Descriptive Statistics Parameters of Normalized Edge-to-Centroid Distances as a Shape Feature that is Rotation and Scale Invariant

• DOI: 10.1145/3568231.3568247
• Publication Date: 2023-01-14T07:18:36Z
• Authors: Hurriyatul Fitriyah; Nur Afny Catur An...; Mochammad Hannats...; Rakhmadhany Prima...; Rizal Maulana
• Keywords: Rectangle; Centroid; Kurtosis

In Computer Vision, Shape Feature is one of the important features besides color and texture. A Geometric type Features uses parameters from Region Interest as input such Area, bounding rectangle's width length, Axis Length, Centroid, Perimeter. The Rectangularity, Eccentricity, Elongation rely on rectangle which invariant to rotation. Circularity rotation scale but only captures general information object's area. This research proposes a basic that invariant. First, it finds centroid using Moment equation, then edges Sobel operator calculates distances between them. distance Euclidian Distance values are normalized Min-Max Normalization maintain their value due variations. Descriptive Statistics Mean, Median, Standard Deviation, Skewness, Kurtosis used parameter extracted distances. proposed shape feature tested in common planar shapes with different sizes rotations circle, oval, square, rectangle, triangle, pentagon shows good grouping. can cluster Silhouette Index 0.578.