Submitted by Maggi Kelly on
Object detection in high resolution imagery is an important research arena across many domains. In environmental applications, the mapping individual tree crowns or gaps in a forest can provide useful data on the start-of-season, end-of-season, plant phenology, growth and yield. There are several protocols to do this; one of the most promising recent development are CNNs. Convolutional Neural Networks are multi-layered models (neural network) that learn how to identify and recognize objects in an image. With the proliferation of high-resolution imagery from drones, finding an accurate and straightforward method to extract individual tree metrics is important. This project will develop and test numerous cutting-edge methods for object detection that rely on CNNs. The SPUR student will work the UC Agriculture and Natural Resource Statewide Program in Informatics and GIS (IGIS) personnel with drone imagery, GIS software, and Jupyter Notebooks to refine our existing protocols. and test transferability across forest and agricultural domains.
Learn Deep Learning protocol and evaluate accuracy of alternative methods using a range of high-resolution drone imagery.
Interact with PI Kelly and IGIS staff.
Perform GIS and image processing tasks.
Experience with ArcGIS Pro; Strong Python programming skills; Knowledge of forestry