Synthetic agricultural sensor

Abstract

In order to predict plant stresses at a localized level, data feeds from many sensor types can be fused and analyzed to create a synthetic sensor estimating plant water stress, predicting microclimatic conditions, and performing localized plant disease and pest modeling. To make this affordable, an array of low-cost, lower precision sensors can be used. Sensor fusion is used to improve the accuracy of each sensing element by using machine learning to fuse data from the other sensing elements in the array. Additionally, machine learning can create a “synthetic sensor” replicating the output of high-cost and maintenance intensive sensing devices by using machine learning to replicate their output.

What is claimed is: 
     
         1 . A method of generating a synthetic sensor for providing an agricultural measurement comprising:
 receiving a plurality of secondary signals from a plurality of secondary sensors;   providing the plurality of secondary signals to a neural network comprising a learning model that transforms the plurality of secondary signals into a value of a primary signal;   transmitting the primary signal from the trained learning model to a device having a display.   
     
     
         2 . The method according to  claim 1 , wherein training the learning model further comprises obtaining a measured value of the primary signal from at least one primary sensor, and modifying the learning model based on the measured value of the primary signal. 
     
     
         3 . The method according to  claim 2 , wherein the neural network is a recurrent neural network that is used to create the learning model that models a relationship between the plurality of secondary signals and the measured value of the primary signal. 
     
     
         4 . The method according to  claim 1 , wherein training of the learning model comprises using secondary signals from at least five secondary sensors. 
     
     
         5 . The method according to  claim 1 , wherein the value of the primary signal is not a value measured by the plurality of secondary sensors. 
     
     
         6 . The method according to  claim 1 , wherein the primary signal replaces one of the plurality of secondary signals. 
     
     
         7 . The method according to  claim 1  further comprising:
 obtaining at least one historical environmental data value; and 
 predicting a future value of the primary signal based on the at least one historical environmental data value. 
 
     
     
         8 . The method according to  claim 7 , wherein the at least one historical environmental data value is weather data. 
     
     
         9 . The method according to  claim 8 , wherein the plurality of secondary sensors is used to measure at least one of soil volumetric moisture, soil tension, soil temperature, air humidity, air temperature, and barometric pressure, and
 the primary signal is a value of soil matric potential.   
     
     
         10 . A system for providing a synthetic sensor for agricultural measurements comprising:
 a plurality of secondary sensors configured to output a plurality of secondary signals;   a neural network comprising a learning model that transforms the plurality of secondary signals into a value of a primary signal; and   at least one device having a display that is configured to output the primary signal that is a transformation of the plurality of second signals from the plurality of secondary sensors from the trained learning model.   
     
     
         11 . The system of  claim 10 , wherein the value of the primary signal is not a value measured by the plurality of secondary sensors. 
     
     
         12 . The system of  claim 10 , wherein the primary signal replaces one of the plurality of secondary signals.