Automating image segmentation for vegetation monitoring

Plants represent a major component of the biosphere, influencing the flow of energy, carbon and water within the Earth system. Human-induced global environmental change is rapidly altering the form and function of vegetation from cells to canopies, with implications for the integrity of the Earth system. Therefore, vegetation monitoring is crucial, to quantify impacts, determine baselines and guide potential management options. Image capturing devices, such as satellites, drones, portable scanners and smartphones are widely used by plant scientists to monitor vegetation. However, processing these images to identify meaningful objects can be time and labour-intensive.

Image-processing tools such as OpenCV can be widely applied to many areas of vegetation science. For example, OpenCV has been used to extract digitised plant size data from drawings made in the field, to examine demographic trends of grassland species under experimentally manipulated rainfall conditions. Advances in artificial intelligence computer vision will lead to models that can be pretrained on large and diverse datasets, which will further reduce the user input and computing resources required for image processing. The use of such tools to speed up image-based vegetation monitoring could help to address many of the challenges associated with environmental change such as deforestation and drought.