A smart classifier for extracting environmental data from digital image time-series: Applications for PhenoCam data in a tidal salt marsh. (October 2016)

• Record Type: Journal Article
• Authors: O'Connell, Jessica L.; Alber, Merryl
• Is Part Of: Environmental modelling & software. Volume 84(2016:Oct.)
• Page Start: 134
• Page End: 139
• Publication Date: 2016-10
• Subjects: Flooding -- Georgia Coastal Ecosystems LTER -- PhenoCam -- Phenology -- Salt marsh -- Tides -- Wetlands; BCC blue chromatic coordinate -- GCC green chromatic coordinate -- RCC red chromatic coordinate -- RGB red, green, blue color space -- ROI region of interest -- WFI weather filtering index; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2016.06.025
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: PhenoCams are part of a national network of automated digital cameras used to assess vegetation phenology transitions. Effectively analyzing PhenoCam time-series involves eliminating scenes with poor solar illumination or high cover of non-target objects such as water. We created a smart classifier to process images from the "GCESapelo" PhenoCam, which photographs a regularly-flooded salt marsh. The smart classifier, written in R, assigns pixels to target (vegetation) and non-target (water, shadows, fog and clouds) classes, allowing automated identification of optimal scenes for evaluating phenology. When compared to hand-classified validation images, the smart classifier identified scenes with optimal vegetation cover with 96% accuracy and other object classes with accuracies ranging from 86 to 100%. Accuracy for estimating object percent cover ranged from 74 to 100%. Pixel-classification with the smart classifier outperformed previous approaches (i.e. indices based on average color content within ROIs) and reduced variance in phenology index time-series. It can be readily adapted for other applications. Highlights: Automated digital image processing provides time-series for documenting ecological phenomena. We developed a smart classifier to detect and quantify vegetation and other object classes in Phenocam imagery. The smart classifier is more effective than previous approaches for filtering PhenoCam images. The smart classifier can be adapted for a wide range of applications.