Remote sensing of coarse woody debris in caribou habitat

The MSc thesis of Gus Lopes Queiroz is centered on the study of automated detection of coarse woody debris (CWD) on high-resolution aerial images of the boreal forest. A field campaign was performed during the summer of 2018 to physically obtain CWD measurements on forest restoration sites near Conklin, Alberta. Next, a machine learning solution to semi-automatically classify CWD on aerial images was performed with reliable accuracy. Rigorous tests on the precision of the method are currently being evaluated.

The following poster contains some of Gus' methods and preliminary results and was presented during the Boreal Ecosystem Recovery and Assessment (BERA) project third annual workshop on Thursday, November 22nd 2018 at the University of Alberta.

Click the image to view the poster larger!

Glossary

• CHM: canopy height model; dataset with height of objects on terrain.

• Commission: proportion of false positives.

• Completeness: 1 – Omission.

• Correctness: 1 – Commission.

• CWD: coarse woody debris; dead trees.

• Decay: stage of decomposition, ranges from 1 to 5.

• DSM: digital surface model; dataset with elevation of terrain.

• Image-object: segment of an image, representing an object.

• Log: downed CWD.

• Machine learning: artificial intelligence algorithm.

• Omission: proportion of false negatives.

• Orthophoto: geometrically corrected aerial photograph.

• Plot: sampling site.

• RF: random forest; machine learning technique.

• Seismic line: petroleum exploration cut line in a forest.

• Snag: standing CWD.

• Spatial attributes: size and shape characteristics of an object.

• Spectral attributes: image layer data; red, green, blue and near infrared characteristics of an object.

• Training samples: example dataset used to prepare the RF classifier for application.