What is it about?

It is important to be able to find and record Japanese knotweed over wide areas of often inaccessible land. Using standard approaches such as ecological survey, this can be very expensive. We used cost-effective aerial photograph imagery and an object-based image analysis algorithm to see if Japanese knotweed could be accurately located and mapped automatically.

Featured Image

Why is it important?

Though our object-based image analysis algorithm performed well in the areas where we manually mapped Japanese knotweed by ecological survey (i.e., ground-truthed), we were not confident about the accuracy of the algorithm elsewhere. This was because of the risk of false positives and negatives - both of which could have serious legal implications in the UK.

Perspectives

This journal article was developed significantly as part of a KESS MRes studies in the Geography Department of Swansea University. This further research highlighted how object-based image analysis accuracy for detection of invasive plants can often be limited by the underlying quality of data, rather than the algorithm in-itself. Therefore, such approaches are often not suitable for automated mapping of invasive plants over wide areas. This was my first peer-reviewed journal article and led to me obtaining a Knowledge Economy Skills Scholarship (KESS) PhD in Swansea University's Department of Biosciences.

Dr Daniel Jones
Swansea University

Read the Original

This page is a summary of: Object-Based Image Analysis for Detection of Japanese Knotweed s.l. taxa (Polygonaceae) in Wales (UK), Remote Sensing, February 2011, MDPI AG,
DOI: 10.3390/rs3020319.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page