Systems, apparatuses, and methods for detecting early drought stress in plants
A remote inspection vehicle for inspecting a plant for early drought stress and a neural network model for calculating a score corresponding to a level of early drought stress in the plant. The inspection vehicle may be equipped with an inspection camera having a radiometric infrared sensor and a red, green, and blue light sensor to capture plant images. Image data from the inspection vehicle, as well as soil moisture and calculated CWSI scores, can be used to train the neural network model, deriving weights and biases therefrom using a plurality of neural network layers. Using a plant canopy temperature value, a red raster value, a green raster value, a blue raster value, and a soil moisture value, the trained neural network model can generate a score which can be used to assess the level of early drought stress in a plant.
1. A method for determining the probability of early drought stress in a plant comprising the steps of:
a) creating a neural network model having at least three layers;
b) obtaining a radiometric infrared image file of said plant;
c) obtaining each of a red, green, and blue light reflectance image file of said plant;
d) providing each of said radiometric infrared image file, said red light reflectance image file, said green light reflectance image file, and said blue light reflectance image file to said model; and
e) generating a score associated with said plant;
wherein said step of creating said model comprises creating a training database, said training database consisting of a plurality of databases of samples, each of said samples comprising a temperature value, a red raster value, a green raster value, a blue raster value, a moisture value, and a calculated CWSI value.
2. The method of claim 1 , wherein said temperature value is extracted from a radiometric infrared image.
3. The method of claim 1 , wherein each of said red raster value, said green raster value, and said blue raster value are extracted from a light reflectance image.
4. The method of claim 1 , wherein said CWSI value is calculated by taking an average of a plurality of said temperature value.
5. The method of claim 1 , further comprising the step of:
f) determining a plurality of weights and biases of the neural network model by using about eighty percent (80%) of said samples in said database.
6. The method of claim 1 , further comprising the step of:
f) optimizing the model with a loss function of mean squared error, and metrics of mean squared error and mean absolute error.
7. The method of claim 1 further comprising the step of:
f) applying an early stopping function to prevent overfitting.
8. The method of claim 1 , further comprising the steps of:
f) obtaining a soil moisture value associated with said plant; and
g) providing said soil moisture value to said model.
9. A system for determining early drought stress in a plant, comprising:
a) a processor and a memory, said memory comprising instructions adapted to generate a drought assessment score associated with said plant from a radiometric infrared image and a red, green, and blue light reflectance image of said plant; and
b) a remote inspection vehicle for obtaining said images of said plant, said remote inspection vehicle comprising a (1) data transceiver and (2) an inspection camera, said inspection camera comprising a first image processor engaged with a radiometric infrared sensor and a color light sensor, and configured to generate said images therefrom,
wherein said instructions comprise a neural network model having at least three layers, and is trained and optimized using a database of samples, each said sample comprising a temperature value, a red raster value, a green raster value, a blue raster value, a moisture value, and a calculated CWSI value.
10. The system of claim 9 , said remote inspection vehicle further comprising:
(3) a plurality of motors;
(4) a motion control circuit comprising a processor electrically engaged with said plurality of motors and a memory for storing instructions, said instructions adapted to cause said motion control circuit to move said motors; and
(5) a movement transceiver electrically engaged with said motion control circuit and adapted to interface with a remote transceiver and receive said instructions.
11. The system of claim 10 , wherein said motion control circuit comprises a position determining circuit.
12. The system of claim 10 , wherein said remote inspection vehicle further comprises:
(6) a main controller, said main controller comprising a processor and a memory for storing instructions, said main controller electrically engaged to said motion control circuit, said data transceiver, said inspection camera, and said movement transceiver.
13. The system of claim 12 , wherein said instructions are adapted to cause said motion control circuit to move said motors.
14. The system of claim 12 , wherein said instructions are adapted to cause said controller to receive and process said images.
15. The system of claim 12 , wherein said instructions are adapted to cause said controller to transmit said images to a remote memory through said data transceiver.
16. The system of claim 9 , wherein said temperature value is extracted from a radiometric infrared image generated from said radiometric infrared sensor.
17. The system of claim 9 , wherein each of said red raster value, said green raster value, and said blue raster value are extracted from a light reflectance image generated from said color light sensor.
18. The system of claim 17 , wherein said color light sensor comprises an RGB sensor with an infrared filter.
19. The system of claim 9 , wherein said model is further trained and optimized with a soil moisture value associated with said plant.