Optical characterization of citrus fruit for quality monitoring with Vis/NIR spectroscopy

The high nutritional value and health benefits of citrus fruit make it a widely consumed food. To maximize customer satisfaction and minimize food waste, the citrus industry demands for non-destructive and efficient quality detection methods. Optical techniques have been proposed as interesting candidates to build detection models for quality monitoring of citrus fruit. However, these detection models have been reported to lack robustness against the biological variability between fruit from different growing seasons, production regions and even orchards. One way to increase this robustness would be to include more samples with different attributes to cover a wider variability and thus enhance the robustness of the detection model. However, this turns out to be challenging in practice. Tracing back to the nature of the light-matter interaction, it can be found that the light propagation in citrus fruit and the acquired spectra are the result of a complex interplay between chemical components-related absorption and physical structures-related scattering phenomena. Therefore, it is hypothesized that separating the bulk optical properties (BOP) that are related to the chemical and physical quality of citrus tissues will contribute to building more robust quality detection models and exploring the light propagation in citrus fruit to optimize the design of optical sensors for citrus fruit quality measurement. In a first experiment, the BOP were characterized for different citrus fruit species, including grapefruit, orange, lemon and lime. Since citrus fruit is composed of different tissue layers: flavedo, albedo and juice vesicles, the BOP measurements were performed on each of these tissue layers. The diffuse reflectance and total transmittance measured from double integrating spheres (DIS), together with the unscattered transmittance measurement, were combined with the inverse adding-doubling (IAD) routine to estimate the BOP. Simultaneously, the size of the oil glands in flavedo, the soluble solids content (SSC) and the titratable acidity (TA) in juice vesicles were respectively measured with traditional destructive methods. Based on these measurements, the relationships between the BOP and the quality properties were investigated. The bulk absorption coefficient (μa) spectra in the near-infrared (NIR) range were found to be dominated by the absorption peaks of water around 980, 1200 and 1450 nm for all the species and tissue layers. In the visible (Vis) range, absorption peaks attributed to carotenoids were found at around 480 nm for all the species, while a specific absorption peak attributed to chlorophyll was also found for lime at around 680 nm. Differences in the bulk scattering coefficient (μs) were observed among different tissue layers, with the μs value of peel (flavedo and albedo) much higher than the flesh (juice vesicles). The changes in anisotropy factor (g) with wavelength varied among different tissue layers. However, the differences in μs and g between different species were typically smaller than the variation among different tissues. A negative correlation was found between the average value of μs and the diameter of the oil glands in the flavedo. The best prediction models for SSC and TA were respectively achieved based on the attenuation coefficient (μt) and μa of juice vesicles. After obtaining the BOP of different tissue layers, the light propagation in citrus fruit was investigated with Monte Carlo multi-layered (MCML) simulations. The light propagation was traced in the wavelength range from 490 nm to 1880 nm for source-detector distances up to 1 cm. The simulated light energies in different detection modes (including diffuse reflectance, absorption and transmittance), absorbed by different tissue layers and their contributions to the diffuse reflectance profiles were recorded. Next, the simulated spatially resolved diffuse reflectance profiles for the intact fruit were compared to those extracted from hyperspectral laser scatter images to validate the accuracy of the simulations. Apart from the deviations caused by the rough surface and nearby oil glands, the simulated results generally matched well with the measurements. Simulation results indicated that most of the light was reflected and absorbed by the flavedo. Around 15 % of the light was absorbed by the albedo for all the wavelengths outside the absorption bands of water and flavedo pigments. Compared to the flavedo and albedo, the light absorbed by juice vesicles was extremely low with fractions mostly below 3 %. The flavedo and albedo layers were found to contribute mostly to the diffuse reflectance spectra at 800 nm and their contributions respectively decreased and increased with increasing source-detector distance. Although the contribution of the juice vesicles increased with distance, the maximal diffuse reflectance intensity contributed by juice vesicles was located within 0.56 cm from the illumination point. Based on these results, the combination of the power intensity and the source detector distance can be optimized to obtain maximal information about the layer of interest. In a next phase, the influences from harvest time, fruit size and cultivar on the BOP of Satsuma mandarin tissues were investigated. As the diffuse reflectance profiles mostly reflected the properties of the outmost flavedo layer while many qualities of interest (e.g., sugar and acid contents) are related to the innermost juice vesicles layer, the variation of the BOP in relation to these influencing factors was explored for both the inner (juice vesicles) and outer (flavedo) tissue layers. Satsuma mandarin of three different cultivars (Iwasaki, Okitsu and Goku Wase) harvested at three different times around the recommended harvest date were used as the samples for this experiment. Similar to the above measurements on the citrus fruit with different species, the BOP, SSC, TA as well as the color parameters (L*, a* and b*) and Brix minus acid (BrimA) were measured for the Satsuma mandarins. Along the harvest time, the absorption peak related to carotenoids in the μa spectrum was found to increase in both tissue layers, while the values attributed to chlorophylls decrease in the flavedo. Moreover, a slightly increasing trend over the harvest time was observed in the μs spectra of juice vesicles, especially at the early stage. The μa values of both tissue layers and the μs values of juice vesicles deviated among different fruit sizes, which can be related to the development stage. Smaller fruit presented a higher carotenoids content in juice vesicles, an advanced color change in flavedo throughout the maturation stages and higher μs values in juice vesicles at earlier maturation stages. The μa differences in pigment absorption among different cultivars were related to the maturation order. Earlier maturing cultivars possessed higher μa values related to carotenoids absorption in juice vesicles and advanced color changes in flavedo. The μs values of the flavedo differed among the cultivars and these differences were enhanced during maturation. BOP values can also be extracted from the non-destructive hyperspectral laser scatter images using a metamodel to link the BOP with the diffuse reflectance profiles. While measurement-based metamodels have been reported to provide a good match with the measured data, simulation-based metamodels would provide more flexibility in designing the search space. Therefore, the mismatches between measurements and simulations were explored by studying the influences from light source, the incident angle and the detection angle on the diffuse reflectance profiles as measured with a hyperspectral laser scatter imaging setup. Results indicated that the diffuse reflectance intensity was in proportion with the light source power and the influence from the size of the light beam diminished with increasing source-detector distance. The profiles extracted from the sides perpendicular to the illuminating beam were found to be less influenced by the incident angle. Moreover, taking into account the correct region of interest and incident angle, the mismatch between measurements and simulations could be reduced further. Within a detection angle range from 0 ° to 30 ° relative to the normal of the sample surface, the relative deviations between measurements and simulations at source-detector distances from 0.04 to 0.07 cm were below 6 %. After reducing the mismatches caused by the factors mentioned above, the simulated diffuse reflectance profiles matched well with the measurements, with R2-values above 0.99. Successively, metamodels were respectively built based on the measured and simulated profiles to link the BOP with the diffuse reflectance profiles. The prediction performance of the inverse simulation-based metamodel was comparable with the one based on the measurements, obtaining R2 values above 0.96. Proper correction for these sources of mismatches between measurements and simulations thus allows to build a simulation-based metamodel with a wide range of desired optical properties that is applicable to different measurement configurations. Early puffing is one of the dominant postharvest disorders in easy-peeling citrus cultivars such as Satsuma mandarin. To obtain more insight in this disorder, the structural properties (tissue thickness, surface/volume, percent volume and surface/tissue volume) of the flavedo and albedo tissue layer in the peel were calculated from X-ray CT scans of both healthy and puffy mandarin. The thickness of both flavedo and albedo tissue layers was found to be lower in the puffy mandarin. No differences were observed in the size of the oil glands in the flavedo, but the pore size in the albedo of puffy mandarin was found to be larger at a later harvest time. Next, the BOP of the individual tissue layers in the 530 nm - 970 nm wavelength range were quantified from the DIS and unscattered transmittance measurements, while those for the intact fruit were estimated from laser scatter images with the measurement-based metamodel. The reduced scattering coefficient (μs') values for the intact fruit were found to be lower in the puffy mandarins compared to the corresponding healthy fruit. The sizes of both the oil glands in the flavedo and the air pores in the albedo were negatively correlated with the μs' values of the respective tissue layers. The μs' values of the intact fruit were found to be dominated by the scattering properties of the flavedo layer and positively correlated with the size of the oil glands. However, as no differences in the size of the oil glands were observed between the healthy and puffy mandarins, the observed differences in the μs' values of the intact fruit could be attributed to changes in the scattering properties of the albedo and were negatively correlated to the size/percentage of the air pores. This provides a solid basis for the non-destructive optical detection of puffing disorder at an early stage. In general, the measurements on the BOP of citrus tissues provided an insight into the optical differences among tissues and the explorations of the relationship between the BOP and quality properties. The study on the influences of harvest time, fruit size, cultivar and puffing disorder on the BOP investigated the possible sources of variations in practice, which better promoted the development of the quality detection. By simulating the light propagation in citrus fruit, the obtained signal is well linked with the properties of the different tissue layers, which laid the foundations for designing and optimizing quality detection sensors for citrus fruit. The measurement-based and simulation-based metamodels of deriving BOP from spatially resolved diffuse reflectance profiles offered a non-destructive method for the BOP detection, moreover, the probe into the mismatch between these two metamodels further facilitated their applications on measurements.

Publication year:2021

Accessibility:Open