Rapid quality control of fruits and vegetables using NIR spectroscopy

Importance of the topic

Near-infrared (NIR) spectroscopy offers a rapid, nondestructive approach for assessing critical fresh-produce quality attributes such as firmness and soluble-solids content (°Brix). These attributes determine consumer acceptance and are routinely measured with destructive laboratory techniques (e.g., penetrometry for texture and refractometry for soluble solids). Implementing robust NIR methods supports faster incoming-inspection, in-line process control, and reduced sampling effort while preserving sample integrity.

Objectives and study overview

This application note evaluated Metrohm NIR spectroscopy for rapid quantification of: firmness in apples, °Brix in peaches, and °Brix in crushed tomatoes. The goal was to develop calibration models and compare NIR predictions with conventional laboratory reference methods to assess accuracy and feasibility for routine quality control in processing environments.

Methodology and used instrumentation

Measurements were acquired in reflection mode using a Metrohm near-infrared (NIR) spectrometer. Samples were measured either directly (without a holder) or in a large sample cup to collect reflectance spectra. Metrohm software handled data acquisition and the development of chemometric quantification models. Reference methods cited for comparison include penetrometer measurements for firmness and ISO 2173:2003 refractometric determination of soluble solids (°Brix).

Main results and discussion

• Apples — Firmness: Calibration yielded an R2 of approximately 0.80 with standard error of calibration (SEC) ≈ 0.93 N and standard error of cross‑validation (SECV) ≈ 0.95 N. Note: only a calibration dataset was used; no external validation set was reported.
• Peaches — °Brix: Calibration achieved R2 ≈ 0.80 with SEC ≈ 0.53 °Bx and SECV ≈ 0.58 °Bx. As with apples, results were based on the calibration dataset only.
• Crushed tomatoes — °Brix: Stronger performance was observed with R2 ≈ 0.85, SEC ≈ 0.14 °Bx, SECV ≈ 0.14 °Bx and SEP (external prediction error) ≈ 0.16 °Bx. This dataset included an independent validation set (predictions shown as separate points), demonstrating transferability beyond calibration data.

These outcomes demonstrate that NIR spectra correlate well with traditional quality indicators. The tomato model, with an independent validation, indicates higher predictive reliability when external validation is performed. For apples and peaches, the absence of a validation set suggests the need for further testing across varieties, harvest seasons, and sample conditions to confirm robustness.

Benefits and practical applications

• Nondestructive, rapid analysis enabling higher-throughput quality screening of incoming and processed produce.
• Reduction of labor and consumables compared with destructive reference methods (penetrometry, refractometry).
• Potential integration into production lines for sorting, grading, and process control (e.g., juice or puree blending based on °Brix).
• Ability to build product-specific calibrations for on-site decision making (accept/reject, ripeness classification, blending targets).

Limitations and considerations

• Model robustness depends on calibration sample diversity (cultivar, maturity, growing conditions, seasonality); limited calibration sets can inflate apparent performance.
• Apples and peaches in this study lacked independent validation, so reported errors may be optimistic until externally validated.
• Measurement geometry (direct vs. cup) and sample presentation influence spectra; standardized sampling protocols improve reproducibility.
• Routine deployment requires periodic model maintenance and validation to accommodate variability in raw material.

Future trends and applications

• Wider adoption of portable and handheld NIR instruments for field and store-level quality checks.
• Integration with hyperspectral imaging or multispectral cameras for spatially resolved quality mapping and defect detection.
• Advanced chemometrics and machine-learning strategies (transfer learning, model updating, cloud-hosted calibration libraries) to improve robustness across varieties and sites.
• Inline and at-line process control implementations with automated sorting and blending guided by real-time NIR measurements.
• Standardized validation protocols and shared calibration datasets to accelerate cross-site method transfer and regulatory acceptance.

Conclusion

This application demonstrates that Metrohm NIR spectroscopy can deliver rapid, reproducible estimates of apple firmness and °Brix in peaches and crushed tomatoes, offering a practical, nondestructive alternative to traditional laboratory assays. While the tomato model included external validation and showed particularly strong performance, apple and peach models require additional validation across broader sample sets to ensure operational robustness. When properly validated and maintained, NIR-based workflows can streamline fresh-produce quality control and support real-time decision making in processing environments.

Reference

• ISO 2173:2003 Fruit and vegetable products — Determination of soluble solids — Refractometric method.

Used instrumentation

• Metrohm near-infrared (NIR) spectrometer operated in reflection mode.
• Sample presentation: large sample cup or direct measurement (no holder) depending on sample type.
• Metrohm software for spectral acquisition and chemometric model development.