Phenotyping with multispectral imaging

In physiological phenotyping, multispectral channel-wise imaging gives more informative data compared to simple photography. Presence and combination of various biochemical compounds are the main determinant of the plant’s physiological status. As the biochemical properties of the plant’s surface determine optical properties, multispectral imaging can give insight in the physiological status of the plant. Spectral signatures serve as indicator of stress that is evoked by abiotic and biotic factors. We designed a multispectral imaging system that uses a 16 MPix Greyscale camera in combination with different LEDs and a Filter wheel. The LED array has a range of channels, and our current setup consists of UV, blue, green, yellow, red, deep red, far red and near-infrared wavelength LEDs. Thereby, an advanced analysis of the red edge is enabled as different images in the red range are provided and can be used for ratio calculation. The filter wheel enables filtering for reflectance or fluorescence, e.g., in the chlorophyll fluorescence range.

Having the images in the different channels, we can not only calculate an RGB image for the objects, but we can use the channels to assess physiological features. The RGB image provides an insight how the objects look like and allow the same analyses as they can be done with classical photographs.

In the example there is a fresh leaf, a dead dry leaf, and a leaf with some stress-induced damages. With the RGB image, size, shape, and color of the leaves are accessible and the stress-related spots on the bottom leaf can be seen. The dried leaf has completely lost chlorophyll so it looks brown.

While such analyses that are congruent with the human visual impression already give important information about the phenotypes, multispectral imaging can do more. Looking at the red, green, and blue channels separately, the imaging system can better differentiate healthy living tissue from dead tissue or damaged tissue. The capabilities to analyze stressed tissues with damages from disease or from adverse environmental factors are improved by the broader range of spectral information.

The dry leaf massively reflects in the red and blue channels whereas the fresh leaf does not, because red and blue light are absorbed by the chlorophyll. The chlorophyll loss of the dry leaf causes that this leaf does not absorb red or blue light anymore. In turn, the spectral data indicate that the chlorophyll was degraded. In the stress damaged leaf, particularly the red channel is indicative for the damages. Here, the reflectance in the red range was enhanced, but not in the blue range, which is a hint that there is still some chlorophyll present, however it is partly overlaid by physiological damage that causes the brown-looking spots. In the example the damage was induced by shortage of water and nutrients, so it is likely that the photosynthetic apparatus war at least partly impaired.

Combining red and NIR channels, the system can image NDVI data. The NDVI is one of the most common indices for plant vitality. NDVI imaging is not only used at field scale, but increasingly serves to monitor plant quality in various applications. Living healthy tissue has high NDVI values, while these values are low in the damaged areas of the stressed leaf. The dryed dead leaf has lowest NDVI values. The water- and nutrient stress induced damage on the bottom leaf led to partial reduction of the vitality in the leaf, as seen by the reduced NDVI from the leaf base through the lower half of the leaf.

As chlorophyll fluorescence is a key indicator of plant health and performance, measuring this type of fluorescence is frequently done in biology. While there are specialized systems that capture dynamics of chlorophyll fluorescence, the multispectral fluorescence imaging system allows static images of chlorophyll fluorescence. Although they are not informative about photosynthetic activity, they still indicate presence and abundance of chlorophyll and indicate damages of the photosynthetic apparatus.

In the example, the dry leaf has very weak chlorophyll fluorescence in the transition zone of petiole and blade, indicating that this area was not yet completely dead. Major parts of the dry blade have no chlorophyll fluorescence, indicating that the tissue is dead. In the leaf with the stress symptoms, the stressed area has much less chlorophyll fluorescence compared to the rest of the leaf. Even in the fresh unstressed leaf chlorophyll fluorescence is not distributed equally in all parts of the blade. This might be related to physiological differences within the leaf, or with the curvatures of the leaf surface.

click image to enlarge

Multispectral Imaging
Calculated RGB image
Red, green, and blue channel imaging – color-coded visualization of reflectance
Combining red and infrared image (color-coded visualization of reflectance) to NDVI image (greyscale intensity of NDVI values: light-grey high NDVI, dark-grey low NDVI)
Excitation of fluorescence with different blue LEDs and recording using a bandpass filter at 695 nm