Near-Infrared Diffuse Reflection Analysis of Fruit

The quality of fruit and produce cannot be judged by appearance alone. A colorful mango or vibrant avocado must do more than add a splash of colour to the plate. Taste and freshness are critical. Since it is not practical to determine fruit quality by tasting each individual piece, an objective, nondestructive measurement is needed to determine the quality of the fruit hidden beneath the peel. In this application note, near-infrared (NIR) diffuse reflection spectra are measured for avocados and mangoes to demonstrate the power of NIR measurements for the noninvasive assessment of fruit quality.

Background

Consumers use many techniques to assess fruit quality including smell, firmness, sound, appearance and even intuition. Everyone has their own special approach, with mixed results. While these qualitative techniques are sufficient for consumers, commercial fruit growers and packers require a quantitative approach to determining fruit quality to ensure customer satisfaction and retain or expand market share (consider that annual worldwide production of mangoes alone exceeds 62 million metric tons). Determination of critical quality parameters such as sugar, starch and moisture content requires a rapid, noninvasive, online measurement to test fruit prior to picking or packaging. NIR spectroscopy meets these requirements.

NIR spectroscopy has been used since the 1970s for the analysis of agricultural products. Spectral data for NIR light reflected from an agricultural sample like grain or produce is acquired and compared with a calibration model generated from spectral data acquired for samples with known levels of the constituents of interest. For fruit or produce covered by a peel, the longer wavelengths used for NIR analysis are weakly absorbed so they pass through the peel, enabling sampling of the fruit pulp beneath. The measurement of NIR reflection is rapid (no sample preparation is required), quantitative (with the assistance of carefully constructed calibration models) and nondestructive. Since its initial use as an industrial tool in agriculture, NIR spectroscopy has grown significantly in areas ranging from materials analysis to pharmaceutical monitoring and medical diagnostics.

The NIR region extends from 780-2500nm. Absorption of light in this region causes molecules to vibrate. These molecular vibrations result in spectral data with features dependent on the chemical composition of the sample. In the case of agricultural samples, the NIR spectra are typically composed of broad peaks due to overlapping absorptions caused by overtones and combinations of vibrational modes of organic functional groups like C-H, O-H and N-H chemical bonds. The NIR spectrum provides a snapshot of the sample with information for multiple components available in a single NIR spectrum. These characteristics and others make modern NIR spectroscopy instrumentation ideal for online monitoring and process control.

Starch and sugar (primarily fructose, glucose and sucrose) are commonly measured to determine fruit maturity and quality. While the peaks for these constituents are located near one another, starch has some specific wavelengths that enable construction of a multi-parametric model for determination of fruit quality based on starch components. An extended range NIR spectrometer like the Ocean Optics NIRQuest256-2.5 is a great option for these measurements because it can detect critical starch peaks near 1722nm, 2100nm and 2139nm, as well as sugar peaks that occur primarily between 900-1200nm (some peaks also occur >2100nm). The NIRQuest256-2.5 enables detection of these wavelengths in a single spectrum.

In addition to the spectrometer, a bright light source like the Ocean Optics Vivo Light Source is necessary. Vivo has four powerful tungsten halogen bulbs and fibers that transmit light efficiently for effective NIR measurements of fruit. Since much of the light will scatter off the surface of the fruit, a large core diameter fiber (600 microns) is recommended for these measurements to increase throughput and improve sensitivity.

Sampling configuration is critical for these measurements. In addition to the light lost by scattering from the surface of the fruit, water in the fruit will absorb NIR wavelengths. In addition, the constituents of fruit (or any natural or agricultural products) are not uniformly distributed within the sample. Sampling over a large surface area of the fruit is recommended to provide an average value for the constituents in the fruit. A light source with a large illumination area is a great option for sample illumination when testing fruit and produce.

While the results reported here are qualitative, a carefully constructed chemometrics model is required for a multi-parameter, quantitative assessment of fruit quality. With a good set of reference spectra and PLS (partial least squares) modeling, it is possible to develop a calibration model that can be used to measure multiple fruit parameters (sugar, starch and other fruit constituents) for the prediction of fruit quality. The ability to quantitatively measure multiple parameters simultaneously makes NIR spectroscopy a powerful tool for the agricultural industry.

Measurement Conditions

NIR spectra were measured for avocados and mangoes using the NIRQuest256-2.5 extended range NIR spectrometer (900-2500nm) and Vivo direct illuminated reflection stage. The Vivo provided bright, diffuse illumination with four tungsten halogen bulbs illuminating the sample. A 2-meter VIS-NIR fiber with a 600 micron core diameter arranged at a 45 degree angle relative to the tungsten halogen bulbs was used for the measurement of diffuse reflection from the fruit. Reference measurements were made with a diffuse reflection standard. The dark measurement was made with the lamps turned on and an empty optical stage. The stage was shielded from overhead illumination during the dark measurement with a black shroud. The setup used for the measurements is shown in Figure 1.

Figure 1: An NIR spectrometer used with a powerful tungsten halogen light source and optical stage provides a convenient setup for diffuse reflection analysis of fruit.

NIR diffuse reflection was measured for whole fruit samples with measurements made at four different locations on the fruit. The fruit was placed on the magnetic ring of the optical stage to keep the fruit from rolling off the stage. Multiple measurements were made due to the variable nature of fruit - bruising, non-uniformity in colour and differences in sugar content (due to differences in sun exposure) all lead to NIR spectral differences. To account for the inhomogeneity and variations in the fruit surface, many more measurements should be made at different points on the fruit surface.

Results

NIR diffuse reflection measurements were made for whole ripe and unripe mangoes and avocados. In Figure 2, the average of the spectra measured at four locations on each piece of fruit is shown for two mangoes and two avocados. Multiple spectra (n=4) were recorded for each piece of fruit to account for the inhomogeneity of the fruit. These spectra demonstrate that even spectra for the same fruit type show variability across the spectral region, with the avocado more consistent in the region above ~1100nm.  While the spectral features are similar for both types of fruit, differences in magnitude are observed throughout the spectra. Sampling additional locations on the surface of the fruit would help to average out variability for a given piece of fruit and improve the accuracy and repeatability of the results. Fortunately, the speed of the NIR technique provides the option to sample a larger surface area of the fruit with multiple measurements without the need for lengthy measurement times.

Figure 2: NIR diffuse reflection measurements of mangoes and avocados reveal spectral variability from sample to sample.

Notably, spectral features observed in these diffuse reflection spectra arise from a combination of phenomena depending on the amount of light scattered from the surface of the fruit and the penetration depth for NIR light into the sample. Light that is not scattered by the surface of the fruit passes through the peel and enters the fruit where it can be absorbed based on chemical composition. While diffuse reflection measurements are relatively straightforward to make, diffuse reflection from a variable rounded sample like a piece of fruit results in a complicated spectrum requiring carefully constructed interpretation models to extract quantitative information.

Also, spectra for peeled versus unpeeled ripe avocados and mangoes were captured. In the case of the avocado, spectral features were more pronounced for the peeled avocado than the unpeeled avocado. This may result from less reflection of light by the peel, which increases absorption based on the chemical composition of the avocado.

For mangoes, the effect of peeling the fruit is similar to the spectral differences observed for peeled and unpeeled avocados. But the impact of removing the peel is not as significant for the mangoes because there is less smoothing due to light reflected from the peel (Figure 3). The spectral differences observed when an avocado or mango is peeled suggest that different fruit peels have different properties that impact the overall fruit spectrum either through chemical composition or reflection properties.

Figure 3: Differences in spectral features of peeled and unpeeled ripened mangoes may result from less reflection of the light by the peel.

The degree of fruit ripening also can be observed in NIR analysis. The spectra for ripe versus unripe avocados and mangoes were captured. The unripe avocado spectra are very consistent in the region from ~1900-2500nm; the spectrum for the ripe avocado in this region flattens relative to the unripe avocado (Figure 4).

Figure 4: Spectral features for unripe avocado are very consistent in the region from ~1900-2500nm

For mangoes, the ripe versus unripe sample spectra are very similar with only very slight differences occurring between ~1900-2500nm. These differences are most likely related to differences in sugar and starch content as the fruit matures. While many of the spectral changes are very subtle, a carefully constructed calibration model and a good sampling approach could be used to extract more quantitative information on fruit maturity from these spectra.

Conclusions

NIR spectroscopy is a powerful measurement tool for the characterization of agricultural samples. In the case of fruit, long NIR wavelengths where absorption is weak allow sampling through the peel of the fruit. This also makes sample preparation unnecessary. Combine these advantages with the ability to make rapid measurements and NIR spectroscopy becomes a great option for at-line measurement of fruit.

While the spectral data shown here illustrate the qualitative differences between avocados and mangoes at different stages of maturity, more quantitative information on fruit quality could be extracted from these spectra using an appropriate chemometric model and careful sampling to account for the inhomogeneity of the fruit.

References

Near-infrared Spectroscopy in Food Analysis, Brian G. Osborne, Encyclopedia of Analytical Chemistry, 1986

The quality of fruit and produce cannot be judged by appearance alone. A colorful mango or vibrant avocado must do more than add a splash of colour to the plate. Taste and freshness are critical. Since it is not practical to determine fruit quality by tasting each individual piece, an objective, nondestructive measurement is needed to determine the quality of the fruit hidden beneath the peel. In this application note, near-infrared (NIR) diffuse reflection spectra are measured for avocados and mangoes to demonstrate the power of NIR measurements for the noninvasive assessment of fruit quality.

Background

Consumers use many techniques to assess fruit quality including smell, firmness, sound, appearance and even intuition. Everyone has their own special approach, with mixed results. While these qualitative techniques are sufficient for consumers, commercial fruit growers and packers require a quantitative approach to determining fruit quality to ensure customer satisfaction and retain or expand market share (consider that annual worldwide production of mangoes alone exceeds 62 million metric tons). Determination of critical quality parameters such as sugar, starch and moisture content requires a rapid, noninvasive, online measurement to test fruit prior to picking or packaging. NIR spectroscopy meets these requirements.

NIR spectroscopy has been used since the 1970s for the analysis of agricultural products. Spectral data for NIR light reflected from an agricultural sample like grain or produce is acquired and compared with a calibration model generated from spectral data acquired for samples with known levels of the constituents of interest. For fruit or produce covered by a peel, the longer wavelengths used for NIR analysis are weakly absorbed so they pass through the peel, enabling sampling of the fruit pulp beneath. The measurement of NIR reflection is rapid (no sample preparation is required), quantitative (with the assistance of carefully constructed calibration models) and nondestructive. Since its initial use as an industrial tool in agriculture, NIR spectroscopy has grown significantly in areas ranging from materials analysis to pharmaceutical monitoring and medical diagnostics.

The NIR region extends from 780-2500nm. Absorption of light in this region causes molecules to vibrate. These molecular vibrations result in spectral data with features dependent on the chemical composition of the sample. In the case of agricultural samples, the NIR spectra are typically composed of broad peaks due to overlapping absorptions caused by overtones and combinations of vibrational modes of organic functional groups like C-H, O-H and N-H chemical bonds. The NIR spectrum provides a snapshot of the sample with information for multiple components available in a single NIR spectrum. These characteristics and others make modern NIR spectroscopy instrumentation ideal for online monitoring and process control.

Starch and sugar (primarily fructose, glucose and sucrose) are commonly measured to determine fruit maturity and quality. While the peaks for these constituents are located near one another, starch has some specific wavelengths that enable construction of a multi-parametric model for determination of fruit quality based on starch components. An extended range NIR spectrometer like the Ocean Optics NIRQuest256-2.5 is a great option for these measurements because it can detect critical starch peaks near 1722nm, 2100nm and 2139nm, as well as sugar peaks that occur primarily between 900-1200nm (some peaks also occur >2100nm). The NIRQuest256-2.5 enables detection of these wavelengths in a single spectrum.

In addition to the spectrometer, a bright light source like the Ocean Optics Vivo Light Source is necessary. Vivo has four powerful tungsten halogen bulbs and fibers that transmit light efficiently for effective NIR measurements of fruit. Since much of the light will scatter off the surface of the fruit, a large core diameter fiber (600 microns) is recommended for these measurements to increase throughput and improve sensitivity.

Sampling configuration is critical for these measurements. In addition to the light lost by scattering from the surface of the fruit, water in the fruit will absorb NIR wavelengths. In addition, the constituents of fruit (or any natural or agricultural products) are not uniformly distributed within the sample. Sampling over a large surface area of the fruit is recommended to provide an average value for the constituents in the fruit. A light source with a large illumination area is a great option for sample illumination when testing fruit and produce.

While the results reported here are qualitative, a carefully constructed chemometrics model is required for a multi-parameter, quantitative assessment of fruit quality. With a good set of reference spectra and PLS (partial least squares) modeling, it is possible to develop a calibration model that can be used to measure multiple fruit parameters (sugar, starch and other fruit constituents) for the prediction of fruit quality. The ability to quantitatively measure multiple parameters simultaneously makes NIR spectroscopy a powerful tool for the agricultural industry.

Measurement Conditions

NIR spectra were measured for avocados and mangoes using the NIRQuest256-2.5 extended range NIR spectrometer (900-2500nm) and Vivo direct illuminated reflection stage. The Vivo provided bright, diffuse illumination with four tungsten halogen bulbs illuminating the sample. A 2-meter VIS-NIR fiber with a 600 micron core diameter arranged at a 45 degree angle relative to the tungsten halogen bulbs was used for the measurement of diffuse reflection from the fruit. Reference measurements were made with a diffuse reflection standard. The dark measurement was made with the lamps turned on and an empty optical stage. The stage was shielded from overhead illumination during the dark measurement with a black shroud. The setup used for the measurements is shown in Figure 1.

Figure 1: An NIR spectrometer used with a powerful tungsten halogen light source and optical stage provides a convenient setup for diffuse reflection analysis of fruit.

NIR diffuse reflection was measured for whole fruit samples with measurements made at four different locations on the fruit. The fruit was placed on the magnetic ring of the optical stage to keep the fruit from rolling off the stage. Multiple measurements were made due to the variable nature of fruit - bruising, non-uniformity in colour and differences in sugar content (due to differences in sun exposure) all lead to NIR spectral differences. To account for the inhomogeneity and variations in the fruit surface, many more measurements should be made at different points on the fruit surface.

Results

NIR diffuse reflection measurements were made for whole ripe and unripe mangoes and avocados. In Figure 2, the average of the spectra measured at four locations on each piece of fruit is shown for two mangoes and two avocados. Multiple spectra (n=4) were recorded for each piece of fruit to account for the inhomogeneity of the fruit. These spectra demonstrate that even spectra for the same fruit type show variability across the spectral region, with the avocado more consistent in the region above ~1100nm.  While the spectral features are similar for both types of fruit, differences in magnitude are observed throughout the spectra. Sampling additional locations on the surface of the fruit would help to average out variability for a given piece of fruit and improve the accuracy and repeatability of the results. Fortunately, the speed of the NIR technique provides the option to sample a larger surface area of the fruit with multiple measurements without the need for lengthy measurement times.

Figure 2: NIR diffuse reflection measurements of mangoes and avocados reveal spectral variability from sample to sample.

Notably, spectral features observed in these diffuse reflection spectra arise from a combination of phenomena depending on the amount of light scattered from the surface of the fruit and the penetration depth for NIR light into the sample. Light that is not scattered by the surface of the fruit passes through the peel and enters the fruit where it can be absorbed based on chemical composition. While diffuse reflection measurements are relatively straightforward to make, diffuse reflection from a variable rounded sample like a piece of fruit results in a complicated spectrum requiring carefully constructed interpretation models to extract quantitative information.

Also, spectra for peeled versus unpeeled ripe avocados and mangoes were captured. In the case of the avocado, spectral features were more pronounced for the peeled avocado than the unpeeled avocado. This may result from less reflection of light by the peel, which increases absorption based on the chemical composition of the avocado.

For mangoes, the effect of peeling the fruit is similar to the spectral differences observed for peeled and unpeeled avocados. But the impact of removing the peel is not as significant for the mangoes because there is less smoothing due to light reflected from the peel (Figure 3). The spectral differences observed when an avocado or mango is peeled suggest that different fruit peels have different properties that impact the overall fruit spectrum either through chemical composition or reflection properties.

Figure 3: Differences in spectral features of peeled and unpeeled ripened mangoes may result from less reflection of the light by the peel.

The degree of fruit ripening also can be observed in NIR analysis. The spectra for ripe versus unripe avocados and mangoes were captured. The unripe avocado spectra are very consistent in the region from ~1900-2500nm; the spectrum for the ripe avocado in this region flattens relative to the unripe avocado (Figure 4).

Figure 4: Spectral features for unripe avocado are very consistent in the region from ~1900-2500nm

For mangoes, the ripe versus unripe sample spectra are very similar with only very slight differences occurring between ~1900-2500nm. These differences are most likely related to differences in sugar and starch content as the fruit matures. While many of the spectral changes are very subtle, a carefully constructed calibration model and a good sampling approach could be used to extract more quantitative information on fruit maturity from these spectra.

Conclusions

NIR spectroscopy is a powerful measurement tool for the characterization of agricultural samples. In the case of fruit, long NIR wavelengths where absorption is weak allow sampling through the peel of the fruit. This also makes sample preparation unnecessary. Combine these advantages with the ability to make rapid measurements and NIR spectroscopy becomes a great option for at-line measurement of fruit.

While the spectral data shown here illustrate the qualitative differences between avocados and mangoes at different stages of maturity, more quantitative information on fruit quality could be extracted from these spectra using an appropriate chemometric model and careful sampling to account for the inhomogeneity of the fruit.

References

Near-infrared Spectroscopy in Food Analysis, Brian G. Osborne, Encyclopedia of Analytical Chemistry, 1986