Problem Statement

We have developed a new marking technique for mark‑recapture studies of the mountain pine beetle. In traditional (typical) mark‑recapture studies insects are collected and marked using fluorescent powders by shaking them in a jar to coat them before releasing them en-masse from a central release point. Obviously, dispersal resulting from this type of mass release is very unnatural.

Our technique involves by coating trees in paper, which emerging insects would chew through as they emerge and thereby become marked by paper dust. This paper dust fluoresces under black light. In this way beetles would become marked in a much more natural way and would release themselves for later recapture.

A former student on this project imaged all of the beetles emerging from papered trees and from control trees that were not papered under a light microscope Illuminated with black light.

The workshop project challenge will be to conduct batch image analysis of a large number of images in order to automate classification of beetles as marked or unmarked. All available data will be made available for this project.

Once the images have been analyzed there is a classification problem and the data could be divided into training and validation subsets to demonstrate classification skill of an algorithm chosen by the team.

Skills

• Image Analysis
• Machine Learning