Tomato Analyzer measures morphological and color attributes via image analysis in an objective, highthroughput, and semiautomatic manner. This software allows for reproducible quantification of phenotypic data that previously were done by hand or visual analysis. The new version has improved the accuracy of all measurements and reduced the need to make timeconsuming manual adjustments. In this paper new morphological and color attributes available in Tomato Analyzer 3.0 as well as how the color test module was made more user-friendly are described.
The species in the plant kingdom are characterized by a great diversity in color, shapes and size displayed in organs such as leaves, flowers, and fruits. Even within a particular species the individuals also can be distinguished by the morphology and color displayed in those organs. Biologists trying to understand the genetic and molecular basis for this variation need to measure morphological and color attributes in an objective and reproducible way. Most of this type of phenotypic analysis consists of timeconsuming manual measurements or subjective visual scoring of characteristics that reduce the success of identifying genomic regions or physiological causes underlying this variation.
Tomato Analyzer (TA) is a software program designed to
collect objective data from digital images obtained from
plant organs
This paper describes how morphological and color analysis of plant organs can be precisely done using Tomato Analyzer. Lastly, some possible applications of Color Test in Tomato Analyzer 3.0 are discussed.
Tomato Analyzer can do a high-throughput analysis of morphological traits in images obtained from plant organs. Moreover, the data obtained are unbiased compared to those manually measured by different researchers or using different instruments. This impartiality allows the reproducibility of the experiments as well as the compilation and analysis of data obtained from experiments conducted in several environments and years. To date, most of the morphological classifications in plants were made based on eyeball observations. Instead, attributes of TA can be used to objectively classify plant organs into various morphological categories.
Tomato Analyzer 2.0 has been a valuable and effective tool to indentify and confirm genomic regions that control tomato fruit shape as well as performing in-depth analyses of the effect of key fruit shape genes on plant morphology. It was possible due to color of tomato fruits are contrasting enough with a black background used in the scanned images. However, the software was unable to detect other darker fruits or leaves.
Scan plant organs against a black background to eliminate shadows.
Open the image in TA, select the attributes to measure, tell TA to analyze the image.
TA separates objects from background. It does this by looking at a histogram of luminance for the image and finding the separation point (area of low histogram values) between the foreground (lighter colors) and the background (darker colors). It then finds contiguous areas of foreground pixels (the objects) and calculates the boundaries around them.
Resulting data appears in a spreadsheet panel in TA (screenshot) and can be exported as .csv file.
Some measurements can be manually adjusted.
The previous version of Tomato Analyzer read only JPEG (.jpg) files but this new version can open both JPEG (.jpg) and TIFF (.tif) files. TIFF files are recommended because they preserve the image as it was originally scanned. JPEG images alter some of the colors in the image, reducing the accuracy of object boundary detection and color analysis. For example, this new feature has improved the boundary detection in dark green leaves and cucumber fruits. This improves the accuracy of all measurements and reduces the need to make time-consuming manual adjustments.
One of the most important features to analyze fruit shape in tomato is the fruit shape index (defined as the ratio between the width and the height of the fruits). When the tomato fruit or other type of fruits as cucumber is curved, the values for this index do not represent the actual value, they are underestimated. A new measurement, named curved height, was added in the new version of Tomato Analyzer (Figure 1). This attribute allows an accurate estimation of the length on curved fruits as well as is possible to be measured a new fruit shape index attribute. Width Mid-height (W_mid) – The width measured at ½ of the fruit’s height. Maximum Width (W) – The maximum horizontal distance of the fruit. Height Mid-width (H_mid) – The height measured at ½ of the fruit’s width. Maximum Height (H) – The maximum vertical distance of the fruit. Curved Height (CH) – The height measured along a curved line through the fruit (passing through the midpoints of opposing pairs of points on either side of the distal and proximal points). Fruit Shape Index External I (H / W) – The ratio of the Maximum Height to Maximum Width. Fruit Shape Index External II (H_mid / W_mid) – The ratio of Height Mid-width to Width Mid-height Curved Fruit Shape Index (CH / CW) – The ratio of Curved Height to the width of the fruit at mid-curved-height, as measured perpendicular to the curved height line.
To measure shape without selecting individual attributes,
TA offers a morphometric or geometric analysis of each
object. This function finds points along the boundary of
each tomato slice in the loaded image. Statistical tools such
as Principal Component Analyses can be used to analyze
the points in the exported data and it has been used to
identifying tomato genome regions that control fruit
morphology
Color is an important quality attribute in horticultural and floricultural crops defined by particular genes. However, color changes in plant organs also are indicators of biotic or abiotic stresses. The genetic bases for color attributes as well as factors affecting plant health would be better understood if computer-based analysis of digital images is applied instead of their subjective characterization. The Color Test module in Tomato Analyzer (TACT) is able of collecting and analyzing color parameters in an efficient, accurate and high-throughput manner from scanned images that contain plant organs. However, the scanner needs to be calibrated if the user intends to translate RGB values to L*, a* and b* parameters. This is because scanners may change in accuracy and how they capture the color scheme over time. Moreover, scanners differ in how well they capture color information.
Scan a color checker. Open the image in TA and perform a color calibration.
Scan objects and open the image in TA, as you would for morphological analysis.
Select the color attributes to measure, and tell TA to perform color analysis.
Resulting data appears in the spreadsheet panel, as in morphological analysis.
The most important improvement related to the color test module in this new version of TA is that the results are shown in the real time on the screen shot of the software (Figure 3). In the previous version, the results only could be visualized in .csv files after the analysis was done. The user should chose a color checker with a black or very dark background based on the broad range of colors observed in the object of interest (Figure 4). Color checkers can be purchased custom made or standard.
After the color checker was scanned should be opened and analyzed in TA. The software recognizes each tile as an object. Then, the user needs to enter the actual L*,a*,b* values for each tile in the color checker, as provided by the manufacturer (Figure 4). TA uses the actual and observed L*, a*, and b* values to calculate the linear regression. After that, the color module is calibrated.
Average color values. The values displayed are: Average Red, Average Green, Average Blue, Average Luminosity, Average L* Value, Average a* Value, Average b* Value, Average Hue, Average Chroma. These average values are calculated taking account all pixel within the object.
L*, hue, chroma distributions. These measurements provide histogram data for L*, hue, and chroma. The data appear in the tabs called L* Distributions, Hue Distributions, and Chroma Distributions, respectively. Each column shows the fraction of the object whose color falls within a certain range. For example, if the L[40..50) column in L* Distributions has the value 0.3, then 30% of the fruit has L* between 40 (inclusive) and 50 (exclusive).
Set custom color parameters. Based on the L*, hue,
chroma distributions, the user can define custom
ranges of L*, hue, chroma, or a combination of the
three. The User-Defined Color Ranges dialog appears
as shown in Figure 5. The user can define up to 6
combinations of color ranges. In the example in
Figure 5, Parameter 1 includes all colors where the
hue is between 30 (inclusive) and 45 (exclusive); L*
and chroma may be anything. Parameter 2 includes all
colors where the L* is between 0 and 30 and the hue is
between 0 and 90; the chroma may be anything. The
data will appear in the data window tab called Custom
Color Parameters.
The Color Test module in Tomato Analyzer 3.0 is able to
define the average for several color attributes inside the
boundaries of a plant organ as well as the proportion of six
different user-defined color parameters. This new feature
can be useful to study pattern of color variation in some
plant organ such as petals or leaves. In the plants, soil
nutritional deficiencies, pesticides toxicities and even the
severity of pathogen attacks affect the color pattern on
some plant organ. For example, it has been demonstrated
that the estimation of severity in a specific corn disease is
highly affected by the rater experience when they directly
estimate percentage of diseased leaf area and even more
when they use a 0 to 9 ordinal rate scale
This project was funded by the National Science Foundation DBI 0227541 to Esther van der Knaap