Abstract:As a key factor in determining the optimal harvest period of Camellia oleifera fruit maturity is significant in increasing camellia oil yield. Due to the interference of environmental factors, capturing hyperspectral images of Camellia oleifera fruit on-site often suffers from a low spectral signal-to-noise ratio. In this study, the multi-head attention parallel residual block-cycle generative adversarial network (MAPRB-CycleGAN) model was used to reconstruct on-site spectral data of Camellia oleifera fruit. Meanwhile, the characteristics and quality of the reconstructed spectra were analyzed through subjective visual inspection of the curve effect and the principal component analysis (PCA) algorithm. Then, the modeling effect of the reconstructed on-site spectral data after being reconstructed by wavelet transform, PCA, and kernel principal component analysis was compared. After the conventional pretreatment method in chemometrics, the partial least squares discriminant analysis (PLS-DA) model established with the field original spectra of Camellia oleifera fruit achieved the highest classification accuracy of 93.82%. The maturity classification model of Camellia oleifera fruit established by the PLS-DA method using the reconstructed spectra from this paper achieved 99.54% accuracy, and the classification accuracy of this data set in multiple models was higher than that of other reconstructed data sets. The results indicate that hyperspectral imaging technique combined with MAPRB-CycleGAN can realize high-accuracy maturity classification of Camellia oleifera fruit, and the spectral reconstruction can be exploited as a preprocessing method for on-site spectral data collection, which provides important technical support for obtaining high-quality spectral data of Camellia oleifera fruit in field environments. The randomness and unpredictability of outdoor environmental factors lead to the diversity in data collection, and enhancing the robustness of generative models to cope with environmental influences is the most significant challenge in practical applications. It will also be the focal point of our future research.

Discrimination of maturity of Camellia oleifera fruit on-site based on generative adversarial network and hyperspectral imaging technique

In-field and non-destructive determination of comprehensive maturity index and maturity stages of Camellia oleifera fruits using a portable hyperspectral imager

Maturity Stage Discrimination of Camellia oleifera Fruit Using Visible and Near-Infrared Hyperspectral Imaging

Study on Identification of Immature Corn Seed Using Hyperspectral Imaging

Maturity Identification of Camellia Seeds Based on Mid- and Far-Infrared Data Fusion

Improving Discrimination Accuracy of Pest-Infested Crabapples Using Vis/NIR Spectral Morphological Features

Enhancing model robustness through different optimization methods and 1-D CNN to eliminate the variations in size and detection position for apple SSC determination

Grade Classification of Camellia Seed Oil Based on Hyperspectral Imaging Technology

Maturity Classification of Rapeseed Using Hyperspectral Image Combined with Machine Learning

Authenticity Discrimination and Adulteration Level Detection of Camellia Seed Oil via Hyperspectral Imaging Technology

Investigating the Identification and Spatial Distribution Characteristics of Camellia oleifera Plantations Using High-Resolution Imagery

Study on Adulteration Detection of Camellia Seed Oil Based on Hyperspectral Imaging Technology

Discrimination of Tomato Maturity Using Hyperspectral Imaging Combined with Graph-Based Semi-supervised Method Considering Class Probability Information

Nondestructive Detection of Sunflower Seed Vigor and Moisture Content Based on Hyperspectral Imaging and Chemometrics

Study on Maturity Discrimination of Hami Melon with Hyperspectral Imaging Technology Combined with Characteristic Wavelengths Selection Methods and Svm

Dual-band feature selection for maturity classification of specialty crops by hyperspectral imaging

CenterNet-LW-SE net: integrating lightweight CenterNet and channel attention mechanism for the detection of Camellia oleifera fruits

Nondestructive quality assessment and maturity classification of loquats based on hyperspectral imaging

Application of Hyperspectral Imaging for Maturity and Soluble Solids Content Determination of Strawberry With Deep Learning Approaches

Quantitative Prediction of Acid Value of Camellia Seed Oil Based on Hyperspectral Imaging Technology Fusing Spectral and Image Features

Rapid and Non-destructive Classification of New and Aged Maize Seeds Using Hyperspectral Image and Chemometric Methods