Read doi:10.1016/j.foodres.2004.05.005 text version

Food Research International 37 (2004) 875­881

Determination of acceptability and shelf life of ready-to-use lettuce by digital image analysis

T. Zhou


, A.D. Harrison a, R. McKellar a, J.C. Young a, J. Odumeru b, P. Piyasena a, X. Lu c, D.G. Mercer a, S. Karr d


Food Research Program, Agriculture and Agri-Food Canada, 93 Stone Road West, Guelph, Ont., Canada N1G 5C9 b Laboratory Services, University of Guelph, Guelph, Ont., Canada N1G 5C9 c Department of Mathematics and Statistics, University of Calgary, Calgary, Alta., Canada T2N 1N4 d Pride Pak Canada Ltd., Mississauga, Ont., Canada Received 12 May 2004; accepted 16 May 2004

Abstract The potential of digital measurement of browning to determine the acceptability and shelf life of ready-to-use (RTU) lettuce was investigated. Shredded lettuce was treated either with or without 48 °C chlorine water (100 ppm) before being washed with 4 °C chlorine water and packaged in plastic bags. The packaged lettuce was then stored at 4 and 10 °C. A human panel visually evaluated the lettuce samples seven times over 18 days using a loss of quality scale from 1 to 5. The same samples were photographed, and the images were analyzed for percent brown area and changes in colour composition with image analysis software (Northern Eclipse). Both the human evaluation and image analyses revealed significant differences among the treatments, with similar trends. Percent brown area as determined by image analysis was a much better indication of lettuce quality than values of colour composition changes. Image analysis of browning corresponded well with storage days, i.e. shelf life. The correlation coefficients between percent brown area and shelf life ranged from 0.9194 to 0.9941, for four different treatments. Also, percent brown area was highly correlated with human visual evaluations. The image analysis of browning is a reliable research tool for objectively and quantitatively determining the quality and shelf life of RTU lettuce, and should also be suitable for use by the food processing industry. Crown Copyright Ó 2004 Published by Elsevier Ltd. All rights reserved.

Keywords: Ready-to-use; Lettuce; Browning; Image analysis

1. Introduction In recent years, ready-to-use (RTU), minimally processed vegetables have gained great acceptance by consumers. RTU vegetables are popular with consumers and the food industry because of the reduction in preparation times, lower costs of shipping and storage, and reduced waste (Odumeru, Boulter, Knight, Lu, & McKellar, 2002). Shredded lettuce is one of the most important RTU products; however, its shelf life is limited due to browning and microbiological deterioration (Bolin, Stafford, King, & Huxsoll, 1977; Priepke, Wei, & Nelson, 1976). Since colour and appearance are imporCorresponding author. Tel.: +1-519-829-2400/780-8036; fax: +1519-829-2600. E-mail address: [email protected] (T. Zhou).


tant quality aspects for shoppers when selecting fresh fruits and vegetables, a colour analysis technique would provide a useful tool for assessing consumer acceptance and predicting shelf life of fresh produce (Jilliffe & Lin, 1997). The shelf life of lettuce can be defined as the length of time which lettuce can maintain an appearance that appeals to the consumer. This appearance consists of a crisp green vegetable with little browning or wetness present. The shelf life of prepackaged lettuce is dependant on several factors. Shredding the lettuce disrupt the protective epidermal layer of the leaf and ruptures the cells, which then decompartmentalizes and releases cell contents leading to biochemical reactions (Delaquis, Stewart, Toivonen, & Moyls, 1999). These reactions include substrate and enzyme interactions that may result in tissue browning. There are several important

0963-9969/$ - see front matter. Crown Copyright Ó 2004 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.foodres.2004.05.005


T. Zhou et al. / Food Research International 37 (2004) 875­881

enzymes involved in enzymatic browning. Oxidative enzymes such as polyphenyl oxidase (PPO), peroxidase, and phenylalanine ammonia lysase (PAL) oxidize soluble phenolic compounds to produce orthoquinones, which polymerize to form insoluble brown pigments (Casta~r, Gil, Ru & Arts, 1999; Delaquis et al., e iz, e 1999). Also, the release of cell contains allow microbes to access easily the nutrients contained inside the cells and provides a positive environment for bacterial reproduction, and hence shortens the shelf life of the lettuce (Delaquis et al., 1999). Many food processing industries have adapted computer technology to assess the quality of products such as fruits, cheese, and shrimp (Coles, Peter, Ammerink, & Wallace, 1993; Kranzier, 1985; Studman, 2001; Tunde & Feldoldi, 2000; Wang & Sun, 2001). By use of an image capture device such as a video camera or digital camera, an image can be analyzed by application of colour analysis software. With computer based colour analysis, thresholds can be set to provide a consistent pass/fail evaluation of a product. This provides an unbiased and reliable evaluation without the inconsistencies of visual evaluation by a human panel. Two approaches were used to examine colour in this study. Red, green, and blue (RGB) colour space can describe over 16 million colours in 64 colour ranges by combining different amounts of red green and blue, with possible values ranging from 0 to 255 (Luzuriaga, Balaban, & Yerlan, 1997). Although the RGB method is capable of distinguishing between millions of colours, it is not representative of how humans perceive colour. The second approach is hue, saturation, and value (HSV) colour space. With this method, values range from 0 to 1.0 for each of the three variables. As hue moves from 0 to 1.0, colour varies from red, yellow, green, cyan, blue, magenta, and then back to red. Saturation values vary from 0 to 1.0, increasing with the amount of black in the colour. Value represents the brightness of the colours specified by hue; from 0 to 1.0 colours become increasingly brighter. HSV values can also be represented as values ranging from 0 to 255, making it easier to move between the RGB and HSV colour representations. The HSV colour space better represents how humans perceive colour, due to the varying levels of saturation and value. Spectrophotometers and tristimulus colorimeters can be used to evaluate objects with uniform surfaces. Nonuniform surfaces may be blended or processed to provide uniformity but this will provide only an average colour and not represent the original colour and appearance of the sample (Abbot, 1999). Computer analysis allows quantification and classification based on the original colour and appearance of the sample even if the samples are of varying size, shape, or texture (Studman, 2001). The aim of this investigation was to evaluate the use of image analysis by computer for

evaluation of shredded lettuce. Comparing evaluations by a human panel and by a computer will allow the development of an acceptable limit which can be used as a threshold for determining whether lettuce is still acceptable or should be discarded. The application of this technology was tested during the evaluation of four different lettuce treatments and the effectiveness in extending the shelf life of the lettuce in three experimental trials.

2. Materials and methods 2.1. Lettuce acquisition and treatments Lettuce samples were obtained from Pride Pak Canada, Inc. (Mississauga, Ont.) and were processed by using a method similar to that described previously (McKellar et al., 2004). Shredded lettuce was removed from the processing line prior to the cold (4 °C) chlorinated (100 ppm) wash stage. The lettuce was treated with warm (48 °C) chlorinated (100 ppm) water for 30 s in a stainless steel flume system, then immediately cooled to <20 °C using chilled tap water. The lettuce was then put back into the regular processing line, washed in cold (4 °C) chlorinated (100 ppm) water for 25 s, dried by centrifugation, and packaged in 1.150 kg amounts using 28 cm  30 cm plastic bags. The bags were stored at 4 °C for 24 h prior to shipping to the laboratory. The samples were received, coded, and then stored at 4 or 10 °C for up to 18 days. A randomized compete block design was used for the experiment, and three trials were conducted with similar treatments. 2.2. Sensory panel A human sensory panel was used to grade lettuce based on acceptability as described previously (McKellar et al., 2004). At each sampling time, a lettuce sample (100 g) from each of the four treatments was placed in a clear plastic bag and assigned a random number. The bags were placed on a laboratory bench, and evaluated by 12 panelists for overall acceptability based on visual qualities such as colour, dryness and texture. Panelists rated each sample on a scale of 1­5, called a loss of quality scale (LOQ). The following descriptors were used: 1, acceptable; 2, mostly acceptable; 3, somewhat acceptable; 4, mostly unacceptable; 5, unacceptable. Mean values with standard errors were reported (n ¼ 12). A score of 3 was considered to be the cut-off point between acceptable and unacceptable quality. 2.3. Image acquisition Images of the lettuce were captured using a digital camera (Nikon Coolpix 900). The camera was mounted

T. Zhou et al. / Food Research International 37 (2004) 875­881


on an adjustable stand 30 cm above the base and light was provided by two, 120 V frosted photographic floodlights, in a position providing minimum shadow and glare when photographing the samples. A white balancing was conducted on a gray board each time before photographing. The camera was set to 1.2Â digital zoom on automatic indoor focus with the flash turned off. These settings provided a close up view of the lettuce and covered the entire field of view. The camera was also set on XGA which resulted in a picture at 1024 Â 786 pixels. This picture was saved as a JPEG in RGB format at a 1/8 compression ratio on a personal computer for later analysis. Lettuce samples that had been evaluated by the human panel were removed from the bags and filled into a Petri dish to cover the entire bottom of the dish. The dish with lettuce was placed on the base of the camera stand and positioned to cover the entire field of view of the camera. Three to four lettuce samples were photographed for each treatment. Lettuce treatments were assessed seven times during the experiment at 1, 4, 6, 8, 11, 14, and 18 days, respectively, after the lettuce were processed and packaged. 2.4. Digital image analysis The stored digital images were analyzed for percent brown area and RGB changes using an image analysis program (Northern Eclipse version 6, by Empix Imaging, Inc., Mississagua, Ontario). Shades of brown were selected from images taken during the last two sampling days of the study. Several areas were sampled from the selected images using the eyedropper tool in Northern Eclipse program. Images were analyzed using the HSV colour model. A range of hue, saturation, and value was developed to specify which colours, shades, and intensities of brown should be included in colour measurements (Table 1). Hue outside the range in Table 1 began with green (>50) and black (<30). A wide range of saturation was used included as many varying intensities as possible without including browns that were too dark. Values greater than 165 included any glare that appeared in the image as well as the very white parts of lettuce that had lost pigment. Values below 45 included any black present because of shadows. The same HSV ranges were applied to each image to measure the same specified ranges of colour. After thresholding the image, the percentage of brown present in the image was measured. This meaTable 1 HSV ranges used to select brown colours from digital images of lettuce samples Hue Lower limit Upper limit 35 50 Saturation 145 255 Value 45 165

surement was repeated for each of the three to four images taken of each treatment and averaged to give an average percent brown. To store and analyze the data, a Dynamic Data Exchange Library was set up between Northern Eclipse and Microsoft Excel. Appropriate data were placed in a summarizing spreadsheet and used to create graphs of the data. RGB values were scanned by using the line scan tool in Northern Eclipse. A 1-pixel diagonal line was drawn manually across the image from top left to bottom right. Between 400 and 500 pixels were exported for each picture. The analysis program then took RGB values from under the line and exported them to Excel. These values were then averaged, summarized in a spreadsheet, and displayed graphically. 2.5. Statistical analysis All data were analyzed using Unistat 5.5 software (Unistat Ltd., London, UK). General linear model (GLM) procedures were used for the analysis of variance. Trials were considered as blocks in the analyses. Percentage data with a range greater than 40% were subjected to an arcsine-square-root transformation before the analysis of variance (Little & Hills, 1978). Differences among treatments were determined by the Tukey-HSD multiple comparison test. Correlation coefficients between human evaluation and digital measurement were obtained using the Spearman rank correlation test, and the regression equations in Fig. 4 were from the polynomial regression analyses.

3. Results and discussion 3.1. Human panel evaluation The stored lettuce was graded by a human sensory panel at each sampling time. Each treatment displayed varying degrees of quality throughout the study (Fig. 1). The four treatments were significantly different in terms of average LOQ readings of three trials when the data was analyzed using the Friedman two-way ANOVA with multiple comparisons with trials as blocks (Table 2). Lettuce that was incubated at 4 °C had better LOQ scores than lettuce stored at 10 °C (Fig. 1). From these particular trials, lettuce in treatments with 30-s warm water wash seemed to be more unacceptable than their corresponding controls. However, many factors could affect the shelf life of RTU lettuce as described previously (McKellar et al., 2004). The objective of this research was to develop a technique to evaluate quality objectively for shelf life of RTU lettuce, rather than focussing on an efficacious treatment. The direct visual observation by humans is subjective and may not be able to avoid biases. Also, because the


T. Zhou et al. / Food Research International 37 (2004) 875­881

Fig. 1. Loss of quality (LOQ) as determined by a human sensory panel for lettuce washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C). The values in the chart are means and their standard errors of three experimental trials.

Fig. 2. Percent brown area of lettuce washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C). The values in the chart are means and their standard errors of three experimental trials.

Table 2 Friedman's two-way ANOVA with multiple comparisons (TukeyHSD) for human visual sensory evaluation Treatmentsa 0s-4C 30s-4C 0s-10C 30s-10C Total


Cases 202 202 202 202 808

Rank sum 357.00 556.00 477.50 629.50 2020.00

Mean rank 1.77 2.75 2.36 3.12 2.50

Comparisonb A B C D

Lettuce was washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C), respectively. b Treatments with different letter are significantly different at the 0.05 level.

human panel evaluation used a numeric grading system where the scale ranges were only 1­5 in this case, detailed measures for each treatment, which are often needed, may not be obtained by visual evaluation. Since data collected from human panel evaluations were ranks of acceptance, nonparametric ANOVA must be used for its statistical analyses, which may result in loss of details in the data analyses as compared to parametric statistical analyses (Steel & Torrie, 1980). 3.2. Imaging analysis of percent brown area Data of percent brown area were obtained by image analysis using the developed HSV colour ranges (Table 1). Although some colours included in the ranges may have been too dark or too light to be considered ``brown'', each measurement applied the same HSV ranges. Consistent colour thresholds applied to images taken in the same lighting environment allowed for the measurement of changes in the same ``browns'' when measured in each image. The percent brown area showed a steady increase in the selected brown colour in all treatments (Fig. 2). Different treatments exhibited varying rates of brown-

ing in each sampling time. Analysis of variance was carried out using a GLM module of parametric statistics (Table 3) and clearly indicated that all factors, including with or without a warm water wash, storage temperature, and storage days, were significant as percent brown area was measured by digital image analysis. Similar to the trends identified by the human visual evaluation, the best treatment was the one that consisted of a cold wash and storage at 4 °C (0s-4C). The treatment had the lowest value in percent brown area, still below 10% after 18 day incubation (Fig. 2). The highest value was found with the treatment consisting of a 30-s wash of 48 °C water and stored at 10 °C (30s10C). Percent brown area was greater than 15% after the 8th day and increased rapidly, reaching to 30% on the 18th day. Comparison of an ineffective treatment and an effective treatment illustrate the possibility of digital measurement in distinguishing between varying degrees of browning, and the results clearly demonstrated that image analysis of browning was effective in determining difference in lettuce quality. 3.3. Digital Image analysis of colour changes As the shredded lettuce aged, overall colour appearance also changed. The colour compositions of each treatment were tracked using digital measurement. The starting values for each of the experimental trials were similar. This indicates that the method used was consistent and that each lettuce treatment had the same initial colour composition. The 30s-10C treatment illustrates a typical RGB change for aging lettuce. These colour changes were measured using the RGB colour space, as colour measurement could not be made in HSV with the Northern Eclipse software. In the graph of the RGB changes vs. time the 30s-10C treatment showed a dramatic drop in blue within the first few days (Fig. 3). Noticeable changes in red and green were also observed. A small drop in green, which is expected as the lettuce loses its green pigments was accompanied by a

T. Zhou et al. / Food Research International 37 (2004) 875­881 Table 3 Analysis of variance of percent brown area of lettucea Factorsb Main effects Treat Store Days Treat  store Treat  days Store  days Treat  store  days

a b


Sum of squares 9.972 0.778 1.352 4.976 0.001 0.385 0.833 0.131

DF 10 1 1 6 1 6 6 6

Mean square 0.997 0.778 1.352 0.829 0.001 0.064 0.139 0.022

F-values 148.308 115.669 201.097 123.337 0.080 9.554 20.660 3.235

Significance <0.0000 <0.001 <0.001 <0.001 0.7774 <0.0000 <0.0000 0.0048

Percentage data were arcsine-square-root transformed before analysis. Factors tested include: Treat (treatments) indicates different lettuce wash methods, with or without a warm water wash. Store (storage) means that treated lettuce was incubated at 4 and 10 °C; Days are number of days that lettuce has been stored post treatments.

few days of each treatment. Sensitivity of blue colour might be exploited if the reasons for this drop in colour were better understood (Luzuriaga et al., 1997). 3.4. Correspondence of acceptance and browning to days of storage Data obtained from the human panel evaluation were submitted to regression analyses. The correlation coefficients between LOQ and storage days were high and ranged from 0.9396 to 0.9987, with R2 values of 0.8385­0.9960 (Fig. 4). This clearly demonstrated that this technique might be applied not only to evaluate the acceptability of the produce from consumers' point of view, but also to predict the shelf life of the product. Regression analysis showed that the progression of lettuce browning corresponded well with days of storage, i.e. shelf life. The correlation coefficients between percent brown area and lettuce shelf life ranged from 0.9194 to 0.9941 with relatively high R2 values of 0.7622­0.9389 (Fig. 4), just slightly lower to those achieved in the human panel evaluation. In this study, 10­12 people were involved in the panel evaluations, producing 10­12 readings per sample each time every trial. Only three to four images were analyzed for percent brown area in each trial. Improved correlation and consistence should be achieved with increased number of images. 3.5. Relationship between browning and acceptance Correlation analyses were performed using percent brown area data vs. LOQ for individual trials, and the pooled data from all three trials are shown in Table 4. Although there was considerable variation among the trials and treatments, most values of the correlation coefficients are greater than 0.7, and many of them over 0.9, indicating lettuce browning measured by digital image analysis is highly correlated to human visual

Fig. 3. RGB values of lettuce washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 or 10 °C (4C and 10C).

small rise in red. The most significant changes all happen within the first 4­6 days of storage. The colour changes of the 0s-4C treatment were similar but much less dramatic; the change in blue was only about half that of the treatment of 30s-10C. After 4 days of aging, the 0s-4C treatment had lost much of its original crispness but was still very green. The samples seemed to get slowly darker; however, there was very little change in either the red or green values. Green decreased slightly while red increased slightly, similar to that observed with the 30s10C treatment. Because of the similar reactions of the red and green values they would not be suitable for characterizing changes in lettuce quality. These values may be useful as a general indicator over the first few days of aging. The best indicator in terms of RGB values is blue. It consistently showed greater changes in value within the first


T. Zhou et al. / Food Research International 37 (2004) 875­881

Treatment: 0s-4C

5 4 3 2 1 0 15 12.5 10 y = 0.0120x 2 - 0.0916x + 1.1522 7.5 CC=0.9987 R2 = 0.9893 5 2.5 y = 0.0553x 2 - 0.6268x + 1.3870 0 -2.5 CC=0.9194 R2 = 0.7622 -5 4 6 8 11 14 18

Treatment: 0s-10C

5 4 3 2 1 0 -1 1 4 y = 0.0058x 2 + 0.1812x + 0.3767 CC=0.9396 R2 = 0.914 40 35 30 25 20 15 10 y = 0.0762x 2 +0.6690x -2.9303 5 0 CC=0.9344 R2 = 0.9042 -5 6 8 11 14 18

40 35 30 25 20 15 10 5 0 -5

Loss of quality


Treatment: 30s-4C

5 4 3 2 1 0 -1 1 4 6 y = 0.0469x2 + 0.5975x - 1.6454 CC=0.9442 R2 = 0.9389 8 11 14 18 y = 0.0053x 2 + 0.1268x + 1.1656 CC=0.9641 R2 = 0.9107 30 25 20 15 10 5 0 -5

5 4 3 2 1 0 -1

Treatment: 30s-10C

y = - 0.0145x 2 + 0.5311x + 0.2787 CC=0.9946 R2 = 0.9142

y = -0.1155x 2 + 4.3088x - 8.6119 CC=0.9941 R2 = 0.8995








Days of storage

Fig. 4. Regression equations for LOQ (j, solid line) and digital image analysis of browning (, broken line) and their correlation coefficients (CC) of lettuce washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C). The lines in the graphs are polynomial regressions.

Table 4 Correlation coefficients between human visual evaluation and digital measurement of percent brown area analyzed using Spearman's rank correlation test Treatmentsa 0s-4C 30s-4C 0s-10C 30s-10C Means


Trial 1 0.6742 0.7487 0.9231 0.8469 0.8850

Trial 2 0.6301 1.0000 0.9550 0.9910 0.9730

Trial 3 0.5447 0.5932 0.9856 0.9000 0.9428

Pooledb 0.8571 0.9643 0.9643 1.0000 0.9822

Lettuce was washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C), respectively. b Analyzed using pooled data from three experimental trials.

evaluations. Fig. 4 illustrates the close correlations between browning development and acceptability changes. Digital image analysis of browning proved to be able to parallel the LOQ scores of the human panel. Using a LOQ threshold of 3, lettuce from 0s-4C, 30s-4C, 0s-10C,

and 30s-10C treatments was unacceptable at days 18, 11, 11, and 8, respectively, corresponding to 9.21%, 9.72%, 12.60%, and 16.94%, respectively, in browning digital analysis (Table 5). With the formula generated using the experimental data of this study, it is predicted that the lettuce shelf life at an LOQ reading of 3 would be 16.8, 10.2, 10.8, and 6.2 days, which corresponds to 6.46%, 9.32%, 13.18%, and 13.66% browning, for the treatments 0s-4C, 30s-4C, 0s-10C, and 30s-10C, respectively. These predicted values are very close to actual readings in the experiments. It may be difficult to create an universal formula for evaluating or predicting shelf life of RTU lettuce from varied treatments and storage conditions. A formula specific to an particular product or a treatment could easily be generated, however, and applied to monitoring quality and shelf life of the product, or evaluating the research treatment. Consumer acceptance of RTU lettuce may be determined by various sensory measurements, such as fresh

Table 5 Shelf life of RTU lettuce measured by human visual evaluation and digital image analysis of browning Treatmentsa Experimental datab Shelf life (days) 0s-4C 30s-4C 0s-10C 30s-10C

a b

Estimationc Acceptance 3.4 3.4 3.6 4.0 Browning (%) 9.21 9.72 12.60 16.94 Shelf life (days) 16.8 10.2 10.8 6.2 Acceptance 3.0 3.0 3.0 3.0 Browning (%) 6.46 9.32 13.18 13.66

18 11 11 8

Lettuce was washed with warm chlorine water for 30 s (30s) or without (0s) and stored at 4 and 10 °C (4C and 10C), respectively. Values are means from three experimental trials. c Values are estimated with regression equations generated using the experimental data in this study, shown in Fig. 4.

Percent brown area


T. Zhou et al. / Food Research International 37 (2004) 875­881


appearance, overall colour changes, portion of browning and possibly other sensing criteria (Casta~r et al., e 1999; Rocha & Morais, 2003). However, this study has demonstrated that it is possible to use digital image analysis in assessing the quality of and prediction of shelf life of RTU lettuce. With further more specific development of digital image analysis, this versatile and inexpensive technique (Papadakis, Malek, Kamdem, & Yam, 2000) could be adopted by the food processing industry as a reliable objective and quantitative tool for monitoring the quality and shelf life of RTU products.

Acknowledgements The authors thank J. Boulter, X.Z. Li, K. Knight, H. Zhu for their excellent technical assistance. References

Abbot, J. A. (1999). Quality measurement of fruits and vegetables. Postharvest Biology and Technology, 15, 207­225. Bolin, H. R., Stafford, A. E., King, A. D., Jr., & Huxsoll, C. C. (1977). Factors affecting the storage stability of shredded lettuce. Journal of Food Science, 42, 319­1321. Casta~r, M., Gil, M. I., Ru M. V., & Arts, F. (1999). Browning e iz, e susceptibility of minimally processed Baby and Romaine lettuces. European Food Research and Technology, 209, 52­56. Coles, G. D., Peter, J., Ammerink, J., & Wallace, A. R. (1993). Estimating potato crisp colour variability using image analysis and a quick visual method. Potato Research, 32, 127­134. Delaquis, P. J., Stewart, S., Toivonen, P. M. A., & Moyls, A. L. (1999). Effect of warm chlorinated water on the microbial flora of shredded iceberg lettuce. Food Research International, 2, 7­14.

Jilliffe, P. A., & Lin, W. C. (1997). Predictors of shelf life in long English cucumber. Journal of American Society for Horticultural Science, 122, 686­690. Kranzier, G. A. (1985). Applying digital image processing in agriculture. Agricultural Engineering, 66, 11­13. Little, T. M., & Hills, F. J. (1978). Agricultural experimentation. New York, NY, USA: Wiley, p. 159. Luzuriaga, D. A., Balaban, M. O., & Yerlan, S. (1997). Analysis of visual quality of white shrimp by machine vision. Journal of Food Science, 62, 1­13. McKellar, R. C., Odumeru, J., Zhou, T., Harrison, A., Mercer, D. G., Young, J. C., Lu, X., Boulter, J., Piyasena, P., & Karr, S. (2004). Influence of a commercial warm water/chlorine treatment on the shelf life of ready-to-use lettuce. Food Research International, 37, 343­354. Odumeru, J. A., Boulter, J., Knight, K., Lu, X., & McKellar, R. (2002). Assessment of a thermal­chemical process to extend the shelf life of ready-to-use lettuce. Journal of Food Quality, 26, 197­ 209. Papadakis, S. E., Malek, S. A., Kamdem, R. E., & Yam, K. L. (2000). A versatile and inexpensive technique for measuring color of foods. Food Technology, 54, 48­51. Priepke, P. E., Wei, L. S., & Nelson, A. I. (1976). Refrigerated storage of prepackaged salad vegetables. Journal of Food Science, 41, 379. Rocha, A. M. C. N., & Moraise (2003). Shelf life of minimally processed apple (cv. Jonagored) determined by colour changes. Food Control, 14, 12­20. Steel, R. G., & Torrie, J. H. (1980). Principles and procedures of statistics ­ a biometrical approach (2nd ed.). New York, NY, USA: McGraw-Hill, p. 534. Studman, C. J. (2001). Computer and electronics in postharvest technology ­ a review. Computers and Electronics in Agriculture, 30, 109­124. Tunde, V., & Feldoldi, J. (2000). Enhancing colour differences in images of diseased mushrooms. Computers and Electronics in Agriculture, 26, 187­198. Wang, H. H., & Sun, D. W. (2001). Evaluation of the functional properties of cheddar cheese using a computer vision method. Journal of Food Engineering, 49, 49­53.



7 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in