•  
  •  
 

Abstract

Shape recognition is an essential task in machine vision applications. Many techniques have been adopted for shape recognition all of them distributed according to two directions of shape features belonging to boundary or region, in many categories of applications in addition to recognition like shape simplification, restoration classification and retrieving. This paper presents a proposed testing algorithm for shape recognition according to boundary or global base features where the approximated contour of the shape is derived by the Douglas Peucker line simplification approach. The initial diversity value starts with fixit value according to an undetermined number of iterations until reaching to target set of approximated points for testing, where in each iteration the diversity value increases gradually with the fixit interval. This algorithm evaluates the state, "Is there an optimal number of approximated points as features vector can be adopted that satisfy the maximum recognition rate.'' Before learning and testing stat the features vector derived after applied a distance matrix among the approximated set. The recognition process experiments on the MPEG-7 dataset when classified using Support Vector Machine (SVM) through different kernel functions. However, the experiment results show the proposed testing recognition achieved to high rate about 0.961 according to a specific number of approximated datasets when adopting a radial basis function (RBF) as kernel function based on Matlab testing environment.

Keywords

Douglas Peucker, Line simplification, Pattern recognition, Shape recognition, SVM

Subject Area

Computer Science

First Page

349

Last Page

360

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Receive Date

9-18-2023

Revise Date

2-25-2024

Accept Date

2-27-2024

Download

Share

COinS