A new genetic algorithm for polygonal approximation.

Area of Science:

• Computer Vision
• Computational Geometry
• Artificial Intelligence

Background:

• Digital curve representation is crucial for shape analysis and pattern recognition.
• Simplifying complex curves into dominant feature points preserves essential shape information.
• Existing approximation methods may suffer from premature convergence and suboptimal solutions.

Purpose of the Study:

• To develop an efficient algorithm for approximating closed digital curves using dominant feature points.
• To leverage genetic algorithms (GAs) for robust contour simplification.
• To enhance GA performance by addressing premature convergence.

Main Methods:

• An approach based on genetic algorithms (GAs) is employed for curve approximation.
• Chromosomes represent approximating polygons using binary strings, where '1' indicates a dominant point.
• The algorithm incorporates enhanced selection and mutation phases to prevent premature convergence.

Main Results:

• The proposed GA method demonstrates superior performance in approximating digital curves.
• Experimental results show a lower error norm compared to original curves.
• The method effectively identifies dominant points, preserving crucial contour information.

Conclusions:

• The enhanced genetic algorithm provides an efficient and accurate solution for digital curve approximation.
• This approach offers a significant improvement over existing methods for contour simplification.
• The identified dominant points accurately represent the essential features of the original digital curves.