Non-Destructive Detection of Pomegranate Blackheart Disease via Near-Infrared Spectroscopy and Soft X-ray Imaging Systems

Abstract

Pomegranate blackheart disease, as an internal disease affecting the global pomegranate industry, is difficult to identify externally and urgently requires non-destructive detection methods for rapid diagnosis. This study established discriminative models for blackheart disease severity in pomegranates by using near-infrared (NIR) spectroscopy and soft X-ray imaging techniques. The results showed that the optimal NIR-based discriminative model, constructed with a Random Forest (RF) algorithm based on spectra preprocessed by the second-derivative (D2) denoising and a Competitive Adaptive Reweighted Sampling (CARS) algorithm, achieved a prediction set accuracy of 86.00%; the optimal soft X-ray imaging-based discriminative model, built with an RF algorithm using textural features extracted from images preprocessed by median filtering and a Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm combined with gray-level co-occurrence matrix (GLCM) and gray-gradient co-occurrence matrix (GGCM) algorithms, reached a prediction set accuracy of 93.10%. In terms of model performance, the model based on soft X-ray imaging exhibited superior performance. Both techniques possess distinct advantages and limitations yet enable non-destructive detection of pomegranate blackheart disease. Further technical optimizations in the future could provide enhanced support for the healthy development of the pomegranate industry.

Keywords: blackheart disease; food non-destructive detection; near-infrared spectroscopy; pomegranate; soft X-ray.

Figure 1

Flowchart of the pomegranate blackheart disease detection experiment.

Figure 2

Cultivated Alternaria alternata.

Figure 3

Cross-sectional views of pomegranates at different days after infection with blackheart disease. (a) Healthy; (b) 1 day after infection; (c) 3 days after infection; (d) 5 days after infection.

Figure 4

Spectral acquisition system.

Figure 5

Soft X-ray imaging device.

Figure 6

The segmentation process of pomegranate soft X-ray images. (a) Original image. (b) The image after threshold segmentation.

Figure 7

Image processing workflow for pomegranate cross-sections: (a) original RGB cross-sectional image (b) HSV color space conversion; (c) seed region extraction; (d–f) morphological operations; (g) seed segmentation; and (h) pathological region segmentation.

Figure 8

Spectral profiles of pomegranate specimens. (a) Absorbance curves of all specimens; (b) mean absorbance curves across infection severity grades.

Figure 9

Soft X-ray image preprocessing (original image, after median filtering and CLAHE), (a–c) healthy specimens, (d–f) Grade 1 infected specimens, (g–i) Grade 2 infected specimens, and (j–l) Grade 3 infected specimens.

Figure 10

Removal of outliers, (a) LOF algorithm, (b) marginal distance algorithm.

Figure 11

The spectral denoising processing results of pomegranate samples. (a) Raw spectral data. (b) Spectral preprocessed by the SNV algorithm. (c) Spectral preprocessed by the MSC algorithm. (d) Spectral preprocessed by the SG algorithm. (e) Spectral preprocessed by the D1 algorithm. (f) Spectral preprocessed by the D2 algorithm.

Figure 12

Process of optimal wavelength selection based on the CARS algorithm.

Figure 13

Confusion matrix of the prediction set for the optimal discrimination model of pomegranate blackheart disease established using NIR technology.

Figure 14

Confusion matrix of the prediction set for the optimal discrimination model of pomegranate blackheart disease established using soft X-ray imaging technology.