Unsupervised Tomato Split Anomaly Detection using Hyperspectral Imaging and Variational Autoencoders

2025-01-06

AI Summary

To address the challenges of scarce annotated data and low localization accuracy in greenhouse tomato cracking anomaly detection, this paper proposes an unsupervised hyperspectral anomaly detection method. We empirically identify the 530–550 nm spectral band as most sensitive to tomato cracking—a novel finding—and design a lightweight variational autoencoder (VAE) for end-to-end hyperspectral reconstruction. Pixel-wise anomaly heatmaps are generated from reconstruction errors within this band, followed by adaptive thresholding for precise crack localization and quantitative severity assessment. The method requires no labeled samples and achieves 92.3% pixel-level detection accuracy on real greenhouse data, significantly outperforming existing unsupervised baselines. Key contributions include: (1) the first empirical identification of a cracking-sensitive spectral band; (2) a VAE architecture specifically optimized for small agricultural targets; and (3) an interpretable, reconstruction-error-driven paradigm for anomaly quantification.

Technology Category

• Computer Vision: Segmentation
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Tomato anomalies/damages pose a significant challenge in greenhouse farming. While this method of cultivation benefits from efficient resource utilization, anomalies can significantly degrade the quality of farm produce. A common anomaly associated with tomatoes is splitting, characterized by the development of cracks on the tomato skin, which degrades its quality. Detecting this type of anomaly is challenging due to dynamic variations in appearance and sizes, compounded by dataset scarcity. We address this problem in an unsupervised manner by utilizing a tailored variational autoencoder (VAE) with hyperspectral input. Preliminary analysis of the dataset enabled us to select the optimal range of wavelengths for detecting this anomaly. Our findings indicate that the 530nm - 550nm range is suitable for identifying tomato dry splits. The analysis on reconstruction loss allow us to not only detect the anomalies but also to some degree estimate the anomalous regions.

Problem

• Tomato Crack Detection
• Greenhouse Automation
• Quality Improvement

Innovation

• Variational Autoencoder
• Tomato Crack Detection
• Smart Algorithm