1 Introduction

Visible-near infrared (Vis-NIR) spectroscopy has been widely applied for nondestructive assessment of the physical and chemical quality parameters of fruits by many research groups as well as companies. In conventional Vis-NIR spectral analysis, the spectra are always described by the absorbance of light based on Beer-Lambert Law (or Beer’s Law). The concentration of the compounds is proportional to the absorbance (Chen and Wang, 2001). Quality assessments are typically executed based on large datasets using chemometrical methods. In this way, scattering is weakened or neglected, as well as are the complex light propagations inside the fruits. However, as a kind of turbid biological material, the constituents and structures of fruit tissue are very complex. Information about the interaction between light and fruit tissue is essential in the evaluation of fruit qualities, since optical signals are significantly affected by constituents and the physical structure of tissue. In light penetration in multiple scattering media, such as fruit, both scattering and absorption contribute to the distance-dependent attenuation. For a quantitative understanding of light interaction with fruit tissue, it is critical to obtain two fundamental optical properties characterized by the absorption coefficient (μa) and the scattering coefficient (μs) (Tuchin, 2007). The measured absorption and scattering coefficients can be used not only for characterizing the absorption and scattering features, respectively, of fruit and vegetable tissues, but also for such investigations and applications as simulating light propagation inside tissues, assessing qualities of fruits and vegetables, and providing guides for their process optimization. Most studies measured the reduced scattering coefficient (μs′) instead of μs. The relationship between μs′ and μs is μs′=(1−g)μs, where g is the anisotropy factor.

In biomedical engineering research, different methods have been used to measure absorption and scattering properties of different types of biological tissue (Cheong et al., 1990; Tuchin, 2007). More information about the tissue is obtained because absorption and scattering properties are distinguished. All these methods can be classified into direct and indirect methods according to the theory and technique based on the study of Kim and Wilson (2011). Because of the critical experimental conditions that direct methods need, most methods reported are indirect methods (Tuchin, 2007). Inverse adding-doubling (IAD) is one of the indirect methods based on adding-doubling theory. It usually uses an integrating sphere-based system to measure the transmittance and reflectance of the tissue. Then, the optical properties of the measured tissue can be calculated in the inverse form of adding-doubling (Pickering et al., 1993; Prahl et al., 1993; Prahl, 2011). Compared with other commonly used methods such as the time-resolved method, the spatially-resolved method, and the frequency-domain method, the integrating sphere method has the characteristics of low cost, being easy in execution, but slow speed. It is suitable for initial research on tissue optical properties. It is also a favorable method for comparison purposes to validate the measurement results of other methods because of the relatively accurate results that can be gained by this method.

Research on light transportation in agro-food products can be dated back to the report of Birth (1978), who studied the relationship between the light scattering properties and internal structure of raw plant and pork tissue with a high moisture content. They pointed out that the physical structure of agro-products determined its scattering and the composition determined its absorption. Little progress was made until the 21st century; an important reason for this may be that Vis-NIR spectroscopy attracted the attention of researchers because of its fast speed and simple sample preparation advantages. In the past decade, investigations for measuring tissue absorption and scattering coefficients of agro-products became popular for the previously mentioned reasons. Much attention was paid to the optical properties of apples perhaps because of the universality of this crop (Cubeddu et al., 2001a; 2001b; Zerbini et al., 2002; Qin and Lu, 2008; 2009; Vanoli et al., 2009; 2014; Lu et al., 2010; Rizzolo et al., 2010). Other fruits (peach (Cen et al., 2012), pear (Zerbini et al., 2002), plum (Qin and Lu, 2008), etc.) and vegetables (tomato (Cubeddu et al., 2001a), cucumber (Qin and Lu, 2008), onion (Wang and Li, 2013), etc.) were also studied but not as frequently. In the studies by two groups, the method of time-resolved spectroscopy (TRS) was mostly used by the research group at CRA-IAA (Consiglio per la Ricerca e la Sperimentazione in Agricoltura, Unità di ricerca per i processi dell’industria agroalimentare) in Italy and the hyperspectral imaging-based spatially resolved (HISR) method was mostly used by the research group in ARS/USDA (Agricultural Research Service, United States Department of Agriculture) at East Lansing, Michigan, USA. Although rare, research about absorption and scattering properties of kiwifruit was carried out by the two groups (Zerbini et al., 2002; Qin and Lu, 2008). There were some reports on optical property measurements of fruits and vegetables based on single integrating sphere: Saeys et al. (2008) used single integrating sphere measurements combined with IAD to estimate the optical properties of apple skin and flesh tissue in the 350–2200 nm range for three cultivars. Wang and Li (2012; 2013) measured the optical properties of skin and fleshy tissues of onion at 632.8 nm to better understand their light propagation. The results indicated that onion dry skins have much higher absorption and reduced scattering coefficients compared with onion fleshy tissues. They also found that, for both skin and flesh, the μa varied between cultivars, while the differences of the μs′ between cultivars were less profound.

A preliminary work was presented at 2014 American Society of Agricultural and Biological Engineers (ASABE) Annual International Meeting (Fang and Fu, 2014). After that, the measuring system was improved from manual detection to automatic detection. The current research was aimed to measure the optical properties of kiwifruit tissues using the improved automatic single integrating sphere system. Specific objectives were to: (1) establish an integrating sphere system for measuring the optical properties of biological tissue; (2) estimate the absorption coefficient μa and the reduced scattering coefficient μs′ of the kiwifruit tissue using the established system and the IAD method; (3) compare the optical properties of kiwifruit tissues from three different parts (the flesh part, the seed part, and the seed-base part).

2 Materials and methods

2.1 Integrating sphere system

A schematic of the developed single integrating sphere system is shown in Fig. 1. The system was built on an optical table (Beijing Optical Century Instrument Co., Ltd., Beijing, China), including a light source, a neutral density (ND) filter, an integrating sphere with a diameter of 152 mm (model 4P-GPS-060-SF, Labsphere, North Sutton, NH, USA), a sphere port cover, a light barrier, a sample holder, an optical fiber (model QP1000-2-VIS-NIR, Ocean Optics, Dunedin, FL, USA), a spectrometer (model QE65pro, Ocean Optics, Dunedin, FL, USA), three motorized translation stages (PX245M-01A, VEXTA; TU4036-220L, IKO; Japan), a controller (STC89C516), and a computer. The integrating sphere was coated by Spectraflect® material, which had a reflectivity of 98% at 632.8 nm. Two 25.4-mm-diameter ports located oppositely on the equator of the sphere were used as entrance and exit. Optical fiber was used to connect the port plug on the north pole of the sphere and the spectrometer. The light source used in this study was a He-Ne laser with an output of 632.8 nm and a beam diameter of 1 mm. The laser source was mounted on a kinematic laser holder, which could provide two axes of precision angular adjustments of the laser beam. The ND filter was used to attenuate the incident light intensity. The ND filter in the filter holder can be replaced by a series of ND filters with different optical densities (ODs) but the same size, so that the light intensity can be adjusted into an appropriate range when the light source is replaced. The OD was 1.0 in this study. The sphere port cover and the light barrier were used to cover the exit and entrance ports for different measurement purposes. The sample holder was originally designed to hold the glass cuvette or sheet sample. It has four bolts through two sides to fix the sandwiched tissue sample to make the tissue a slab. The light barrier and the sample holder were installed on a two-dimensional adjustable stage, which was combined with two motorized translation stages. The port cover was installed on another motorized translation stage. The three translation stages were controlled by the controller and the computer. The whole system was enclosed in a dark chamber consisting of optical rails and black hardboards.

Fig. 1
figure 1

Schematic of the single integrating sphere system

Full size image

2.2 Reflectance and transmittance measurement

To obtain the absorption and reduced scattering coefficients based on the integrating sphere system, total transmittance (Tt), total reflectance (Rt), and the scattering anisotropy coefficient g must be known. Measurements of the total reflectance and transmittance were referenced from the setup recommended by Saeys et al. (2008). Before measuring the transmittance and reflectance signals, the reference and dark signals were measured. The dark background (Idark) was measured by covering all ports of the integrating sphere and turning off the laser. The reference signal (Iref) was measured by opening the entrance port and covering the exit port with the port cover of the sphere and turning on the laser. To measure the Tt, the sample was automatically moved to the entrance port of the sphere, while the exit port was blocked with the port cover. In this case, all light transmitted through the sample (It) was collected by the sphere and detected by the spectrometer through the port plug. To measure the Rt, the sample was automatically moved to the exit port of the sphere, so that all light reflected by the sample (Ir) was collected in the sphere and directed toward the spectrometer. The Tt and Rt were calculated using the detected signals of Idark, Iref, It, and Ir.

The IAD program provided by Prahl (2011) was used to calculate the optical properties for each kiwifruit tissue sample. Using the IAD program, μa and μs′ can be calculated using the measured Tt and Rt signals, and g. The g value can be estimated when the collimated transmittance (Tc) was detected following the method recommended by Prahl (2011). In this study, Tc detected by the established system was too small, which influences the calculated g values significantly. Some sensitivity analysis has shown that μa and μs′ estimations are not very sensitive to the value of g (Tuchin, 2007). The values of g reported were 0.55–0.75 for apple flesh (Saeys et al., 2008) and 0.427–0.688 for onion flesh (Wang and Li, 2013). Our previous research also indicated that the optical properties of kiwifruit tissue were not sensitive to the value of g from 0.4 to 0.9. Therefore the g value was set to a constant of 0.7 here.

2.3 Measurement of kiwifruit tissue

“Zespri Green” kiwifruits were purchased from a local market. They were stored in a cold room at (4±1) °C before being tested. Thirty disease-free kiwifruits were manually selected. The samples were taken from the cold room 12 h prior to tests and stored at room conditions (about 15 °C, about 70% relative humidity). The kiwifruits were sliced by a slicer. When slicing, the blade was parallel to the neck-root direction, while the long axis of the kiwifruits was perpendicular to the blade. The thickness of the piece was set to 4 mm. For each kiwifruit, the first and last pieces were discarded. Then, three consecutive pieces from one end to the middle were selected as a set of tissues (including the flesh part, the seed part, and the seed-base part), as shown in Fig. 2. In total, 90 pieces of kiwifruit tissue samples were measured. The thickness of each sample tissue was measured using an electronic micrometer. On each sample, three replicates were made on the central area and the average value was used as the thickness. The refractive indices (nD) of the tissues were measured by a refractometer (model DR-A1 plus, ATAGO, Japan).

Fig. 2
figure 2

Kiwifruit tissue samples of different parts

Full size image

When measuring the spectral signals, the sample was sandwiched by two pieces of quartz glass (ϕ50 mm× 1 mm). Quartz glass was used because of its good homogeneity and high transmittance at the visible range. The nD of the quartz glass is 1.53. The Tt and Rt were measured at five positions of each sample in the same horizontal plane. Every two contiguous positions of each sample have the same distance of 2 mm.

2.4 System validation

To quantify the estimation errors of the absorption and reduced scattering coefficients, the integrating sphere system was validated using the phantoms made from Intralipid 20% (Sigma Aldrich, USA) and a Biomimic solid phantom. First, we measured the μa and μs′ of the Biomimic phantom for continuous 5 d. The marked μa and μs′ of the Biomimic phantom at 632.8 nm are 0.00387 mm−1 and 0.922 mm−1, respectively. Then, the μa and μs′ of Intralipid 20% solution with a different concentration were measured. The validation method was similar to that described by Troy and Thennadil (2001) and Saeys et al. (2008). According to the reports, the μa values of kiwifruit were 0.055–0.283 mm−1. The μs′ values of kiwifruit were 0.15–0.76 mm−1. Pure absorption solutions were first made by mixing small amounts of India ink with distilled water. Intralipid 20% was then pipetted into the pure absorption solutions to make emulsions diluted to the final concentrations of Intralipid 20% at 1%, 2%, 4% (v/v) and of India ink at 0.005%, 0.010%, 0.015%, 0.020% (v/v).

3 Results and discussion

3.1 Validation results of the integrating sphere system

Fig. 3 shows the μa and μs′ values of the Biomimic solid phantom measured at three different time points (9:00 a.m., 2:00 p.m., 9:00 p.m.) over the continuous 5 d. Fig. 3a shows that the relative difference of the absorption coefficient is 45%, 44%, and 45%, respectively. Fig. 3b shows that the relative difference of the reduced scattering coefficient is 3.8%, 4.2%, and 4.8%, respectively, all less than 5.0%. There were no significant changes of the two parameters in 5 d detected by the system, which confirmed its stability. Fig. 4 shows the μa and μs′ values of Intralipid 20% solution with different concentrations and optical paths. The result shows that the relative differences of the absorption and reduced scattering coefficients of Intralipid 20% solution with different concentrations are both less than 10%.

Fig. 3
figure 3

μ a (a) and μ s ′ (b) values of the Biomimic solid phantom measured at three different time points (9:00 a.m., 2:00 p.m., 9:00 p.m.) over continuous 5 d

Full size image
Fig. 4
figure 4

μ a and μ s ′ values of Intralipid 20% solution with different concentrations (1%, 2%, 4%) and optical paths (3, 5, 7 mm)

Full size image

Thus the above results indicate that the system is of good stability. The relative differences of the absorption and reduced scattering coefficients are acceptable, being in the range of 0.02–0.30 mm−1 and 0.02–0.80 mm−1, respectively, as well as the optical path in the range of 3–7 mm. The relative difference of the absorption coefficient is large when measuring the optical properties of the Biomimic phantom. The reason may be that the Biomimic phantom is a kind of high albedo medium. The absorption coefficient of the Biomimic phantom is significantly lower than the scattering coefficient of the Biomimic phantom, 0.00387 mm−1 and 0.922 mm−1, respectively. It leads to large errors measuring the absorption coefficient of a high albedo medium using an integrating sphere-based system. Overall, the accuracy of the system for estimating the μs′ could meet the detecting requirement.

3.2 Reflectance and transmittance of kiwifruit tissue

Fig. 5a shows the total reflectance (Rt) values of the kiwifruit samples. The Rt values of the seed-base part were the highest among the three parts. In contrast, the Rt values of the seed part were the smallest. The deviations of the Rt values among different positions on one piece and among different pieces of three parts were small. Fig. 5b shows the total transmittance (Tt) values of the samples. The trend shown in Fig. 5b was different from that in Fig. 5a. For the Tt, the deviations among different positions of the seed part were the largest. The deviations among different pieces of the three parts were similar.

Fig. 5
figure 5

Total reflectance R t (a) and total transmittance T t (b) of each piece of kiwifruit tissues (flesh part, seed part, and seed-base part) at 632.8 nm

Data are expressed as mean±standard deviation (SD) at five positions of each sample (total 30 samples)

Full size image

The main reason for this situation was that the kiwifruit tissue was a heterogeneous medium, especially the seed part. As Fig. 2 shows, seeds were randomly distributed at the center area of the seed part. When the translation stages moved for the total transmittance and the total reflectance measurements at different positions of each sample, whether the light beam irradiated the seeds or not made a significant difference to the results. The IAD method can only be accurate on a spatially homogenous medium since it assumes that layers to be doubled have identical optical properties. The seed part was so heterogeneous that the optical properties in the light path were significantly different. Thus, the measurement error of IAD can be quite significant. The results on seed part may therefore be unreliable. Fig. 2 also shows that the center area (the white area) of the seed-base part was the most homogeneous among the three parts. That explains why the deviations of both the Rt and Tt values among different positions of the seed-base part were smaller than those of the other two parts. The detected spectral signals concern the optical properties of the tissue in a very small area that the laser beam irradiated. Therefore, it is necessary to acquire spectral signals from different positions, especially for heterogeneous media.

3.3 Absorption and reduced scattering coefficients of kiwifruit tissue

The average thickness and refractive index of samples of the flesh part, seed part, and seed-base part were 4.35, 4.43, 4.24 mm and 1.338, 1.335, and 1.342, respectively. Fig. 6 shows the mean and standard deviation values of μa and μs′ of the tissues of different parts estimated by the IAD program. According to the mean values of the μc, the light absorption of tissue in the seed part was about 10 times higher than that of tissue in the seed-base part. The μa of the flesh part was between those of the other two parts. The μs′ of the seed-base part was about 8 times higher that of the seed part. The μs′ of the flesh part was between those of the other two parts, which was similar to the situation of μa. Table 1 shows results of analysis of variance (ANOVA) for the optical properties of different parts of kiwifruits. The results indicated that the μa and μs′ of the three parts were significantly different, with the P-values being less than 0.01.

Fig. 6
figure 6

μ s ′ (a) and μ a (b) of kiwifruit tissues of the flesh part, seed part, and seed-base part at 632.8 nm

Data are expressed as mean±SD (total 30 samples)

Full size image
Table 1 Results of ANOVA for the optical properties of kiwifruit tissues at 632.8 nm
Full size table

The absorption differences of the three parts were due to their pigments. The dark color of the seeds in the seed part and the white color of the tissue in the seed-base part were the main reason for the highest and lowest μa values, respectively. The significant differences of the μs′ values of the three parts might be caused by the different tissue microstructures of the three parts. For tissues of the flesh part and the seed-based part, the μs′ values were much larger than the μa values, which was consistent with the statement that fruits are high scattering biological media. The exception of tissue of the seed part may due to the randomly distributed seeds in this part.

Generally, the μa values of kiwifruit tissue detected by the established integrating sphere system and the IAD method were comparable to those reported in other research (Zerbini et al., 2002; Qin and Lu, 2008). The μa values of kiwifruit reported by Zerbini et al. (2002) were 0.005–0.450 mm−1 in the range of 650–1000 nm based on the TRS technique. In the study of Qin and Lu (2008), the μa values of kiwifruit were 0.02–0.10 mm−1 in the range of 500–1000 nm based on the HISR method. However, the μs′ values of kiwifruit tissues detected in this study were much smaller than those detected by Zerbini et al. (2002) (1.0–1.6 mm−1 in the range of 650–1000 nm) and Qin and Lu (2008) (0.7–0.8 mm−1 in the range of 500–1000 nm). The differences of the μa and μs′ values detected in these studies may partly due to the different varieties of the kiwifruit samples, the different methods used as well as the different wavelength ranges.

4 Conclusions

The optical properties of kiwifruits at the wavelength of 632.8 nm were investigated. The mean values of μa and μs′ of tissues in different parts of kiwifruit were 0.031–0.308 mm−1 and 0.120–0.946 mm−1 at 632.8 nm, respectively. At this wavelength, the μs′ values were much larger than the μa values of tissue in the flesh part and the seed-based part. In contrast, μa was larger than μs′ for tissue in the seed part. The results revealed the significant differences of optical properties of kiwifruit tissue in different parts. It can be confirmed that studying the optical properties of fruit tissue is very important for understanding light interaction with different kinds of tissue. The measured absorption and scattering characteristics of fruit tissue may help to better design optical systems for quality evaluation. Based on this study, further investigation will be extended to a broader spectral region and different kinds of fruits.

Compliance with ethics guidelines

Zhen-huan FANG, Xia-ping FU, and Xue-ming HE declare that they have no conflict of interest.

This article does not contain any studies with human or animal subjects performed by any of the authors.