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

Abstract

Citrus black spot (CBS) is a quarantine fungal disease caused by Phyllosticta citricarpa that can limit market access for fruit. It causes lesions on fruit surfaces and may lead to premature fruit drops, reducing yield. Leaf symptoms are uncommon for CBS, although the fungus reproduces in leaf litter. Similarly, citrus canker is another serious disease caused by the bacterium Xanthomonas citri subsp. citri (syn. X. axonopodis pv. citri) and leads to economic losses for growers from fruit drops and blemishes. Therefore, early detection and management of groves infected by CBS or canker via fruit and/or leaf inspection can benefit the Florida citrus industry. Manual inspection to classify disease symptoms on either fruits or leaves is a tedious and labor intensive process. Hence, there is need to develop computer vision system for autonomous classification of fruits and leaves that can speed up their management in fields. In this paper, we demonstrate the capability of convolution neural network (CNN)-based deep learning along with classical machine learning (ML) based computer vision algorithms to classify 'Valencia' orange fruit surfaces with CBS infection along with four other conditions and 'Furr' mandarin leaves with canker and four other conditions. Fruits with CBS and four other conditions (marketable, greasy spot, melanose and wind scar) were classified using a custom shallow CNN with SoftMax and RBF SVM at an overall accuracy of 89.8% and 92.1%, respectively. Similarly, a custom VGG16 network with SoftMax could classify canker leaves with F1-score of 85% and overall accuracy of 82% including other four conditions (control/healthy, greasy spot, melanose and scab). In addition, it was found that by replacing SoftMax with RBF SVM in the VGG16 network, the overall classification accuracy improved to 93% i.e., an increment of 11% points (13.41%). The preliminary findings reported in this paper demonstrate the capability of HSI system for automated citrus fruit and leaf disease classification using shallow and deep CNN generated features and ML classifiers.

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

“…The overall goals of this study, which is a derivative of our previous study, 16 are to explore the application of shallow CNN with SoftMax and SVM to classify HSI images of CBS infected "Valencia" orange fruit from four other conditions (greasy spot, melanose, wind scar, and marketable) and to use VGG16 deep CNN with SoftMax and SVM to classify citrus canker-affected leaves from four other conditions (control, greasy spot, melanoses, and scab). The specific objectives are as follows: (i) use PCA to select the top five discriminant bands from the 92 HSI bands used in imaging CBS infected orange fruits, (ii) train a custom shallow CNN with SoftMax and SVM classifiers using the selected five bands for the classification of orange fruits with CBS and four other conditions, (iii) use PCA to select the top five discriminating bands from the 348 HSI bands used in imaging citrus leaves affected with canker and four other conditions, and (iv) train VGG16 with SoftMax and SVM classifiers using the selected five bands for classification of citrus leaves with canker and four other conditions.…”

“…In another study by Zhang et al, 15 they detected yellow rust on winter wheat at an accuracy of 85% using HSI and a DL-based CV algorithm. Similarly, Yadav et al 11,16 used a DL-based CV algorithm with HSI to detect citrus canker and other disease conditions on Ruby Red grapefruit at an average accuracy of 98.87%. Most of the image-based classification and detection tasks using DL rely on convolution neural network (CNN) generated features.…”

“…Selecting optimal bands by reducing the high-dimensional vector space to a reduced one using the PCA technique not only eliminates the issues posed by curse of dimensionality but also provides an opportunity to develop a multi-spectral imaging (MSI) system that is commercially viable option for fruit and leaf inspection. 11,16 The optimal bands selected by the PCA method can be used to train ML and DL-based classification algorithms for detecting CBS-infected fruit and cankerous leaves from healthy ones or with other conditions.…”

Detection of citrus black spot fungus using fluorescence imaging and deep learning on leaf surface

“…Citrus black spot (CBS) is a quarantine fungal disease caused by Phyllosticta citricarpa that was first reported in the U.S.A. in March 2010, in Collier County of Florida and still remains a concern in southwest Florida [1]. This disease can not only limit market access for citrus fruits but also the lesions caused by it on fruit surfaces may lead to premature drops, reducing yield [1], [2], [3], [4]. Leaf symptoms are not very common for CBS, although the fungus reproduces in leaf litter.…”

“…are sprayed monthly during the growing season i.e., from early May to mid-September [5]. In addition, the current practices of field scouting and collecting leaf and fruit samples for polymerase chain reaction (PCR) tests in the laboratory for identification of CBS fungus is extremely labor intensive, time consuming and a costly process [3], [4]. Therefore, it is desired to develop an automated sensing system which can rapidly detect the fungus on citrus leaves or fruit surfaces before they become symptomatic and start showing lesions.…”