Team
- Faculty: Samarth Swarup (UVA), Abhijin Adiga (UVA)
- Undergraduate students: Andrew Ma (UVA), Christopher Goodhart (UVA)
- Graduate student: Supriya Savalkar (WSU)
- Other collaborators: Sid Chaudhary (NASA Impacts Centers, UA)
Objective
- To develop a deep learning method that can accurately determine the age of orchards from satellite imagery. This method would enable data collection from these orchards on a broad scale. Such a data collection method would eliminate the need to acquire the data for each orchard individually, which is expensive, time-consuming, and inaccurate. Orchard age is critical for informing decisions about resource allocation (including water), especially during times of drought, because orchard age is correlated with the value of the orchard.
Approach
We first recreated and slightly modified an existing method to determine orchard age which does not rely on deep learning. This method analyzed a time series of the average Normalized Difference Vegetation Index (NDVI) for each orchard and employed three rules to identify a planting year in the time series. We then used the AlexNet deep learning architecture to classify orchards as having a trellis or not. Because orchards with a trellis are generally younger, this is a first step towards a future deep learning method for predicting orchard age. models.
Related Publications
Sorry, no publications matched your criteria.
