← zurück
Blog
Writing about things I build, break, and occasionally understand.
-
From cones to track — EMA tracking, KNN graphs, and DFS path search
How a list of detected cones becomes a pair of track boundaries — exponential smoothing, K-nearest-neighbor graph construction, depth-first path search with geometric pruning, conflict resolution, and center line computation.
-
Voxel-based cone detection from point clouds
A 3D perception pipeline for detecting traffic cones from LiDAR data — voxel downsampling, RANSAC ground removal, Euclidean clustering, and geometric classification using physical cone dimensions.