System for detection of disease in plants
The present inventors have recognized that various diseases in plants, such as Phytophthora infestans (late blight) and Alternaria solani (early blight), and/or various stages of such diseases in plants, can be reliably detected by applying measurements from electromagnetic reflections detected from a plant in a model to produce an output indicating a probability of the disease and/or stage. In one aspect, coefficients can be applied to each measurement at each wavelength to emphasize identification of a given disease or stage. In another aspect, an imager can capture images comprising spectral pixels in which each pixel comprises measurements from the electromagnetic reflections for application in a model to identify a given disease or stage.
1. A system for detection of disease in plants, comprising:
a spectrometer configured to detect a spectrum of electromagnetic radiation reflected from a plant at a plurality of wavelengths, the spectrum comprising reflection measurements corresponding to wavelengths; and
a processor executing a program stored in a non-transient medium to apply the reflection measurements as variables in a predetermined model configured to indicate a likelihood of presence of a disease in the plant and output the likelihood of presence of the disease in the plant;
wherein the model provides an array of coefficients multiplied by each reflection measurement, wherein each coefficient corresponds to a given wavelength and emphasizes identification of a disease or stage of infection of disease.
2. The system of claim 1 , wherein the coefficients are determined by applying a partial least squares discriminant analysis with respect to first and second spectrums, the first spectrum comprising reflection measurements corresponding to wavelengths from a plant free from the disease or stage of infection of disease, and the second spectrum comprising reflection measurements corresponding to wavelengths from a plant having the disease or stage of infection of disease.
3. The system of claim 1 , wherein the reflection measurements correspond to wavelengths between 400 and 2400 nanometers.
4. The system of claim 3 , wherein the reflection measurements are provided in increments of at least 1 nanometer.
5. The system of claim 1 , wherein the reflection measurements are applied to indicate a likelihood of presence of Phytophthora infestans ( P. infestans ) or Alternaria solani ( A. solani ).
6. The system of claim 1 , wherein the reflection measurements are applied to indicate a likelihood of presence of one or more stages of infection of P. infestans.
7. The system of claim 6 , wherein the stages of infection comprise:
early infection; biotrophic growth; necrotrophic lesion formation; and sporulation.
8. The system of claim 6 , wherein the model produces an output ranking the likelihood of presence of each stage of infection.
9. The system of claim 1 , wherein the reflection measurements are transformed into spectral values, each spectral value quantifying a relative difference between reflection measurements at differing wavelengths to emphasize identification of the disease.
10. The system of claim 9 , wherein the spectral values are Normalized Differential Spectral Index (NDSI) values, each NDSI value being calculated as a difference between reflection measurements at differing wavelengths divided by a sum of the reflection measurements at the differing wavelengths.
11. The system of claim 10 , wherein the model applies no more than ten NDSI values for indicating the likelihood of presence of the disease or stage of infection of disease.
12. The system of claim 1 , wherein the processor further executes to send the likelihood of presence of the disease or stage of infection of disease to a graphic display.
13. The system of claim 1 , wherein the spectrometer is attached to a contact probe, an agricultural implement or an aerial vehicle.
14. A system for detection of disease in plants, comprising:
an imager configured to a capture an image comprising a plurality of spectral pixels, each spectral pixel corresponding to a spectrum of electromagnetic radiation reflected at a plurality of wavelengths, the spectrum comprising reflection measurements corresponding to wavelengths; and
a processor executing a program stored in a non-transient medium to apply reflection measurements at each spectral pixel as variables in a predetermined model configured to indicate a likelihood of presence of a disease of a plant at the spectral pixel and output the likelihood of presence of the disease in the plant
wherein the model provides an array of coefficients multiplied by each reflection measurement, wherein each coefficient corresponds to a given wavelength and emphasizes identification of a disease or stage of infection of disease.
15. The system of claim 14 , wherein the coefficients are determined by applying a partial least squares discriminant analysis with respect to first and second spectrums, the first spectrum comprising reflection measurements corresponding to wavelengths from a plant free from the disease or stage of infection of disease, and the second spectrum comprising reflection measurements corresponding to wavelengths from a plant having the disease or stage of infection of disease.
16. The system of claim 14 , wherein the reflection measurements correspond to wavelengths between 400 and 2400 nanometers.
17. The system of claim 14 , wherein the reflection measurements are transformed into spectral values, each spectral value quantifying a relative difference between reflection measurements at differing wavelengths to emphasize identification of the disease.
18. The system of claim 14 , wherein the processor further executes a filter to remove spectral pixels which do not correspond to plants.