Moment Invariants for 2D Flow Fields via Normalization in Detail.

Area of Science:

• Fluid dynamics
• Image recognition
• Pattern analysis

Background:

• Analyzing 2D flow data requires identifying meaningful structures.
• Current methods face challenges in defining similarity and pattern descriptors invariant to transformations like rotation and scaling, yet sensitive to deformations.

Purpose of the Study:

• To propose moment invariants as robust pattern descriptors for 2D flow fields.
• To develop a complete and independent set of descriptors invariant to translation, rotation, and scaling.
• To enable the distinction of flow features based on velocity profiles.

Main Methods:

• Utilizing moment invariants, a technique popular in image recognition, for describing 2D vector patterns.
• Applying moment normalization for translation, rotation, and scaling invariance.
• Developing pattern recognition algorithms using these invariant descriptors for flow analysis.

Main Results:

• Moment invariants provide a complete and independent set of descriptors for flow fields.
• The proposed method is invariant to translation, rotation, and scaling, while sensitive to deformations.
• The approach successfully distinguishes flow features with different velocity profiles.
• Demonstrated robust extension of theoretical results to discrete functions on real-world datasets.

Conclusions:

• Moment invariants offer a powerful tool for pattern recognition in 2D flow analysis.
• The developed descriptors provide a robust method for identifying and comparing flow structures.
• This approach enhances the understanding of fluid dynamics by enabling detailed feature analysis.