Detection of Venation Intersections in Bee Wings

Image processing, Image Detection

Project

The goal of this project was to find an algorithms pipeline that would get the venation intersection detected. Microscopic images and some ground truth were given, but purposely not enough not for deep learning to be used.

Here are a sample of given pictures with their ground truth

Chosen Solution

• Sato filter to highlight the venation
• Enhance and select highlighted venation with morphology
• Morphological reconstruction
• Skeletonize the result to ease corner detection

Results at each step of the pipeline

• Corner detection with Harris corner detector
• Postprocessing to remove points too close from eachother (avoid duplicates)

Result of raw corner detection and after post processing

Final results for the 5 pictures with ground truth. The calculated average F-Score is 0.809