Developing a low-cost digital crop monitoring system
Supervisors: Warren Jin (CSIRO), Rose Broderick (CSIRO) and Dani Way (ANU).
In farming operations, farmers (or their consultants) are keen to know very detailed information about their crops and land, such as the location of damaged areas from the frost event last night, or when and where to irrigate in upcoming days. Economic losses due to crop damage caused by frost in Australia are estimated to be many hundreds of millions of dollars (An-Vo et al., 2018; Fletcher et al., 2019; March et al., 2015). Overuse and poor irrigation practices may waste precious water, energy and money, reduce quality or yield, and can lead to leaching or run-off of nutrients that can pollute groundwater, rivers and estuaries.
A low-cost crop monitoring system is valuable to agriculture in Australia, however it is often expensive to collect these kinds of on-farm detail. On the one hand, thermal images from drones, helicopters and satellites can cover large areas, but normally with low temporal granularity and/or low spatial resolution. On the other hand, in-situ sensors can collect continuous time-series observations, but are too costly to cover large areas in this way.
This PhD project will develop fusion techniques to combine low frequency image observation with high-frequency point-specific time series to provide low-cost high-frequency information about crops. Some intelligent computation techniques will be explored, such as low-rank approximation, local kriging, spatially-varying coefficient modelling, and periodic modelling, the potential of which has been demonstrated by CSIRO (Shao et al. 2020, Bakar et al. 2016, Jin et al 2021). These techniques will be applied to monitoring frost damage and crop irrigation scheduling in case-study regions.