A Novel Approach to Shape Recognition Using the Shape Outline

作者：Lee Sang Hee^*; Bardunias Paul; Su Nan Yao

来源：Journal of the Korean Physical Society, 2010, 56(3): 1016-1019.

DOI：10.3938/jkps.56.1016

• Branch length similarity (BLS) entropy
• Branching network
• Object shape
• Correlation coefficient

We have developed a new approach to shape recognition that uses the outline of an object and is based on a novel "branch length similarity" (BLS) entropy. The BLS entropy is defined on a simple branching network consisting of a single node and branches. The simple pattern is referred to as a "unit branching network" (UBN). The approach involves obtaining BLS entropy profiles from UBNs that are created by joining each pixel in the outline of a shape with every other pixel in the shape's border. To test whether they can be used effectively to recognize an object's shape, we obtained BLS entropy profiles from the shapes of twenty battle tanks depicted with 460 x 350 pixel resolution and from identical shapes depicted at lower resolution. The resized shapes were generated by decreasing the resolution of the original shapes to 60% in 10% increments. Each BLS entropy profile calculated from the original shape set was compared with all BLS entropy profiles obtained from the resized shapes by using correlation coefficients. The BLS entropy profile for a given shape was more similar to identical shapes depicted at different resolutions than to different shapes. Therefore, this process was able to successfully discriminate between the shapes tested regardless of resolution, and the proposed approach can be a useful tool for recognizing an object's shape.

• 出版日期2010-3