Measuring phenology uncertainty with large scale image processing

One standard method to capture data for phenological studies is with digital cameras, taking periodic pictures of vegetation. The large volume of digital images introduces the opportunity to enrich these studies by incorporating big data techniques. The new challenges, then, are to efficiently proce...

Full description

Saved in:
Bibliographic Details
Published in:Ecological informatics 2020-09, Vol.59, p.101109, Article 101109
Main Authors: Alles, Guilherme Rezende, Comba, João L.D., Vincent, Jean-Marc, Nagai, Shin, Schnorr, Lucas Mello
Format: Article
Language:English
Subjects:
Computer Science
Distributed, Parallel, and Cluster Computing
Mathematical modeling
Modeling and Simulation
Parallel workflow
Phenological visualization
Phenology analysis
Uncertainty quantification
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:One standard method to capture data for phenological studies is with digital cameras, taking periodic pictures of vegetation. The large volume of digital images introduces the opportunity to enrich these studies by incorporating big data techniques. The new challenges, then, are to efficiently process large datasets and produce insightful information by controlling noise and variability. On these grounds, the contributions of this paper are the following. (a) A histogram-based visualization for large scale phenological data. (b) Phenological metrics based on the HSV color space, that enhance such histogram-based visualization. (c) A mathematical model to tackle the natural variability and uncertainty of phenological images. (d) The implementation of a parallel workflow to process a large amount of collected data efficiently. We validate these contributions with datasets taken from the Phenological Eyes Network (PEN), demonstrating the effectiveness of our approach. The experiments presented here are reproducible with the provided companion material. •Improvements about the CPM visualization technique, using colors from the inputs to build the histograms representation.•New phenological metrics derived from the RGB and HSV color channels for the CPM-based visualization.•A mathematical model to tackle the natural variability and uncertainty of phenological cameras.•A scalable workflow combining contributions into a flexible, reproducible, script-based implementation.
ISSN:1574-9541
DOI:10.1016/j.ecoinf.2020.101109