You are on page 1of 7

Scanalyzer3D

Approaches in Heat imaging

First issued 1st October 2010

Introduction
Progress in technology and an increasing knowledge of modern kinetic plant phenotyping
are opening up new perspectives on the measurement and understanding of plant development. Particularly advanced, non-destructive imaging in combination with specific
plant sensors and appropriate modelling of the resulting data provides the best options
for gaining sufficient knowledge in optimally targeted breeding, even for complex kinetic
phenotype traits such as stress resistance and tolerance.
For a better understanding of how plant phenotypes are expressed in relation to genetic
and environmental conditions, it becomes increasingly important to state that a phenotype is quite a complex functional network, represented by a large set of quantitative
parameters. These include, besides genetics, all morphological and growth patterns,
depending on time and development, as well as specific reactions towards short- and
long-term environmental conditions, particularly stressors like nutrient deficiency, water
logging, pests, extreme temperatures or drought.
To gain such an understanding, it is important to grow larger numbers of plants under
individually well-controlled conditions. In any case, a comprehensive and non-destructive
measurement of each individual plant over a prolonged time span is absolutely essential.
To be able to separate factors that finally lead to a certain phenotype or measurement
value, the use of appropriate data models is extremely important.
The following paragraphs will highlight some aspects of test design, measurement and
model requirements for plant heat imaging with LemnaTec scanalyzer3D conveyor systems.

The Special Dynamics of Heat Imaging


Heat imaging of plants has some very specific differences to most other types of imaging,
such as visible light imaging, fluorescence imaging or near infrared imaging. In no other
range of electromagnetic waves is the radiation emission so closely related to environmental conditions. At the same time, plants can change leaf temperature much faster
than most other emission parameter by simply opening or closing their stomata.
This huge dependence of the imaging result on one external and one internal factor is a
decisive aspect for the design of heat imaging approaches.
Plants with closed stomata exhibit practically ambient temperature, which is of course
highly changeable (with variations of more than 10 C during the course of the day). In
case of direct solar irradiation, leaves exposed to the sun can have leaf temperatures that
are even 10 or more degrees higher than the ambient air temperature.
While thin plant parts adapt their temperature within seconds or minutes to changes in
ambient temperature, bigger parts like stems need much longer, because of their large
heat capacity.
Plants with open stomata cool down relative to the temperature the leaves would have
without water evaporation. The measured leaf temperature is therefore a result of both
ambient conditions and an incremental reduction of temperature caused by water evaporation. The amount of water evaporating from completely opened stomata again depends
2

on additional environmental conditions such as air humidity and wind speed. Both factors
can show massive fluctuations, both in the field and in greenhouses with air ventilation
and evaporative cooling.
As a result, it becomes obvious that leaf temperature is the integrative result of several
influencing factors. While some of these factors can be controlled, many others must be
assessed and normalised by appropriate statistical and imaging approaches.

Heat Imaging a Short Physical Note


Whenever heat imaging is performed, the results of a calibrated and colour-encoded image
suggest a measurement of an absolute temperature, either in a defined spot on the plant,
on the leaf or in the background. The results look quite similar to a temperature measurement taken with a normal contact thermometer. In fact, there is one specific difference as
any imaging system measures the radiation emission of the object. The physical heat
radiation emission of objects, however, is not only influenced by their temperature, but
also by material properties. These properties are accumulated in the parameter emissivity
(EPSILON), which is around 0.9 for water or wood, but only in the range of 0.1 for
polished aluminum. The variations lead to quite different emission values (and consequently colour variations on a false colour thermal image), even if the test objects themselves have an identical temperature. This physical background of heat radiation measurement is another point that makes it advantageous to focus on relative differences, as
explained below. One should simply not expect a physical temperature measurement as a
result of plant imaging.
In no way does this mean a reduced applicability of plant phenotyping technologies, as in
the end all visible light imaging is also a relative measurement and as such does not provide information on the concentration of a specific dye in the leaves (which nobody would
expect).

Controlling the Influencing Factors


While absolute control of environmental parameters in greenhouses is limited due to the
strong dependency on the solar energy input, the LemnaTec MovingField conveyor concept allows randomisation of plants during the days to equilibrate factors caused by air
humidity and wind speed.
As an extremely important additional feature, all plants go through the same environmental conditions before entering the imaging unit. All plants stand in the same queue, and
environmental conditions in this queue may be either strictly controlled or merely
represent the overall conditions in the greenhouse. Most important is that the same conditions for all tested plants are provided.
The highly automated plant-by-plant transport from the storage area to the imaging area is
the best way to deal with fluctuating climate conditions in the greenhouse. The plants
fast reaction towards heat changes and their inherent heat capacity integrate a kind of
thermal history over the last minutes to hours.
3

Fig. 1: Dynamics of temperature change in a banana plant (over a time span of 300 s),
after reducing light intensity from sunlight to dim light. Increase in yellow colour signifies
that the plant is warming up due to reduced evaporation. The green stem of the plant
reacts much more slowly because of its high heat capacity.

The LemnaTec Heat Imaging Concept


Heat imaging is also extremely susceptible to environmental factors such as reflections
from metal parts in the greenhouse or glass panels. For this very reason, LemnaTec systems take all images in a closed box to eliminate eternal influences. To avoid irradiation of
the cabin on the plants, it is kept at the current environmental temperature. As a result,
plant parts with close to ambient temperature are relatively difficult to separate from the
background. This issue is solved by matching the visible light image with the infrared
image. In this way, heat silhouettes of plants can be easily separated from any kind of
background in the IR-imaging cabin.

The Importance of Relative Differences


While the points mentioned above greatly help to provide reproducible results through
heat imaging, the parameter inherent influence of the absolute environmental temperature
(+ air humidity, + wind speed) cannot be entirely eliminated in greenhouses and can only
be reduced to a certain minimum level in growth chambers. Therefore, the absolute value
of leaf temperature has almost no importance. Relative values are much more important,
and therefore cameras are optimised to image small differences, and not to measure
absolute temperatures with a maximum of accuracy. Consequently, even differences in
colour shades between IR-cameras are of much less importance than the appropriate relative differences in identification.
4

Various Kinds of Relative Differences Providing Information


General Approach
As described above absolute values of leaf temperature data have virtually no significance
as such for an individual plant as the leaf temperature distribution is a highly integrative
parameter of many plant and environmental factors. For this reason, a significant amount
of specific data extraction as well as statistics and modelling are needed. The following
subchapters will highlight some of the key issues in this field.

Mean Temperatures Versus Temperature Distributions


While calculation of mean plant temperatures looks easier than calculation of mean
colours, the main problem remains similar: How to average temperatures or colours while
retaining a sound meaning of the result? How much information is lost in averaging? In
theory, averaging is generally based on the hypothesis of a near to normal distribution of
the individual numbers, giving a meaning to the average value. In the case of thermal
imaging, different plant parts show different temperatures for very specific reasons.
These may be different orientation and exposure to heat or wind, different abilities to
evaporate water or different heat capacities under changing conditions. Most of these
influential factors are quite closely related to the plant phenotype and may therefore generate quite different patterns. For this reason, the basic hypothesis of averaging by normal
distribution or by assuming at least some self-similarity between plants does not hold.
Similar to the approach for colour classifications, LemnaTec recommends working with a
set of temperature classes, thus forming a quantitative temperature distribution for each
individual plant.
It remains a scientific question to be tested for individual conditions of species and greenhouses in which cases average temperatures are sound enough estimates.

General Plant to Plant Differences


If it is expected that a majority of plants behave normal concerning leaf temperature
distribution, it makes sense to calculate a kind of moving average for a sequence of
plants imaged one after the other. Based on such a moving average, outliers become
interesting. In how far these outliers are identified in a dynamic, statistical procedure after
the data has left the LemnaBase, or if such values should be calculated as a kind of deviation from a mean value within the LemnaBase: These are questions that would need to be
answered for each experiment, based on a set of real data. It will also be an issue of research if a simple mean value contains sufficient information or if relative differences of
heat classes (here substituting the normal colour classes) are a better basis for any outlier
statistics or normalisation.

Fig. 2: Model graph for a moving average mean temperature of a sequence of plants,
including the local outliers. It becomes obvious that the sequence of the plants is decisive, not the absolute values.

Differences to Control Plants


As in many tests specific control plants (wild type, etc.) are also placed in any blocked
design of plants on each conveyor belt, another option could be simply to take these
plants as reference for normalising the temperature values of all real screen plants that are
to be phenotyped. Only real datasets, which are analysed by different strategies, can
deliver exhaustive information on best approaches. These may differ greatly, from species
to species as well as within specific environments in the greenhouse.

Plant to Material Differences


While plants may change their evaporation and therefore their temperature behaviour significantly, many other parts of the pot and car transported together with the plants cannot achieve this. These car or pot parts just adjust to the environmental irradiation conditions and keep a certain temperature, depending on their colour, orientation and material.
Under fluctuating conditions, these parts may change their surface temperature more
slowly than the plants themselves, according to the specific heat capacity of the respective material. Please do also consider emissivity issues (as mentioned above) when the
focus is NOT strictly on relative differences.
With regard to the plants on the conveyor, the pots and different parts of the cars can be
used as such reference parts for the normalisation of the plant temperature. Of special
interest could be the pots (high heat capacity), the blue plate (low heat capacity) and the
black pillars (presumably medium heat capacity).
In all cases, it should be considered that infrared cameras do not measure a surface temperature, but an emission of heat irradiation related to the surface temperature and also to
material and surface properties. For further details, please consult the specific literature
on heat imaging.

Again, it is a scientific problem to prove that the concept of using car parts for the normalisation of leaf temperature is successful, and it needs to be checked against larger real
datasets. Infrared imaging under high-throughput conditions that will produce reliable
data/phenotypic parameters is a new field, going far beyond depicting two plants with
different leaf temperatures at the same point in time on one image. Lastly, this is a question of appropriate data modeling of heat exchange processes.

Conclusion
Optimised combinations of phenotyping hardware, imaging and sensor technology,
together with extensively enhanced test designs and sound data processing in models and
advanced statistics form the best basis for any high-throughput or high-content plant
phenotyping. This is valid for imaging in general, but even more important for highly dynamic processes such as heat imaging.
The whole domain of high-throughput heat imaging especially under fluctuating conditions that really integrate the dynamics of leaf temperature changes, and not just as a
snapshot of interesting plants on one image is a relatively new area of research. Having
the appropriate technical measurement structures with the LemnaTec scanalyzer3D imaging and the MovingField conveyors will provide the best opportunity to develop this field
of research further, in a scientifically valuable way.
For further information, please contact
Matthias Eberius
LemnaTec GmbH
Pascalstr. 59
52076 Aachen, Germany
Tel. +49 2408 9383 000
Fax +49 2408 9383 300
matthias.eberius@lemnatec.com

You might also like