multispectral fluorescence images with discrete wavelet transform, deep learning and support vector machine.

“…VGG19 uses SoftMax classifier in its architecture as the last FC layer (Fig. 3) which is a generalized version of logistic regression for multi-class classification (Stanford, 2013;Yadav et al, 2022a). In our application, in the first approach, we used the SoftMax classifier and in the second approach, it was replaced by linear and non-linear (i.e., with radial basis function (RBF) kernel) support vector machine (SVM) classifiers.…”

“…In addition to accuracy, areas under the ROC curve (i.e., AUC) was also used to measure the performance of the VGG19 network. This was done because it considers the entire range of threshold values between 0 and 1 and is not affected by class distribution and misclassification cost (Bradley, 1997;Wang et al, 2009;Yadav et al, 2022a). The AUC can be treated as a measure of separability and the lines belonging to a class that reaches close to the top-left corner is the most separable one.…”

“…Similarly, in order to harness the power of both classical machine learning and CNNs, a combination of both was used in this study as it has been found in many cases that combining both of them can provide improved results (Li et al, 2019;Ragab et al, 2019;Wu et al, 2020). VGG16 is a popular CNN architecture that has been successfully used in wide variety of image classification tasks (Cai et al, 2019;Swasono et al, 2019;Tian et al, 2019;Yadav et al, 2022a).…”

Automated classification of citrus disease on fruits and leaves using convolutional neural network generated features from hyperspectral images and machine learning classifiers

“…3,[8][9][10][11] HSI systems have been widely used in various applications of disease detection and classification tasks by many researchers. 8,[11][12][13] HSI systems are used widely in detection and classification of diseased fruits, vegetables, plants, crops, etc. because they provide unique spectral signatures of these in a wide range of spectra.…”