algorithm - clustering a set of rectangles -
i have set of rectangles need cluster together, based on euclidean distance between them.the situation explained in attached image. .
one possible approach take center of each rectangle , cluster center points using k means (distance function euclidean distance in xy plane). however, know if there other approach problem, not approximate rectangle it's central point, takes actual shape of rectangle consideration.
have @ algorithms such dbscan , optics can used arbitrary data types long can define distance between them (such minimum rectangle-to-rectangle distance).
k-means not good, designed point data squared euclidean distance (= sum of squares, within-cluster-variance).
Comments
Post a Comment