Launch monitor from scratch

[golftechie]

Perhaps one day, mobile technology can catchup to perform that of photometric launch monitors.  I definitely wouldn't hold my breath, even at the highest frame rate say 240fps which is what an iPhone 6 can record, the individual frames exhibit too much motion blur.  This prevents both ball and golf ball dimples from being properly identified which is a prerequisite to both image segmentation and analyzing spin rates.

Some current solutions use pulsed IR which can freeze a frame on the order of nanoseconds which is needed to identify spin patterns on the ball.

[Lihu]

11 minutes ago, golftechie said:

Perhaps one day, mobile technology can catchup to perform that of photometric launch monitors.  I definitely wouldn't hold my breath, even at the highest frame rate say 240fps which is what an iPhone 6 can record, the individual frames exhibit too much motion blur.  This prevents both ball and golf ball dimples from being properly identified which is a prerequisite to both image segmentation and analyzing spin rates.

Some current solutions use pulsed IR which can freeze a frame on the order of nanoseconds which is needed to identify spin patterns on the ball.

 

It might be more possible to use existing technology, and make a cheap enough pulsed light technology to do this? Very fast frame transfer is memory bandwidth limited and will always be.

Radar is much faster at sampling.

Line sensing at 60,000 fps is possible, but that also requires a large amount of light or IR.

[Lihu]

These are relatively cheap 24GHz doppler radar sensors.

I'm thinking of using them for a 2D or 3D DIY LM type project.

I also have a few line sensors and other 3D photonics based sensors. Seeing what is the cheapest and easiest to do. Looks like the modules above will work nicely, since most people trust Radar to be accurate enough anyway.

[Lihu]

Hi All, this is a continuation of the "launch monitor from scratch" type of thing. Got some of the sensors used in professional launch monitors and am starting to play with them a bit.

Finally got some time to hook up one of the sensors using just the "Q" output for now. You need I and Q to determine direction.

Me moving the camera in to take a picture.

 

Keeping perfectly still.

 

This sensor is really sensitive, because I haven't even added any amplification yet. Planning on doing that and adding an ACD converter. Probably going to try a simple USB stereo codec. Once I get the signals into Sw it will be realtively easy to process using either SciPy, Octave or SAGE (open source Matlab type programs). When I run it on a Raspberry pi to make it more portable I'll likely use something lightweight like SciPy.

The code will likely be a multiple order high pass filter to eliminate anything less than 20m/s or something like that. I'll have to figure all that out as I go. 

For now the velocity is approximately

delta f= 2 * delta v / c * f0

Where f0= 24.1 GHz; c= speed of light in m/s

v ~= delta f / 160

Obviously, most of the calculations are not going to be simple fourier components. Probably going to use z-space to calculate "things".

 

Step by step.

[clarke20]

On 1/30/2018 at 5:25 PM, Lihu said:

Hi All, this is a continuation of the "launch monitor from scratch" type of thing. Got some of the sensors used in professional launch monitors and am starting to play with them a bit.

Finally got some time to hook up one of the sensors using just the "Q" output for now. You need I and Q to determine direction.

Me moving the camera in to take a picture.

 

Keeping perfectly still.

 

This sensor is really sensitive, because I haven't even added any amplification yet. Planning on doing that and adding an ACD converter. Probably going to try a simple USB stereo codec. Once I get the signals into Sw it will be realtively easy to process using either SciPy, Octave or SAGE (open source Matlab type programs). When I run it on a Raspberry pi to make it more portable I'll likely use something lightweight like SciPy.

The code will likely be a multiple order high pass filter to eliminate anything less than 20m/s or something like that. I'll have to figure all that out as I go. 

For now the velocity is approximately

delta f= 2 * delta v / c * f0

Where f0= 24.1 GHz; c= speed of light in m/s

v ~= delta f / 160

Obviously, most of the calculations are not going to be simple fourier components. Probably going to use z-space to calculate "things".

 

Step by step.

I'm curious did you get this working?? 

[02bluesuperroo]

On 6/10/2020 at 4:13 PM, clarke20 said:

I'm curious did you get this working?? 

I'm interested as well.  Maybe if a bunch of us band together with varying knowledge we could put together a parts list and some open source software for at least a rudimentary DIY solution.  There must be some people out on the forums who have (in a past life of course) worked at some of these launch monitor companies?  Nothing for-profit.... just for-fun!

[Crisgolf93]

The question that's left is who is crazy enough to embark on this challenging yet fun and interesting journey on making this happen? Sign me up!! who else is in??

[pganapathy]

I know the idea behind building a launch monitor from scratch made a lot of sense a few years ago.  However, with the advance in camera technology on smart phones, if you are building something like this, wouldn't just creating an app, like Shot Vision, make more sense today.  No need to break your head with building the device, just create a program or app for it.  Of course, this doesn't work for a doppler based system