Nondestructive Analysis on Agricultural Products Using Near-Infrared Spectroscopy

The logo of Louisiana Agriculture magazine stands against a white background.

Kun-Jun Han, Buddy Pitman and Wink Alison

Conventional sample analysis is accomplished through laboratory-based chemical analysis procedures that quantify target constituents. However, these procedures involve hazardous chemicals, labor and long processing times. A nondestructive analysis using near infrared spectroscopy (NIRS) realizes rapid and chemical-free assessment on a broad spectrum of samples. The NIRS calibration procedure and analytic application has great potential for agricultural researchers and stakeholders.

Near-infrared light was first found in the 1800s, but it was not until the development of its application in agricultural quality assessment in 1954 that NIRS technology became widely used in industry and academia. Light with wavelengths between 780 and 2,500 nanometers can penetrate an object further than mid-infrared light, which causes various types of vibrations from molecular bonds. These vibrations can be processed by computer software into spectral data unique to each product sample (Figure 1). The recent development of new mathematical algorithms further assisted prediction model development for target chemical constituents. Additionally, with an improved computer interface and optical technology, NIRS instruments are more sophisticated with enhanced analytical accuracy and precision.

Procedure for NIRS calibration development

Determining the quality of agricultural commodities using NIRS is advantageous compared to conventional chemical analysis because it reduces cumbersome and time-consuming chemical analysis procedures. However, NIRS analysis still requires chemical analysis at the initial calibration stage in order to efficiently conduct routine sample analysis. The calibration procedure provides reference samples with known constituent values to the NIRS device to train it.

The calibration procedure begins with collecting at least 150 representative samples covering a product population of interest. The collected samples are then scanned with NIRS to collect the spectral data of samples. A brief workflow describing the calibration procedure is presented in Figure 2.

The collected spectral data is processed using a mathematical algorithm to categorize samples into the subgroups of reference, validation, redundancy and outlier. The redundant samples are eliminated from the calibration model development because of their similarity to the reference samples. This step of identifying redundant samples is another significant advantage of using NIRS to avoid unnecessary chemical analysis. The samples in the outlier group are also eliminated from the calibration model development, but those samples are valuable for later expansion of the calibration model. The reference samples are subject to conventional chemical analysis at a lab and are to be used for calibration model development. The spectral data of reference samples are regressed against the chemical analysis results as the last step in developing a calibration model. Finally, the validation samples are saved to check the calibration models' robustness and potential overfitting.

Any NIRS device installed with a calibration model can then conduct routine analysis of samples at a rate of one sample in less than a minute. While third-party-developed calibration models are available for common nutrient constituents of major agricultural commodities on a subscription or license basis, they often do not include the more specific constituents or characteristics required by research and plant breeding programs. Therefore, calibration models must be developed by researchers. Developing a custom calibration model requires substantial effort at the calibration model development stage; however, it can provide more control to researchers for target constituent settings and flexibility for future refinement of developed calibration models.

Current and future extended applications at the LSU AgCenter

At the LSU AgCenter, NIRS has been used for the assessment of nutrient value, digestibility, and potential intake of forage and feed samples. This function has enabled the LSU AgCenter Forage Analysis lab to serve Louisiana forage and livestock producers by recommending ways to optimize forage feeding and fertilization to boost hay sales and livestock management.

At the LSU AgCenter School of Plant, Environmental and Soil Sciences, expansions of NIRS to row crop and specialty crop analysis are planned through collaborative efforts with researchers so that tailored calibration models can be custom-made for less common analytes, the substances that are being analyzed. The NIRS analysis has great potential value to developers of food and feed products, particularly where new products or processes may require product quality or safety verification.

Kun-Jun Han is an associate professor in the LSU AgCenter School of Plant, Environmental and Soil Sciences in Baton Rouge; William "Buddy Pitman is a professor at Hill Farm Research Station in Homer; and Montgomery "Wink" Alison is an associate professor at the Macon Ridge Research Station in Winnsboro.

This article appeared in the summer 2022 issue of Louisiana Agriculture.

A colorful chart shows data.

The spectrum patterns of pet food, cattle feed and corn mash are charted on analytical software designed for near infrared spectroscopy analysis. Figure adopted from Foss DS2500 demonstration screen.

A workflow chart shows how to examine data.

Workflow of NIRS (near infrared spectroscopy) calibration model development from sample collection to application of calibration model in sample analysis. Figure by Kun-Jun Han

9/16/2022 3:30:58 PM
Rate This Article:

Have a question or comment about the information on this page?

Innovate . Educate . Improve Lives

The LSU AgCenter and the LSU College of Agriculture