Fruit and Vegetable Scanner for Quality Assessment in Retail

Automatic and precise quality assessment

In retail, the freshness and safety of fresh food must be continuously guaranteed. Fruit and vegetables in particular have a short shelf life that varies and depends on a number of factors. If impending spoilage is not detected in time, the products often end up in the trash. In the EU alone, around 58 million tons of food end up as waste every year – worth around 132 billion euros.

To counteract this, forward-looking strategies regarding the timing of sales and pricing are essential. Continuous quality controls along the supply chain are therefore indispensable. Traditionally, destructive laboratory methods are usually used for this purpose, but these are costly and time-consuming. As a result, quality control approaches using spectral sensor technology are becoming increasingly important.
 

AI-supported quality inspection with hyperspectral imaging

As part of a research project, Fraunhofer IOSB is developing a scanner that uses spectral imaging to capture entire crates of fruit and vegetables and derive AI-supported quality assessments. The aim is to transfer near-infrared spectroscopy, which is well established in research, into robust, practical applications.

The system is designed to enable fast, non-destructive analysis of various quality parameters of fruit and vegetables, thus providing a basis for:

• Objective, automated quality control in goods receipt,
• Predictions of remaining shelf life,
• Data-supported control of sales and logistics,
• Digital data acquisition,
• Rapid determination of non-visible quality characteristics.
   

Using physics instead of gut feeling

Unlike conventional RGB cameras, hyperspectral imaging captures not only three color channels, but also many closely adjacent wavelength ranges. Each pixel thus contains a spectrum of reflected light that serves as a characteristic fingerprint of the material. Since many properties relevant to food (e.g., water, sugar, and dry matter content) exhibit specific absorption patterns in the near-infrared range, quality characteristics such as incipient spoilage or internal damage can be derived from this.

Spectral imaging offers the following advantages over conventional methods for quality control of fruit and vegetables:

• Non-destructive: The goods remain intact.
• Comprehensive: Not only individual samples, but entire crates are scanned.
• Fast: Measurement in seconds.
• Contactless: Hygienic analysis without complex mechanics.

Spectral sensors reveal differences in quality that are not visible to the human eye. This enables a reliable assessment of the degree of ripeness, remaining shelf life, and homogeneity of the goods.

Automated quality assessment using AI

The high information density of spectral image data requires the use of machine learning and modern image processing. Features are extracted from spectral images, linked to reference analyses, and models are trained to automatically recognize quality classes and objects and detect anomalies such as foreign objects or atypical spectra. Through continuous real-world data collection, these models can be continuously improved and adapted to new varieties, products, and requirements. This can result in an adaptive system that can be transferred to other fresh products and process steps in the long term.

The system: multimodal sensor technology, local evaluation

Benefits for retail, industry, and development

In fruit and vegetable wholesale and in the processing industry, quality, predictability, and minimal reject rates are crucial. With the help of the fruit scanner, quality decisions are automated, standardized, and less dependent on subjective assessment than manual visual inspections. This provides retailers with a basis for informed decisions on goods receipt, sorting, and marketing. Problem batches are identified early on, stable goods are used more efficiently, and write-offs and complaints are reduced.

The technology has the potential to replace a significant portion of manual visual and laboratory inspections, reduce staffing requirements, and lower quality costs—while at the same time increasing process and planning reliability.

For plant manufacturers and industry, the technology provides a basis for better sorting and processing systems or the development of new applications in food and agricultural machinery. The combination of standardized measurement conditions, digitally available quality data, and flexible AI models forms the basis for scalable, industrial solutions.

Contact us

Would you like to evaluate spectral imaging for your quality processes or integrate it into your products? As your independent R&D partner, we accompany you through the entire development process: from feasibility studies and data-driven model development to prototype construction, valida­tion, and pilot integration. We develop solutions tailored to your specific requirements and production environment.

Let‘s talk about your quality challenges and explore our technology.