search

vortexntnu / vortex-cv

A repository for a set of perception and robotic vision modules, developed by students in Vortex NTNU for use in the AUV and ASV software stacks.
MIT License
4 stars 3 forks source link

Clustering/segmentation of RGB data (Feature Detection) #38

Open theBadMusician opened 2 years ago

theBadMusician commented 2 years ago

Time estimate: 10 hours
Deadline: N/A

Description of task: It would be beneficial to apply a type of segmentation on the RGB data for easier HSV filtering afterwards. Clustering methods that should be looked into: K-means, superpixels, hierarchical (agglomerative and divisive).

theBadMusician commented 2 years ago

A recommendation from Stahl was to try out Directional Anisotropic-Diffusion filtering as a segmentation method. Mentioned that usual K-means clustering should work pretty well as well.