Shape Recognition Methods Used for Complex Polygonal Shape Recognition

Shape Recognition Methods Used for Complex Polygonal Shape Recognition

Copyright: © 2014 |Pages: 19
DOI: 10.4018/978-1-4666-4896-8.ch008
OnDemand PDF Download:
No Current Special Offers


Two new algorithms, optimal and sub-optimal solution algorithm, for shape recognition using geometric calculations are proposed. They are mode-based, which means that shape from the data base which is considered as a model is compared with another shape extracted from the sequence of images, such as, for example, a moving object. The proposed algorithms are efficient and tolerate severe noise. They have the ability to identify the close match between the noisy polygon that has a significantly greater number of sides and the assigned polygon. They work for convex and concave polygons equally well. These algorithms are invariant under translation, rotation, change of scale, and are reasonably easy to compute. The proposed criterion is a metric. The polygonal shapes are compared based on their areas and gravity centers. One polygon is placed over the other one so that one polygon has a fixed center of area (gravity center). The area of the intersection of these two polygons is calculated after one of the polygons is rotated one degree at time. The angular position with the best match is recorded.
Chapter Preview

1. Introduction

A very important domain in video surveillance systems is object identification. This is a phase that comes after the moving object detection and moving object tracking blocks. After designing the intelligent software capable of extraction of moving objects from video sequence without any aside intervention, the next problem to be solved is how to train the system to identify the detected object and report its identification automatically. That is why shape recognition has its theoretical and practical importance in image processing and computer vision.

Many authors have done research in the shape resemblance domain (Schwartz, 1984), (Shapiro, 1982), (O’Rourke, 1985), (Hong, 1988), (Cox, 1989), (Avis, 1983), but these methods are position, range scale and orientation dependent. They are not metrics and are also complicated from computational point of view. Some of these methods (Schwartz, 1984) and (Cox, 1989), apply only to convex polygons and that additionally restricts the proposed algorithms. Huttenlocher and Kedem in (Huttenlocher, 1990) develop the method that computes a distance between two shapes based on the Hausdorff metric. Their metric compares polygonal shapes independently of affined transformation. Some of the proposed algorithms as in (Wolson, 1987) use the curvature function as a shape signature. The methods presented in (Schwartz, 1984), (Shwartz, 1987) and (Arkin, 1991) have a similar approach. They are all based on convolution.

The authors in (Arkin, 1991) represent the simple polygon boundary with the turning function that measures the angle of the counterclockwise tangent as a function of the arc-length, measured from some reference point on polygon’s boundary. The turning function represents the angle that the tangent at the reference point makes with some reference orientation associated with the polygon (such as the x-axis). It keeps track of the turning that takes place, increasing with left-hand turns and decreasing with right-hand turns. The metric proposed in (Arkin, 1991) turns out to be very problematic for noisy and destroyed shapes.

Complete Chapter List

Search this Book: