Tracking Temporal Changes in Vegetative Growth Patterns

It is a challenge to track vegetative growth over time and to re-identify a moving vegetation object between collection intervals. In this paper we propose a set of methods to re-identify a patch, a procedure which will provide insight into the growth patterns of the vegetation. After describing our model, we propose two methods for re-identification, a centroid approach based on distance, and using a patch’s area to calculate the maximum intersection. We make a non-split and non-merge assumption for each of these methods and apply the techniques on sample data set collected in the Cosumnes River floodplain of an invasive weed Lepidium latifolium over a three year period. Based on the results, we can determine how to combine, and refine, the techniques to yield the most probable matches. We also discuss query optimization and index structure to accelerate the re-identification process for large scale problems. We believe that our methods will facilitate applications on moving vegetation objects.