Accelerated PIV (Particle Image Velocimetry) in C++ with Python Bindings

This package provides accelerated Particle Image Velocimetry (PIV) functionality implemented in C++ and CUDA with Python bindings.
PIV is a widely used technique for measuring velocity fields in fluid flows by tracking the motion of particles suspended in the flow.

The project is currently in progress and may be unstable :(

Dependencies

This project relies on the following dependencies:

• NVIDIA Cuda Toolkit
• OpenCV
• C++20
• Python3-dev package
• Linux (temporary)

Installation

You can compile and install the package using CMake and pip:

git clone --recurse-submodules https://github.com/AsphodelRem/accelerated_piv_2d.git
cd accelerated_piv_2d
cmake build
pip install .

Usage

import accelerated_piv_cpp

# Define PIV parameters
piv_params = accelerated_piv_cpp.PIVParameters(width=512, height=512, channels=3, window_size=32)

# Create an ImageContainer
image_paths = ["path/to/image1.jpg", "path/to/image2.jpg"]
image_container = accelerated_piv_cpp.ImageContainer(image_paths, piv_params)

# Perform PIV analysis
accelerated_piv_cpp.start_piv_2d(image_container, piv_params)

# Retrieve results
# TODO: Add instructions for retrieving results once available

TO-DO

#	Task	Done
1	Replace OpenCV with another library for image and video processing	✗
2	Implement iterative algorithms for filtering erroneous vectors	✗
3	Add export of results to .csv	✗
4	Add benchmarks	✗
5	Add a wrapper for the results storage class	✗

API Reference

PIVParameters

Class for defining PIV analysis parameters.

Constructor

PIVParameters(width: int, height: int, channels: int, window_size: int, overlap: int = 0,
               filter_type: FilterType = FilterType.kNoFilter, filter_parameter: float = 0.0,
               interpolation_type: InterpolationType = InterpolationType.kGaussian,
               correction_type: CorrectionType = CorrectionType.kNoCorrection,
               correction_parameter: int = 0, save_on_disk: bool = False,
               capacity: int = 1200)

Attributes

• width: Width of the images.
• height: Height of the images.
• channels: Number of color channels.
• window_size: Size of the interrogation window.
• overlap: Percentage of overlap between interrogation windows.
• filter_type: Type of filter to apply.
• filter_parameter: Filter parameter (e.g., cutoff frequency).
• interpolation_type: Type of interpolation method.
• correction_type: Type of correction method.
• correction_parameter: Correction parameter (depends on correction type).
• save_on_disk: Whether to save intermediate results on disk.
• capacity: Capacity of the data container.

ImageContainer

The ImageContainer serves as a data container for processing images. It is utilized for loading and pre-processing a set of images intended for PIV analysis.

Constructor

ImageContainer(image_paths: List[str], piv_parameters: PIVParameters)

VideoContainer

The VideoContainer acts as a data container for processing video files. It is employed for loading and pre-processing video footage intended for PIV analysis.

Constructor

VideoContainer(video_path: str, piv_parameters: PIVParameters)

start_piv_2d

Function to start 2D PIV analysis.

start_piv_2d(container: IDataContainer, parameters: PIVParameters)

License

This project is licensed under the MIT License - see the LICENSE file for details.