Significant speed and memory improvements in PIVlab!

Mehdi (Manager Mechanical & Aerospace Engineering) from MATHWORKS (the company behind Matlab) approached me a while ago and we discussed how they can support the development of PIVlab. PIVlab seems to be have some importance a relevance to MATHWORKS as it is a pretty popular 3rd party toolbox for Matlab. Together, we identified potential areas in PIVlab that could be improved and MATHWORKS then initiated an internal proposal for programming support. They selected the dutch company Vortech.nl for this project, and Maarten then started to optimize the most important part of PIVlab "PIV_fftmulti.m". He significantly improved the speed of calculations (factor 3), and lowered the memory consumption enormously. This especially helps when processing large images with fine grids, and also makes it possible to benefit from parallel processing of large images. This hasn't been possible before, because at some point RAM was full and the hard disk was used to store temporary data - making everything very slow.

I really enjoyed to see Maarten, a professional coder, doing his work. It was very impressive to follow his work and his "tricks" to improve the code! Big thanks to MATHWORKS and especially Mehdi for making this possible! And of course thanks to Maarten from Vortech for his great work! Have a look at the magic of professional programming here: https://github.com/Shrediquette/PIVlab/commits?author=mkbosmans

The improved code (along with some other enhancements) was released today: https://github.com/Shrediquette/PIVlab/releases

Furthermore, I finished designing my first "own" camera for PIVlab, the "OPTOcam 2/80", with 2 megapixels and a double image rate of 80 Hz (8-bit mode) respectively 40 Hz (12-bit mode). It is equipped with a remote focus control and has a min. interframe time of 61 µs. Of course it is implemented in PIVlab's image acquisition module already.

Details about all supported camera models are available here: https://docs.google.com/spreadsheets/d/1Bdued3OBxUxvRdtU2z3-VYYIoyNvdTBEEUwfodnnP-Q


Location: Lörrach, Deutschland

Popular posts from this blog

PIVlab 3.00 released

Image
  I have just released PIVlab 3.00. There are several changes in the way the graphics are drawn, the way the data is exported and the way it interacts with masks / calibration and region of interest: New graphics: Transparency, masks are transparent, only ROI is opaque Colorbar is now outside the plot for much improved readability Export image data as .mp4 and as .png image Completely new mask handling, allowing much easier mask creation / modification. Mask stay editable all the time. Much easier to create holes in masks Mask operations like shrink/grow/simplify/subdivide/optimize Automatic mask generator, capable of automatic masking of bright areas / dark aread / low-contrast areas. Possibility to add more automatic mask generators in the future Mask import (of user generated masks) heavily improved Matlab Online fully supported New line extraction and area extraction interface: More straightforward and faster. Better interaction Natural file name sorting when importing image fi...
Read more

Comparison of OpenPIV and PIVlab

Image
Since quite a while, I benefit from discussions with the main developer of OpenPIV (Prof. Alex Liberzon  from Tel Aviv University) and other people that contribute to OpenPIV. It is almost a bit like a cooperation, as we share thoughts on different topics, and Alex is really into practical and theoretical PIV application. I think we have a very friendly rivalry in developing useful PIV software. The results are available here:  https://github.com/alexlib/openpiv_pivlab_von_Karman_data You will notice that they are very noisy. But we intentionally decided to make the analysis very challenging for our software. The final interrogation area is e.g. only 6*6 pixels, and no smoothing is allowed. I don't really see a difference between OpenPIV and PIVlab, which is probably a good sign. It might be interesting to include commercial software too in this comparison in the future. By the way: Happy New Year to all PIVlab users (and of course all OpenPIV users too ;-D). The year 2020 has...
Read more

High-speed low-cost PIV with the pulsed laserdiode: Examples

Image
Recently, I asked Prof. de Payrebrune from the University Kaiserslautern / Computational Physics in Engineering if they could lend me their Chronos 1.4 low-cost high-speed camera for some tests. She agreed to support the development of PIVlab which is very cool and kind of her!! Now I did some tests with the camera (captures data at 1.3 megapixels at 1000 Hz) and my new LD-PS pulsed laser diode. The LD-PS has a built-in synchronizer, so it triggers the Chronos and can also do frame-straddling. The Chronos has absolutely brilliant trigger characteristics, as you have complete control over the exposure timing. Now I am not limited to an interframe-time of 1/1000s anymore, but can go down to 10 µs.  When capturing data at 1000 fps, the duty cycle of the illumination becomes quite high. So I added two little fans to the LD-PS housing. Now a duty cycle of up to 50% can be reached without the laser or the driver becoming too hot. The awesome Chronos high-speed camera from krontech.ca La...
Read more