Abstract
The development of new high-quality apple (Malus ×domestica) cultivars that are resistant to flesh browning is needed to expand the use of apples in the food service and catering industry. However, conventional methods for evaluating apple flesh browning can be both time-consuming and costly, thereby rendering such methods unsuitable for breeding programs that must characterize a large number of product samples. Therefore, it is necessary to develop new, simple, and inexpensive methods. The aim was to develop a method for simultaneously measuring the color values of 42 apple samples using a digital camera. The processing time per sample was reduced to less than one-tenth of that of the conventional method. The measurement dispersion [sd of the color difference between two colors
The processing of fresh-cut fruits has increased their consumption because of the ease and variety of their use in the food service industry (Lu et al., 2007; Oliveira et al., 2015); however, the processing of several kinds of fruits and vegetables, including apple (Malus ×domestica), causes their surfaces to turn a brownish color (Pristijono et al., 2006). These color changes (hereinafter referred to as “browning”) are caused by phenolic oxidation reactions catalyzed by polyphenol oxidases, such as tyrosinases and laccases (Veltman et al., 1999). The quality of the taste and aroma of apples deteriorates with browning (Coseteng and Lee, 1987; Guardo et al., 2013; Murata et al., 1995); therefore, an extra step during processing, namely, coating with an antioxidant, is required to control the damage. For these reasons, apples have limited use in the food service and catering industries, compared with other fruits, such as citrus (Citrus sp.), grape (Vitis sp.), and pineapple (Ananas comosus). The use of nonbrowning apple cultivars may increase the demand and consumption of fresh-cut apples, but only a few such cultivars exist. ‘Aori27’ (Igarashi et al., 2016; Tazawa et al., 2019) and Arctic™ apples (Carter, 2012) are genetically modified cultivars that show natural browning resistance. Increasing the number of nonbrowning cultivars would increase fresh-cut apple availability. To develop them, it would be useful to obtain information on the degree of browning among cultivars, but apple flesh coloration is not consistent across cultivars.
Some cultivars have reddish flesh that becomes darker when exposed to air. In addition, watercore, in which part of the flesh appears translucent and water-soaked, may develop in late-harvested apples. Watercore in fruit such as fully ripe ‘Fuji’ apples is popular in Japan (Arakawa and Komori, 2006), although it is generally treated as an internal physiological disorder because it shortens shelf life (Bennedsen and Peterson, 2005). Studies examining reddish and watercored flesh are limited because these specialized tissues are usually patchy and it is therefore difficult to evaluating browning using conventional methods, such as colorimetry or measuring polyphenol oxidase activity and polyphenol composition.
Moreover, general methods for measuring food browning are not suitable for measuring a large number of samples because these methods only measure one sample at a time, making them both time-consuming and costly. In addition, many other farm products, such as european pear (Pyrus communis), peach (Prunus persica), eggplant (Solanum melongena), globe artichoke (Cynara scolymus), and mushroom (Agaricus bisporus), undergo color changes (Amiot et al., 1995; Cabezas-Serrano et al., 2009; Lee et al., 1990; Prohens et al., 2007; Quevedo et al., 2016). An efficient and easy-to-use color evaluation method may be applicable to other crops. A more effective method for evaluating browning is needed for selecting nonbrowning apple cultivars.
Color gamut measurement is a technique that uses computer imaging systems and color images from digital cameras or scanners in addition to colorimetry (León et al., 2006; Yam and Papadakis, 2004). A color gamut is measured in a given region of the sample image [i.e., the region of interest (ROI)] to analyze the color of multiple samples at one time. Previous reports have proposed similar colorimetric systems, and their advantages have been discussed (León et al., 2006; Quevedo et al., 2014). For example, these systems can observe color changes in a time-dependent, nondestructive, and noninvasive manner that can detect even subtle changes in fruit coloration. However, these systems either do not consider the measurement error that arises from differences in brightness owing to sample position, or they correct for this error using a colorimeter. Correcting for these errors without using a colorimeter may simplify the system and facilitate its broader use. In addition, these systems imitate the measuring method of a colorimeter, and one of the features is that they make an adequate substitute for an expensive and high-performance device. However, recently, colorimeters such as the Nix Pro 2 Color Sensor (Nix Sensor, Hamilton, ON, Canada), which are very inexpensive ($349) and have sufficient accuracy for scientific measurement (Hodgen, 2016; Holman et al., 2018; Stiglitz et al., 2016), have become available. It is necessary to carefully consider the advantages and disadvantages of each method after comparing it with such products.
These values have been used to evaluate fresh-cut apple browning; however, Lunadei et al. (2011) and Quevedo et al. (2009) suggest that there are other indices that better explain the degree of fruit browning. Potential useful indicators include the browning index [BI (Buera et al., 1985; Lunadei et al., 2010)], which represents the degree of browning in a variety of foods, and the CIE DE2000 color difference [ΔE00 (Luo et al., 2001)], which is an index adjusted for
Here, we aimed to select an index that enables accurate browning evaluation for any type of apple flesh to allow the accurate evaluation of hybrid progenies with a variety of flesh colors. The objective of this study was to develop a high-throughput, cost-effective analytical system to measure apple browning with a small measurement error. Thus, we attempted 1) to develop a method for the simultaneous measurement of many samples without a colorimeter or other expensive devices; 2) to establish an evaluation index to identify the ideal degree of apple flesh browning; 3) to establish a high-throughput system to determine the degree of apple flesh browning using this method; and 4) to investigate the parameters required to quantify apple browning in cultivars with uncommon flesh colors.
Materials and Methods
Plant material.
Ten apple cultivars and one hybrid progeny were used in the experiments: two severely browning cultivars—Fuji (FJ) and Cripps Pink (CP); three mildly browning cultivars—Shinanogold (SG), Kinshu (KS), and Tsugaru (TG); one browning-resistant cultivar—Aori27 (AO); four reddish-flesh cultivars—Geneva (GN), Pink Pearl (PP), Rose Pearl (RP), and Ruby Sweet (RS); and one watercored hybrid progeny—7-10011 [Haruka × Aika-no-kaori (PG)]. All materials were planted in the experimental orchard at the Apple Research Station, National Agriculture and Food Research Organization (Morioka, Japan). The experiments were conducted in 2017 and 2018.
Apple flesh samples.
Six (for measurement conditions) or three (other experiments) ripe fruit were harvested from the same tree for each cultivar and the hybrid progeny, and stored at 1 °C for at least 16 h and up to 4 weeks before use. Only one sample was prepared from each fruit. The sunlit side of each fruit was vertically sliced into 3-mm-thick samples using a stainless-steel vegetable slicer. These slices were further processed into 40-mm-diameter samples cut from the slice interior using a circular cutting die. The flesh disks obtained were placed at 42 measuring points on an acrylic board (Fig. 1A). The first measurement was taken within about 5 min of the start of sample preparation. If the color of the sample changed significantly (ΔE00 at 5 min after the start of preparation was more than 1.0), the first measurement was taken within 1.5 min after the start of preparation. We also used grated apple flesh samples for measurement to compare the coloration trends to the sliced samples. The grated samples were prepared after removing the core and pericarp from 100 g of fruit. The samples were processed using a stainless-steel grater and placed on glass laboratory dishes. The first measurement was taken within 1.5 min of sample preparation. The grated samples were measured one at a time.
The sample arrangements for calibration and color measurement. (A) Apple flesh samples and (B) color standards (Formula Guide Solid Coated & Uncoated; Pantone, Carlstadt, NJ) were arranged and photographed as the same 42 points installed on a gray acrylic board. The numbers on the acrylic board were used as a guide for the sample position and are not related to the measurement.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Color standards.
One hundred and forty color standards (Supplemental Table 1) were selected from a commercial formula guide kit (Pantone Formula Guide Solid Coated & Uncoated; Pantone, Carlstadt, NJ) and used for calibration (Fig. 1B).
Lighting system.
A camera and four D65 fluorescent lamps (FL20S·D-EDL-D65; Toshiba, Tokyo, Japan) with a color temperature of 6504 K and a color index (Ra) close to 98, were placed in a light booth. A diffusion light reflector made of foam-core board covered with neutral drawing paper (Luminescence neutral white; Tokushu Tokai Paper Co., Shimada, Japan) was used to light the sample area uniformly (Fig. 2). A digital camera (E-510; Olympus, Tokyo, Japan) and interchangeable lens (EZ-1442, Olympus) were fixed vertically over a gray background at 0.45 m. The zoom lens was fixed at a 75° view angle, and the camera optimum exposure and white balance settings were determined with a silk gray card (version 2; Ginichi, Tokyo, Japan). All settings were set manually (Table 1). The camera used a self-timer function. Forty-two samples were placed in the image-capture area of the camera. The light booth was located in a room without any external light; the room temperature (20 °C) and relative humidity (80%) were fixed. The apple disk samples were photographed seven (reddish and watercored flesh) or 20 (normal flesh) times over 24 h. The grated apple samples were photographed 25 times over 60 min.
Lighting system for photographing samples under stable light conditions. (A) Fluorescent lights were installed at positions where apple flesh and color standard samples (Formula Guide Solid Coated & Uncoated; Pantone, Carlstadt, NJ) were not directly illuminated. (B) The door of the lighting booth was closed at the time of sample imaging.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Digital camera settings used to evaluate apple browning.
Calibration and color measurement.
The digitized images were transferred to a personal computer using a memory card (Compact Flash; Western Digital Corp., San Jose, CA) and transformed to an 8-bit TIFF format in the standard color space [sRGB IEC61966-2.1 (International Electrotechnical Commission, 1999)] using the image organizer software (Photoshop Lightroom CC; Adobe, San Jose, CA). The “Profile Correction” and “Remove Chromatic Aberration” functions were selected after the photographs were converted. The following methods describe the background correction and RGB calibration that were applied to each of the 42 measurement points.
Sample color extraction.
Image analysis was performed using an image processing program (ImageJ, version 1.52; National Institutes of Health, Bethesda, MD) and operated with a macro file that performs a loop handling of the “RGB Measure” plugin. Forty-two circles with diameters of 215 pixels were created over the sample ROIs using the “ROI Manager” tools, with one additional ROI for background correction. Then, the mean intensity values for red, green, and blue (R, G, B) in each ROI were extracted with the “RGB Measure” plugin.
Background correction.
RGB calibration.
The SSANOVA process was implemented using the “ssanova” function in the R package “gss”. The smoothing parameters of these models were selected to minimize the generalized cross-validation score. Models that estimated SSANOVA were used for the calibration of the R, G, B values. If the calibrated values were beyond the range of 0 to 255, they were handled as 0 or 255.
Color space conversion.
To transform the R, G, B values to L, a*, b*, the methods described in Hunt and Pointer (2011) and León et al. (2006) were followed, using a spreadsheet (Excel 2019; Microsoft, Redmond, WA) and the event-driven programming language (Visual Basic for Applications 7.1, Microsoft). The transformation method was applied to all the measured and calibrated sample data.
Browning evaluation.
Measurement accuracy and reproducibility.
To assess the accuracy of the method, the L*, a*, b* values of 51 selected color standards [different from those used for calibration (Supplemental Table 1)] were measured using a colorimeter (NF333; Nippon Denshoku Industries Co., Tokyo, Japan) and estimated using the method described previously. The accuracy was evaluated using the root mean square error (RMSE) scores of the colorimeter values, which indicated the magnitude of the difference between the values extracted from the digital images of each color sample, or the values following calibration, and the values measured by the colorimeter (Supplemental Table 2). In other words, the smaller the RMSE value, the more accurate the measurements.
To verify the effectiveness of the calibration method, the calibrated values were compared with the raw measured values and the colorimeter values. The color standard [425U, Pantone (R, G, B = 124, 126, 127)] was measured 30 times at all 42 points every 30 s. Then, the L*, a*, b* values of the 42 measurement points in the 30 replicates were estimated. Finally, the
Observation of browning in reddish flesh and watercore regions of apples from a variety of genotypes.
After obtaining digital images of slice samples from reddish-flesh cultivars, the strongly reddish and pale regions of each sample were selected by appearance and set as ROIs, and named deep red (Dr) or pale red (Pr), respectively. Then, the L*, a*, b* values of the Dr and Pr regions were measured and their browning values calculated. The corresponding values of watercored regions (Wc) were also measured using this method. These scores were compared with those of normal-flesh samples of FJ. The appropriateness of these browning indicators was verified with frequently used browning indices (i.e., L*, a*, b*, ΔE00, BI, ΔBI).
Results
Method for the evaluation of the degree of apple browning.
A photography booth using indirect illumination, with reference to Yamamoto et al. (2011) (Fig. 2A), reduced the brightness difference in the images to one-fifth or less of the methods described by León et al. (2006) and Lv et al. (2009) (Supplemental Fig. 1). The use of a digital camera made it possible to measure multiple samples simultaneously using a simple operation and to observe time-course color changes. Moreover, by using a replaceable acrylic plate as a sample holder, more than 42 samples, which was the maximum number of simultaneous measurements, could be easily handled. The cost of making this measurement system, including the digital camera, was ≈$1900. This is more than five times the cost of an inexpensive colorimeter [i.e., Nix Pro 2 ($349)], but it is less than one-third the cost of a sophisticated colorimeter [i.e., CR-410; Konica Minolta, Tokyo, Japan ($7000)]. A comparison between our method and a commercial colorimeter is shown in Table 2.
Comparison of the features of the developed method of evaluating apple browning with those of an inexpensive colorimeter [Nix Pro 2 (Nix Sensor, Hamilton, ON, Canada)], and two sophisticated colorimeters [CR-410 (Konica Minolta, Tokyo, Japan) and NF 333 (Nippon Denshoku Industries Co., Tokyo, Japan)]. The measured repeatability and interinstrument agreement were measured using a color standard (425U; Pantone, Carlstadt, NJ). Both
As apple samples, we used apple flesh disks processed using a vegetable slicer and a circular cutting die (Fig. 1A). Sample preparation using this process took an average of 8.5 s per sample. This was a 10-fold increase in efficiency compared with the preparation of 100 g of grated flesh, which took an average of 90 s. The flesh disks of six cultivars were measured 20 times between sample preparation and 24 h after processing. ΔE00 and ΔBI changed in a similar manner in all six cultivars (Fig. 3D–E); furthermore, these changes were in agreement with the degree of color change that was visually recognized (Fig. 3G). Although the maximum value of the indices varied with cultivar, the variation in the indices after the maximum values were obtained was slight. Based on these results, we measured the browning of apple flesh slices eight times within 24 h (at 0, 5, and 15 min, and at 1.5, 4, 7, 12, and 24 h); the maximum value of the observed index was treated as the representative value of browning degree.
Time-dependent color change in apple flesh slice samples. The changes in (A) L*, (B) a*, (C) b*, (D) ΔE00, (E) ΔBI, and (F) BI for evaluating the progress of browning in sliced fresh apple samples. L*, a*, b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Laboratory color space. ΔE00 is the color difference expressed by the CIE DE 2000 color difference formula. BI is the browning index described in Buera et al. (1985), and ΔBI represents the difference between the two colors. Each marker is presented as the mean value for each of six cultivars: Fuji (FJ), Cripps Pink (CP), Shinanogold (SG), Kinshu (KS), Tsugaru (TG), and Aori27 (AO) (n = 6). (G) Images of the apple samples before and after browning. Images in the column BImax were photographed when the maximum BI values were observed. Each sample image was selected by the BImax values that were nearest to the values for each marker.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Accuracy and reproducibility of the developed color measurement system.
From the L* (Fig. 4A) and b* (Fig. 4C) values, we found that the RMSE scores of the calibrated data were smaller than those of the raw measured data at any measurement point. Overall, 5-fold or larger differences in the RMSE scores of the L* values before (RMSE scores calculated at each of the 42 measurement points ranged from 10.806–14.297, Supplemental Table 2) and after (1.786–1.975) calibration were reported. However, for the a* values were observed for any of the 42 measurement points compared with the other measurement points (Supplemental Table 2).
Accuracy, reproducibility, and interpoint dispersion of the colorimetric system for evaluating apple browning. The (A) L*, (B) a*, and (C) b* values in the CIE 1976 Laboratory color spaces of 51 selected color standards (different from those used for calibration, Supplemental Table 1) were measured using a colorimeter (NF333; Nippon Denshoku Industries Co., Tokyo, Japan) and estimated using the tested method. Root mean square error (RMSE) was calculated from the values measured by the colorimeter (the y = x lines indicate ideal measurement) and each estimated value. The markers “measured” and “calibrated” indicate the measurement values taken before and after correction using our method. (D) Change in measurement dispersion after calibration. The lateral axis indicates the standard deviation (sd) of
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
The sd of 30 replicates for each of the 42 points is shown in Fig. 4D. The sd of the calibrated data decreased significantly to approximately half of the sd of the raw data, except for one measurement point located on the edge of the top left side of the imaging region. The interpoint dispersion of the sd, which met the instrumental error of colorimeters, showed that the maximum value of sd in 30 measurements was
Changes in the browning indices of reddish-flesh apples.
In the slices of reddish-flesh cultivars, the changes in the L* and b* values over time were similar to those of FJ (Fig. 5A and C). The a* values of GN and PP decreased over time, in contrast to the increased values in the FJ samples. However, in the RP and RS samples, the a* values increased after a transient decrease (Fig. 5B). In RP in particular, the value increased after 1.5 h, compared with that at the start of measurement, and a trend different from any other cultivar was confirmed. The BI values of the reddish-flesh cultivars were higher than those of the other cultivar groups (Fig. 5F), but the ΔBI values before and after browning coincided with the color changes observed in the photographs (Fig. 5E and G). The ΔE00 values showed large changes in the FJ samples compared with ΔBI; however, there was no difference between the two indices in the ranking of the values in the reddish-flesh cultivars (Fig. 5D and E).
The time-dependent color changes in the reddish- [‘Geneva’ (GN), ‘Pink Pearl’ (PP), ‘Rose Pearl’ (RP), ‘Ruby Sweet’ (RS)], watercored- {the hybrid progeny 7-10011 [‘Haruka’ × ‘Aika-no-kaori’, (PG)]}, and nomal- [‘Fuji’(FJ)]; apple flesh disk samples. The changes in (A) L*, (B) a*, (C) b*, (D) ΔE00, (E) ΔBI, and (F) BI values in reddish-flesh and watercored cultivars during the time course of the experiment (n = 6). L*, a*, and b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Laboratory color space. ΔE00 is the color difference expressed by the CIE DE 2000 color difference formula. BI is the browning index described in Buera et al. (1985), and ΔBI represents the difference between the two colors. (G) Browning in reddish-flesh and watercored apple samples.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
We then measured the differences in the time courses of the indices, based on the depth of the flesh colors in GN and PP. In the Dr region of the GN samples, a characteristic large maximum change in the a* values (Δa* = 14.4) and a slight increase in the L* values were observed over time (Fig. 6B and C). In addition, the BI values of Dr in GN increased 15 min after processing and then gradually decreased, although the measurement error was larger in this cultivar than in the others tested. In the Pr region of GN, the ΔE00, ΔBI, and change in b* (Δb*) values were the largest among the four regions measured (Fig. 6D–F). In particular, the Δb* values had a maximum of 16.3, which was more than twice as large as that in the other regions. In PP, the Δb* value of Dr was larger than that of Pr, but not in the GN samples. The changes in the six indices of PP were about consistent between the two regions, and the differences in ΔE00 and ΔBI were smaller than those for the GN regions.
Changes of the colorimetric values of browning in deep red (Dr)- and pale red (Pr)-colored apple flesh. (A) The assigned Dr/Pr regions in cultivars Geneva (GN) and Pink Pearl (PP). The changes in the (B) L*, (C) a*, (D) b*, (E) ΔE00, (F) ΔBI, and (G) BI values in the four portions during the time course of the experiment (n = 3). L*, a*, and b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Laboratory color space. ΔE00 is the color difference expressed by the CIE DE 2000 color difference formula. BI is the browning index described in Buera et al. (1985), and ΔBI represents the difference between the two colors.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Changes in browning indices within watercored fruit portions.
At the start of measurement, the a* and b* values for the watercored PG samples were higher than those for FJ, while the PG L* values were lower than those for FJ. However, there were few differences in the color change between PG and other yellowish-flesh cultivars during the time course of the measurements (Fig. 5A and B).
The differences in the indices between the Wc and normal portions of the PG samples are shown in Fig. 7. The increase in the a* values of the Wc regions during browning was slow and small, compared with those in the normal portions. Additionally, the b* values of Wc increased, while they decreased in the normal portions of the samples. Before and after browning, the ΔE00 values of Wc were about half those of the normal portions, whereas the ΔBI values were more than eight times greater than those of the normal portions.
Changes in the browning colorimetric values in watercored portions of apple flesh. (A) The portions assigned as the watercored (Wc) and normal areas of the hybrid progeny 7-10011 [‘Haruka’ × ‘Aika-no-kaori’, (PG)]. The time-course changes in (B) L*, (C) a*, (D) b*, (E) ΔE00, (F) ΔBI, and (G) BI values in the two portions (n = 3). L*, a*, and b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Laboratory color space. ΔE00 is the color difference expressed by the CIE DE 2000 color difference formula. BI is the browning index described in Buera et al. (1985), and ΔBI represents the difference between the two colors.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Browning of grated apple samples.
The index change of the grated apple samples was very similar to that observed in the sliced samples, but the changes in their browning indices were larger and occurred faster than those in the sliced samples (Figs. 3 and 8; Supplemental Fig. 2). In particular, the greatest changes in the L* and a* values were ≈3- to 4-fold larger than those in the sliced samples for FJ (Figs. 3A and 8A); furthermore, the b* values decreased in grated FJ samples, whereas those of the sliced samples increased. In FJ, TG, and PP, the largest ΔE00 values of the grated samples were larger than those measured in the sliced samples, whereas there were few (nonsignificant) differences between the sliced and grated AO, GN, and PG samples. However, the ΔBI values in the grated GN samples were considerably greater than in the other samples (Figs. 3, 5, and 8). The degree of browning roughly corresponded to that of the ranking of the cultivars based on BI values; however, the color of the TG samples 24 h after the start of measurement (
Time-series measurements of grated apple flesh samples. The time-course changes in (A) L*, (B) a*, (C) b*, (D) ΔE00, (E) ΔBI, and (F) BI values from grated samples of five apple cultivars [Fuji (FJ), Tsugaru (TG), Aori27 (AO), Geneva (GN), and Pink Pearl (PP)] and one hybrid progeny 7-10011 [‘Haruka’ × ‘Aika-no-kaori’, (PG)] (n = 3). L*, a*, and b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Laboratory color space. ΔE00 is the color difference expressed by the CIE DE 2000 color difference formula. BI is the browning index described in Buera et al. (1985), and ΔBI represents the difference between the two colors. (G) Images were taken before and 60 min after sample processing.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Discussion
In this study, we developed a simple and high-throughput system for the quantitative determination of apple flesh browning without using expensive equipment (Figs. 1 and 2). A vegetable slicer and a circular cutting die were used to prepare the flesh disk samples. This eliminated the heterogeneity of shape of samples processed with a knife, and the time required to prepare each sample was reduced to one-tenth of the time required to prepare a grated sample (data not shown). The time elapsed during sample preparation was considered negligible for most apple cultivars, as only one of the eight cultivars showed a visible color change (
In addition, we attempted to eliminate image color-value deviations owing to the influence of optical conditions only using correction based on inexpensive color standards rather than a colorimeter. The indirect illumination, made with reference to Yamamoto et al. (2011), allowed a uniform light environment, such as that achieved by a colorimeter (Supplemental Fig. 1). The calibration by the SSANOVA regression model and the background correction reduced the deviation of the L* and b* values observed with colorimetry measurements. In our method, the sd of
In addition to its high throughput, high accuracy, and low labor costs, our method demonstrates three other advantages of using digital camera images to obtain color values. The first is that browning is recorded as an image; thus, the numerical value can be compared with the degree of browning appearance. Compared with methods using a colorimeter, in which only numerical data are recorded, in our method, it is easy to specify the cause of the difference when an outlier is observed, because the image of the sample corresponding to the measured value is always available. Hence, the risk of wasting all the experimental data at one time is considerably reduced. Second, our method achieves smaller human error, as the sample colors are measured from a remote location; therefore, the risk of contamination of the samples is limited and stable results are reliably obtained by averaging a wide range of measurements of each sample. The third advantage is that the positional relationship between samples is fixed during measurement, which limits the risk of mixing the data between cultivars and fruit. Evaluating apple flesh browning using our system can greatly reduce the failure of an experiment owing to operation error, especially when processing large numbers of experimental samples. Thus, our method provides a more stable measurement platform, compared with conventional colorimetric methods.
As described previously, our method can simultaneously measure a plurality of samples with little human error, although it is inferior to a colorimeter in ease of use and more expensive than an inexpensive device, such as the Nix Pro 2 (Table 2). These features make this method suitable for experiments with large sample sizes and time-course measurements. In addition, this method facilitates site-specific analysis, as shown in Figs. 6 and 7. This is because the measuring position and measuring range can be freely set.
We also analyzed the changes in color values owing to the browning of apple flesh by measuring flesh disks over time, using our newly developed method (Figs. 3 and 5). The size of the BI values used as an indicator of browning and the color tone of each sample were in good agreement, unless the values of the grated GN samples were unstable, similar to the measurements in the Dr region of the flesh disks (Figs. 5F and 8F). This indicates that it is possible to quantitatively evaluate the degree of browning using the BI values of normal (yellow or white) apple flesh colors (Fig. 3). Because the b* value is included in the calculation formula of BI, it is a more appropriate evaluation of apple color tone.
In the samples from the reddish-flesh cultivars, the BI values were generally higher than those of the other cultivars (Figs. 3F and 5F). In addition, the measured BI values of the grated GN samples were unstable, similar to the measurements in the Dr region of the flesh disks (Figs. 5F and 8F). Since their L* and a* values were significantly different from those of the yellowish cultivars, there is a risk in treating the BI values of the reddish-flesh cultivars the same as those of yellowish-flesh cultivars, especially as an absolute browning degree indicator. In addition, the BI values varied with the depth of the red color of the apple sample (Figs. 6G and 8F).
The ΔE00 and ΔBI values before and after browning were calculated as indices that indicate the degree of coloration change of the cut surface of the flesh disk, and the correspondence between these indices and the visually evaluated color differences were confirmed (Fig. 3D, E, and G).
From the observations of the changes in ΔE00 and ΔBI over time, we show that cultivars can be classified into two groups (Fig. 3): cultivars with ΔE00 and ΔBI values that increased between 12 and 24 h after sample processing (FJ, KS, TG, and AO) and cultivars whose ΔE00 and ΔBI values decreased during this time interval (CP and SG). These results suggest that the changes in flesh disk coloration rate over time are caused not only by the chemical equilibrium in a single enzymatic reaction, but by multiple factors. In this experiment, both ΔE00 and ΔBI reflected the degree of coloration well, but the magnitude of the correlation between FJ and CP 24 h after the start of measurement was opposite in ΔE00 ΔBI. The intercultivar differences in ΔBI agreed more with those observed through visual perception, suggesting that ΔE00 should not be used in the measurement of cultivars that show a large degree of browning. ΔE00 indicates an improved index representing color differences that approximate human vision, and the values were adjusted to the visual experience, such that the deeper the color, the lower the color difference perception (Luo et al., 2001). Considering this adjustment, the ΔE00 of a yellowish-flesh sample with higher color saturation could be calculated to be much smaller than that of the sensory difference.
The relationship between ΔE00 and ΔBI in the reddish-flesh cultivars was similar to that in the yellowish-flesh samples; ΔE00 was relatively smaller than ΔBI in the samples with relatively bright colors (Fig. 5D, E, and G). Therefore, although the degree of color change can be quantified using both ΔE00 and ΔBI, as in the yellowish cultivars, it is necessary to examine the index used for evaluating the color tone in reddish-flesh cultivars in future analyses.
The results for the change in the color values of watercored PG samples over time showed that the color change in the PG samples was similar to that of normal yellowish-flesh cultivars, and that it was possible to determine the degree of browning using either index (Fig. 5). However, the L*, a*, and b* values of the Wc portions, and their changes, were different from those in normal-flesh portions (Fig. 7). Because the watercored area was translucent, its color changes depended on the background, and because gray was used as the background color in our method, the watercored samples were artificially made to look as if they were browning. Therefore, the correction effect of ΔE00 had a greater influence on these results than in the normal-flesh samples. The colorimetry of translucent materials is usually performed by measuring spectral transmittance, not by directly reading the tristimulus value, which was simulated in our method. Therefore, the estimation method used must be reevaluated to accurately measure browning from images of watercored apple samples.
In the grated samples, the trends in the ΔE00 and ΔBI values were not much different from those of the sliced samples, and the difference appeared to be only a greater extent of change. However, with their lower L* and b* values and higher a* values, the accuracy of the quantification of the degree of browning using BI may not be as accurate. When a dark reddish-purple color, such as that of GN, was expressed as RGB luminance, the G and B values were very small [0–30 (Fig. 8G, Supplemental Fig. 4)]. These values may have been outside the dynamic range of the image sensor, or the calibration range of the R, G, and B values may have been insufficient for successful sample measurement. These results might be improved by using a camera capable of obtaining more gradation-rich images or by using more standard color samples for calibration, with G and B values that are close to zero.
In the Dr portion of GN, the Wc portion of PG, and the grated samples of FJ, TG, GN, and PP, the value of ΔE00 did not match the magnitude of the coloration observed from the image (Figs. 6A and E, 7A and E, and 8D and G). These flesh types are all characterized by a particularly deep color after coloration. Therefore, it is presumed that even in these cases, the correction included in the ΔE00 calculation formula is too strong.
5The suitability of the indices used in this study for the quantitative evaluation of apple browning is summarized in Table 3. Both ΔE00 and BI corresponded to more diverse samples than L*, a*, b*, and BI was particularly suitable for evaluating normal yellowish flesh. However, ΔE00 may not be consistent with visual perception for deep colors, and BI was limited when measuring dark-reddish-purple flesh and watercored areas.
The ability of each index to evaluate degree of apple flesh browning. “Excellent” is the best indicator. “Good” is a generally good indicator, but depending on the conditions, browning cannot always be expressed well. “Poor” indicates that the evaluation of browning is difficult.
In conclusion, we have developed a system for quantifying apple flesh browning. The system can reduce human measurement errors while achieving high efficiency at the same level of color measurement performance as a sophisticated colorimeter, by using a digital camera and standardized apple flesh disk samples. BI was useful as an index to quantify the degree of browning and it was possible to observe this index changing over time to evaluate the progress of browning. However, there may be different patterns of browning in reddish and watercored flesh, suggesting that different kinds of apple flesh should not be evaluated using the same conditional parameters. This method can be further improved to more accurately measure the color values of dark reddish-purple apple flesh, and it will be necessary to devise ways to expand the dynamic range of the color measurement and develop an index that can accurately express the color change of reddish-flesh apples as browning proceeds.
Literature Cited
Amiot, M.J., Tacchini, M., Aubert, S.Y. & Oleszek, W. 1995 Influence of cultivar, maturity stage, and storage conditions on phenolic composition and enzymic browning of pear fruits J. Agr. Food Chem. 43 1132 1137
Arakawa, O. & Komori, S. 2006 Apple, p. 34–42. In: Jpn. Soc. Hort. Sci. (ed.). Horticulture in Japan. Shoukadoh Publ., Kyoto, Japan
Bennedsen, B.S. & Peterson, D.L. 2005 An optical method for detecting watercore and mealiness in apples Trans. ASAE 48 1819 1826
Buera, M.P., Lozano, R.D. & Petriella, C. 1985 Definition of color in the non-enzymatic browning process Farbe 32/33 316 326
Cabezas-Serrano, A.B., Amodio, M.L., Cornacchia, R., Rinaldi, R. & Colelli, G. 2009 Screening quality and browning susceptibility of five artichoke cultivars for fresh-cut processing J. Sci. Food Agr. 89 2588 2594
Carter, N. 2012 Petition for determination of nonregulated status: Arctic™ apple (Malus × domestica) events GD743 and GS784. U.S. Dept. Agr., Animal Plant Health Inspection Serv., Riverdale, MD
Coseteng, M.Y. & Lee, C.Y. 1987 Changes in apple polyphenoloxidase and polyphenol concentrations in relation to degree of browning J. Food Sci. 52 985 989
Guardo, M.D., Tadiello, A., Farneti, B., Lorenz, G., Masuero, D., Vrhovsek, U., Costa, G., Velasco, R. & Costa, F. 2013 A multidisciplinary approach providing new insight into fruit flesh browning physiology in apple (Malus × domestica Borkh.) PLoS One 8 e78004
Hodgen, J. 2016 Comparison of nix color sensor and Nix Color Sensor Pro to standard meat science research colorimeters Meat Sci. 112 159
Holman, B.W.B., Collins, D., Kilgannon, A.K. & Hopkins, D.L. 2018 The effect of technical replicate (repeats) on Nix Pro Color Sensor™ measurement precision for meat: A case-study on aged beef colour stability Meat Sci. 135 42 45
Hunt, R.W.G. & Pointer, M.R. 2011 Measuring colour. 4th ed. Wiley, Chichester, UK
Igarashi, M., Hatsuyama, Y., Harada, T. & Fukasawa-Akada, T. 2016 Biotechnology and apple breeding in Japan Breed. Sci. 66 18 33
Lee, C.Y., Kagan, V., Jaworski, A.W. & Brown, S.K. 1990 Enzymic browning in relation to phenolic compounds and polyphenoloxidase activity among various peach cultivars J. Agr. Food Chem. 38 99 101
León, K., Mery, D., Pedreschi, F. & León, J. 2006 Color measurement in L*a*b* units from RGB digital images Food Res. Int. 39 1084 1091
Lu, S., Luo, Y., Turner, E. & Feng, H. 2007 Efficacy of sodium chlorite as an inhibitor of enzymatic browning in apple slices Food Chem. 104 824 829
Lunadei, L., Galleguillos, P., Diezma, B., Lleó García, L. & Ruiz-Garcia, L. 2011 A multispectral vision system to evaluate enzymatic browning in fresh-cut apple slices Postharvest Biol. Technol. 60 225 234
Lunadei, L., Galleguillos, P., Diezma Iglesias, B. & Lleó García, L. 2010 Evaluation of enzymatic browning in fresh-cut apple slices applying a multispectral vision system. Proc. Intl. Conf. Agr. Eng., AgEng 2010. Towards environmental technologies. EUIT Agrícola (UPM), Clermont-Ferrand, France
Luo, M.R., Cui, G. & Rigg, B. 2001 The development of the CIE 2000 colour-difference formula: CIEDE2000 Color Res. Appl. 26 340 350
Lv, B., Li, B., Chen, S., Chen, J. & Zhu, B. 2009 Comparison of color techniques to measure the color of parboiled rice J. Cereal Sci. 50 262 265
Murata, M., Tsurutani, M., Tomita, M., Homma, S. & Kaneko, K. 1995 Relationship between apple ripening and browning: Changes in polyphenol content and polyphenol oxidase J. Agr. Food Chem. 43 1115 1121
Oliveira, M., Abadias, M., Usall, J., Torres, R., Teixidó, N. & Viñas, I. 2015 Application of modified atmosphere packaging as a safety approach to fresh-cut fruits and vegetables—A review Trends Food Sci. Technol. 46 13 26
Pristijono, P., Wills, R.B.H. & Golding, J.B. 2006 Inhibition of browning on the surface of apple slices by short term exposure to nitric oxide (NO) gas Postharvest Biol. Technol. 42 256 259
Prohens, J., Rodríguez-Burruezo, A., Raigón, M.D. & Nuez, F. 2007 Total phenolic concentration and browning susceptibility in a collection of different varietal types and hybrids of eggplant: Implications for breeding for higher nutritional quality and reduced browning J. Amer. Soc. Hort. Sci. 132 638 646
Quevedo, R., Jaramillo, M., Díaz, O., Pedreschi, F. & Aguilera, J.M. 2009 Quantification of enzymatic browning in apple slices applying the fractal texture Fourier image J. Food Eng. 95 285 290
Quevedo, R., Valencia, E., López, P., Gunckel, E., Pedreschi, F. & Bastías, J. 2014 Characterizing the variability of enzymatic browning in fresh-cut apple slices Food Bioprocess Technol. 7 1526 1532
Quevedo, R., Pedreschi, F., Bastías, J.M. & Díaz, O. 2016 Correlation of the fractal enzymatic browning rate with the temperature in mushroom, pear, and apple slices Lebensm. Wiss. Technol. 65 406 413
Robertson, A.R. 1977 The CIE 1976 color-difference formulae Color Res. Appl. 2 7 11
Stiglitz, R., Mikhailova, E., Post, C., Schlautman, M. & Sharp, J. 2016 Evaluation of an inexpensive sensor to measure soil color Comput. Electron. Agr. 121 141 148
International Electrotechnical Commission 1999 Multimedia systems and equipment-colour measurement and management-part 2-1: Colour management-default RGB colour space-sRGB. Intl. Electrotechnical Commission 61966-2-1
Tazawa, J., Oshino, H., Kon, T., Kasai, S., Kudo, T., Hatsuyama, Y., Fukasawa-Akada, T., Yamamoto, T. & Kunihisa, M. 2019 Genetic characterization of flesh browning trait in apple using the non-browning cultivar ‘Aori 27.’ Tree Genet. Genomes 15 49
Veltman, R.H., Larrigaudiere, C., Wichers, H.J., van Schaik, A.C.R., van der Plas, L.H.W. & Oosterhaven, J. 1999 PPO Activity and polyphenol content are not limiting factors during brown core development in pears (Pyrus communis L. cv. Conference) J. Plant Physiol. 154 697 702
Yam, K.L. & Papadakis, S.E. 2004 A simple digital imaging method for measuring and analyzing color of food surfaces J. Food Eng. 61 137 142
Yamamoto, K., Kimura, Y., Togami, T., Yoshioka, Y., Hashimoto, A. & Kameoka, T. 2011 A chromatic image analysis system using content-based image retrieval Agr. Inf. Res. (NogyoJyohokenkyu). 20 139 147 (In Japanese), doi: 10.3173/air.20.139
The unevenness in brightness of photographic ranges depending on lighting conditions. (A) When the light source was arranged above the area to photograph, or (B) when a light reflector was used. The numbers yellow frames are red (
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Images of time-course color changes of the apple ‘Fuji’ flesh samples. (A) The grated sample. (B) The sliced sample.
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
The temporal changes in the color difference values (
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Measured red (
Citation: Journal of the American Society for Horticultural Science J. Amer. Soc. Hort. Sci. 146, 4; 10.21273/JASHS05027-20
Color standard samples used to calibrate the intensities of Red (R), Green (G), and Blue (B). The samples presented with a gray background were only used for the accuracy test.
Root mean square error (RMSE) accuracy scores of each set of 42 measurement points. L*, a*, and b* represent the values on each coordinate axis (lightness, greenness to redness, blueness to yellowness) in the CIE 1976 Lab color space. RMSE scores indicates the magnitude of the difference from the values measured by the colorimeter. The colored values are the lower of the two values, before (bottom side) or after calibration (upper side). The designation "(ref. ROI)" indicates the location of the reference region of interest used for background correction.