Classification and Discrimination of Different Fungal Diseases of Three Infection Levels on Peaches Using Hyperspectral Reflectance Imaging Analysis

Abstract

Peaches are susceptible to infection from several postharvest diseases. In order to control disease and avoid potential health risks, it is important to identify suitable treatments for each disease type. In this study, the spectral and imaging information from hyperspectral reflectance (400~1000 nm) was used to evaluate and classify three kinds of common peach disease. To reduce the large dimensionality of the hyperspectral imaging, principal component analysis (PCA) was applied to analyse each wavelength image as a whole, and the first principal component was selected to extract the imaging features. A total of 54 parameters were extracted as imaging features for one sample. Three decayed stages (slight, moderate and severe decayed peaches) were considered for classification by deep belief network (DBN) and partial least squares discriminant analysis (PLSDA) in this study. The results showed that the DBN model has better classification results than the classification accuracy of the PLSDA model. The DBN model based on integrated information (494 features) showed the highest classification results for the three diseases, with accuracies of 82.5%, 92.5%, and 100% for slightly-decayed, moderately-decayed and severely-decayed samples, respectively. The successive projections algorithm (SPA) was used to select the optimal features from the integrated information; then, six optimal features were selected from a total of 494 features to establish the simple model. The SPA-PLSDA model showed better results which were more feasible for industrial application. The results showed that the hyperspectral reflectance imaging technique is feasible for detecting different kinds of diseased peaches, especially at the moderately- and severely-decayed levels.

Application of Deep Learning in Food: A Review

“…The constraints of deep learning-based APP were pointed out by Sun, Wei, Liu, Pan, and Tu (2018). They aimed at developing a system for detection of diseased peaches using spectral analysis.…”

Research Progress on the Early Monitoring of Pine Wilt Disease Using Hyperspectral Techniques

“…The white frame was obtained by measuring white reference objects with reflectance close to 100%, such as barium sulfate, magnesium oxide, and white reference tiles. Black frames were obtained by closing the camera cover to obtain 0% reflectivity [ 99 , 100 ]. Black and white frames can be collected in the field by similar means, and a unified standard hyperspectral image is obtained after calibration.…”

A Hyperspectral Imaging Approach for Classifying Geographical Origins of Rhizoma Atractylodis Macrocephalae Using the Fusion of Spectrum-Image in VNIR and SWIR Ranges (VNIR-SWIR-FuSI)

“…In this work, supervised pattern recognition models were adopted for the origin identification of RAM slices. There have been a variety of models available for classification, including partial least square-discriminant analysis (PLS-DA) [34,35], linear discriminate analysis (LDA) [36], support vector machine (SVM) [8] and back propagation neural network (BPNN) [5]. PLS-DA and SVM were selected herein.…”