ItemArticle

SS-CNN BruiseFinder: Hyperspectral imaging and CNN-driven spatial-spectral fusion for non-destructive plum bruise analysis

Title
SS-CNN BruiseFinder: Hyperspectral imaging and CNN-driven spatial-spectral fusion for non-destructive plum bruise analysis
Creator
K.S., Shanthini; George, Sudhish N.; Francis, Jobin; George, Sony
Description
Plum fruit is susceptible to damage at various stages, from growth to packaging, and such bruising is often difficult to detect visually due to its subtle surface appearance. This research seeks to develop a convolutional neural network (CNN) model that leverages 3D convolutional layers to integrate spatial and spectral features from hyperspectral data, enabling accurate bruise analysis in plum fruit. In this study, plums sourced from a Norwegian fruit store were intentionally bruised and then imaged using hyperspectral technology at various time intervals (30 min to 48 h post-bruising). A novel CNN model, dubbed SS-CNN BruiseFinder, is developed to harness the spatial and spectral characteristics of these hyperspectral images for accurate bruise detection and classification. The SS-CNN BruiseFinder model demonstrates detection accuracy ranging from 68.5% to 91.5% and categorization accuracy between 67.39% and 98.16%. To further establish the effectiveness of this approach, three additional deep learning models a custom spectral CNN, ResNet 101, and a bidirectional LSTM model are developed and evaluated on the same dataset, providing a comprehensive validation of the proposed method's superiority. Timely detection of bruising helps prevent contaminated plums from entering the supply chain during transportation or storage. By categorizing plums based on bruise age, retailers can offer consumers more accurate freshness and quality information, enabling them to make better-informed purchasing choices and ultimately enhancing the overall shopping experience. To encourage community engagement and re-implementation, our code is available at https://github.com/SS-CNN BruiseFinder. 2025 Elsevier Ltd
Source
Food Control;Volume;182;Issue;;Article No.;111870;
Date
01-01-2026
Publisher
Elsevier Ltd
Subject
Bruise categorization; Fruit quality analysis; Postharvest; Spatial spectral CNN; Vis-NIR imaging
Coverage
K.S. S., Department of Electronics and Communication Engineering, National Institute of Technology Calicut, Kerala, India; George S.N., Department of Electronics and Communication Engineering, National Institute of Technology Calicut, Kerala, India; Francis J., Department of Computer Science, Christ University Bangalore, Karnataka, India; George S., Department of Computer Science, Norwegian University of Science and Technology Gjik, Norway
Rights
Restricted Access; Hardcopy may be available in the library
Relation
ISSN: 9567135; CODEN: FOOCE
Format
online
Language
English
Type
Article

Citation

K.S., Shanthini; George, Sudhish N.; Francis, Jobin; George, Sony, “SS-CNN BruiseFinder: Hyperspectral imaging and CNN-driven spatial-spectral fusion for non-destructive plum bruise analysis,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 18, 2026, https://archives.christuniversity.in/items/show/22260.

Output Formats