WSU Tree Fruit Research & Extension Center

Postharvest Information Network

Friday, July 29, 2016

WSU-TFREC/Postharvest Information Network/Near-Infrared Spectroscopy: Background and Application to Tree Fruit Quality Measurements



Near-Infrared Spectroscopy: Background and Application to Tree Fruit Quality Measurements


Background

This paper presents a brief background on near-infrared (NIR) spectroscopy, its modes of use, major issues in the application of NIR to the measurement of tree fruit quality, and the current state of NIR technology.

Near-infrared spectroscopy has been used since the 1970s for the compositional analysis of low moisture food products. However, only in the last 10-15 years has NIR been successfully applied to the analysis of high moisture products such as fruit. NIR is a form of vibrational spectroscopy that is particularly sensitive to the presence of molecules containing C-H (carbon-hydrogen), O-H (oxygen-hydrogen), and N-H (nitrogen-hydrogen) groups. Therefore, constituents such as sugars and starch (C-H), moisture, alcohols and acids (O-H), and protein (N-H) can be quantified in liquids, solids, and slurries. In addition, the analysis of gases (e.g., water vapor, ammonia) is possible. NIR is not a trace analysis technique and it is generally used for measuring components that are present at concentrations greater than 0.1%.


Short-Wavelength NIR vs. Long-Wavelength NIR

NIR has traditionally been carried out in the 1100-2500 nm region of the electromagnetic spectrum. However, the wavelength region of ~700-1100 nm (short wavelength-NIR or SW-NIR) has been gaining increased attention. The SW-NIR region offers numerous advantages for on-line and in-situ bulk constituent analysis. This portion of the NIR is accessible to low-cost, high performance silicon detectors and fiber optics. In addition, high intensity laser diodes and low-cost light emitting diodes are becoming increasingly available at a variety of NIR wavelength outputs.

The relatively low extinction (light absorption) coefficients in the SW-NIR region yield good linearity of absorbance with analyte concentration and permit long, convenient pathlengths to be used. The depth of penetration of SW-NIR is also much greater than that of the longer wavelength NIR, permitting a more adequate sampling of the "bulk" material. This is of particular importance when the sample to be analyzed is heterogeneous.


Diffuse Reflectance Sampling vs. Transmission Sampling

Traditional NIR analysis has used diffuse reflectance sampling. This mode of sampling is convenient for samples that are highly light scattering or samples for which there is no physical means to employ transmission spectroscopy. Diffusely reflected light is light that has entered a sample, undergone multiple scattering events, and emerged from the surface in random directions. A portion of light that enters the sample is also absorbed. The depth of penetration of the light is highly dependent on the sample characteristics and is often affected by the size of particles in the sample and the sample density. Furthermore, diffuse reflectance is biased to the surface of a sample and may not provide representative data for large heterogeneous samples such as apples.

While transmission sampling is typically used for the analysis of clear solutions, it also can be used for interrogating solid samples. A transmission measurement is usually performed with the detector directly opposite the light source (i.e., at 180°) and with the sample in the center. Alternately the detector can be placed closer to the light source (at angles less than 180°), which is often necessary to provide a more easily detected level of light. Because of the long sample pathlengths and highly light scattering nature of most tree fruit, transmission measurements can be performed only in the SW-NIR wavelength region.


NIR Calibration

NIR analysis is largely an empirical method; the spectral lines are difficult to assign, and the spectroscopy is frequently carried out on highly scattering samples where adherence to Beer's Law is not expected. Accordingly, statistical calibration techniques are often used to determine if there is a relationship between analyte concentration (or sample property) and instrument response. To uncover this relationship requires a representative set of "training" or calibration samples. These samples must span the complete range of chemical and physical properties of all future samples to be seen by the instrument.

Calibration begins by acquiring a spectrum of each of the samples. Constituent values for all of analytes of interest are then obtained using the best reference method available with regard to accuracy and precision. It is important to note that a quantitative spectral method developed using statistical correlation techniques can perform no better than the reference method.

After the data have been acquired, computer models employing statistical calibration techniques are developed that relate the NIR spectra to the measured constituent values or properties. These calibration models can be expanded and must be updated periodically and verified using conventional testing procedures.

Factors affecting calibration include fruit type and variety, seasonal and geographical differences, and whether the fruit is fresh or has been in cold storage. Calibration variables include the particular properties or analytes measured and the concentration or level of the properties. Intercorrelations (co-linearity) should be minimized in calibration samples so as not to lead to false interpretation of a model's predictive ability. Co-linearity occurs when the concentrations of two components are correlated (e.g., an inverse correlation exists when one component is high, the other is always low or vice versa).


Application of NIR to Tree Fruit and Existing On-line NIR Instrumentation

A growing body of research exists for NIR analysis of tree fruit. NIR has been used for the measurement of fruit juice, flesh, and whole fruit. In juice, individual sugars (sucrose, fructose, glucose) and total acidity can be quantified with high correlation (0.95) and acceptable error. Individual sugars cannot be readily measured in whole fruit. Soluble solids is the most successfully measured NIR parameter in whole fruit and can generally be achieved with an error of +0.5-1.0° soluble solids. Recent research results indicate firmness and acidity measurement in whole fruit may be possible.

Only in Japan has the large-scale deployment of on-line NIR for fruit sorting occurred. These instruments require manual placement/orientation of the fruit prior to measurement and early versions were limited to a measurement rate of three samples per second. The Japanese NIR instruments are also limited to a single lane of fruit and appear to be difficult to adapt to multi-lane sorting equipment used in the United States. While earlier Japanese NIR instruments employed reflectance sampling, more recent instruments may be using transmission sampling. Efforts to procure additional information about this equipment have not been successful.

Research groups around the world continue to explore the applications of near infrared spectroscopy to tree fruit. Berkeley Instruments, Inc. is currently funded by the Washington State Tree Fruit Research Commission for a Phase 2 NIR project entitled "Nondestructive Determination of Apple Quality Using Near-Infrared Spectroscopy." Preliminary results indicate both firmness and soluble solids can be measured simultaneously in multiple apple cultivars using a single calibration equation with errors of +1-2 lb. and +1-1.5° soluble solids. The ultimate goal of this work is to produce both portable and on-line NIR analyzers for the simultaneous measurement of multiple quality parameters of fruit.

NIR technology will likely play an important future role as another tool for grading fruit quality. The unique ability of NIR statistical calibration techniques to extract non-chemical "properties" may provide a means for development of a general NIR "quality index" for tree fruit. This general "quality index" would combine all of the information that could be extracted from the NIR spectra and may include information about soluble solids, acidity, firmness, and internal disorders.


Address

Dr. Richard M. Ozanich, Jr.
Berkeley Instruments, Inc.
3100 George Washington Way, Suite 104
Richland WA 99352

References on Near-Infrared Spectroscopy can be found in the 1998 Washington Tree Fruit Postharvest Conference Proceedings. A copy can be ordered from Mackey@wahort.org.

Dr. Richard M. Ozanich, Jr.

Berkeley Instruments

Tree Fruit Postharvest Journal 10(1):18-19
February 1999

Dr. Richard M. Ozanich, Jr.

Berkeley Instruments

Tree Fruit Postharvest Journal 10(1):18-19
February 1999

Dr. Richard M. Ozanich, Jr.

Berkeley Instruments

Tree Fruit Postharvest Journal 10(1):18-19
February 1999

Dr. Richard M. Ozanich, Jr.

Berkeley Instruments

Tree Fruit Postharvest Journal 10(1):18-19
February 1999

Tree Fruit Research & Extension Center, 1100 N Western Ave, Washington State University, Wenatchee WA 98801, 509-663-8181, Contact Us