Object-based image analysis for monitoring plant invasions, can we use an open-source solution?

Introduction

This is a practical exercise testing possibilities of open-source solutions (FOSS) for object-based image analysis (OBIA) to monitor plant invasion using unoccupied aerial system (UAS, drone).

The material is accompanying a chapter Müllerová, J. et al. (2023). Vegetation mapping and monitoring by unoccupied aerial systems – current state and perspectives. In: Manfreda, S. et Eyal B.D. (eds). Unmanned Aerial Systems for Monitoring Soil, Vegetation, and Riverine Environments. Elsevier.

The material is meant for readers to run the workflow and detect invasion of giant hogweed on the UAS data themselves testing different FOSS solutions.

Data

• a subset of UAS-borne data (consumer camera) collected in Czech Republic during the flowering of a noxious invasive plant species giant hogweed (Heracleum mantegazzianum)

• training dataset

• eCognition rulebase (proprietary OBIA software)

• a script for SegOptim package implemented in R

Description

The use case represents a simple application of OBIA approach based on the SegOptim package implemented in R.

Four bands (RGBN) UAS image subset are available, capturing the central area of a heavily invaded location (CZ) by giant hogweed (Heracleum mantegazzianum). Thanks to the proper image timing, the invasive species is clearly observable as white objects (in RGB) representing the various stage of the blossom. Considering the complex shape of the flower heads, detection based on image segmentation outperforms pixel-based classification (Müllerová et al., 2017). Simple segmentation of input imagery is performed (for simplicity only the spectral bands are considered both for segmentation and feature space definition, however additional features such as vegetation indices or textural measures may be included), followed by supervised classification using training data. Finally, a visual comparison of result detection both from proprietary (eCognition) and open-source (SegOptim) implementation is provided, confirming comparable results.

Based on #github("joaofgoncalves/SegOptim")

References

Gonçalves, J., Pôças, I., Marcos, B., Mücher, C. A., & Honrado, J. P. (2019). SegOptim—A new R package for optimizing object-based image analyses of high-spatial resolution remotely-sensed data. International Journal of Applied Earth Observation and Geoinformation, 76, 218-230.

Müllerová, J., Brůna, J., Bartaloš, T., Dvořák, P., Vítková, M.  & Pyšek, P. (2017b). Timing Is Important: Unmanned Aircraft vs. Satellite Imagery in Plant Invasion Monitoring. Frontiers in Plant Science 8:1–13.

Accompanying material for

Müllerová, J. et al. (2023). Vegetation mapping and monitoring by unoccupied aerial systems – current state and perspectives. In: Manfreda, S. et Eyal B.D. (eds). Unmanned Aerial Systems for Monitoring Soil, Vegetation, and Riverine Environments. Elsevier