CSS Menu Buttons Css3Menu.com

Club Tracking Camera Mono Club Tracking Camera Color Ball spin camera VisTrak systems Stereo Cameras LX launch monitor Camera Installation Installation and setup

 

Please note that all GSA Golf systems that measure ball spin (Quantum systems) are currently in development.

Expected release date: 2nd quarter 2012

Note that the only difference between the regular VisTrak systems and the Quantum systems

is the ability to measure ball spin - that just requires an additional $699 camera.

For those that want to purchase now, we recommend purchasing the regular VisTrak system

(i.e. the VisTrak IR VC or the VisTrak Base with the Vcam C )

and adding on the additional Spin Cam C camera ($699) later when the development is completed.

VisTrak Spin Cam C

Add-on to VisTrak Base that already has the Vcam C for ball spin rate

$ 699.00

The Spin Cam C is a hi-resolution camera that - when used in conjunction with the hi-res Vcam C camera -

can calculate spin rate from the two images.

The system functions for both left and right handed players

This camera is currently in development.

Expected release date: 2nd quarter 2021

 

Spin Cam C images

The above images are from the Bcam when mounted overhead.

Bcam overhead mounted images

Left: frame 2 of full shot with 3 wood - Right: frame 1 of ball on mat hitting position

Left: frame 2 of open faced 3 wood. Right: Composite of frames 1 and 2.

Measuring spin rate from an overhead camera system

A little math is required to deduce the spin rate from the circle arc measured from the rotating and shifting key point pattern on the ball.

Knowing the circle arc length (i.e. the vertical shift of the key point pattern on the ball as viewed from above) and the radius of the ball, the arc central angle can be calculated.

This arc central angle is the rotation in degrees (or radians) for a given time between the two frames (1 ms) and from this, the spin rate can be calculated.

 

VisTrak LX ball spin detection


 

Detecting ball spin from ceiling mounted cameras that look down on the ball, requires a different approach.

When looking down on the ball, back spin rotation will appear as image shift downwards and not as regular rotation as when the cameras are viewing the ball from the side.

As the ball is spherical, images of the dimple patterns are distorted as the ball rotates, so that direct dimple pattern matching of two images of the ball will not be possible.

The above method solves this problem by projecting the ball's 3D spherical surface onto a 2D UV map so that dimple spin patterns are not distorted during the rotation

and thus can be located and matched.

As we are only seeing a small arc of the rotation, camera frame timing must be very fast. Currently, a frame time of 0.2 ms or 200 micro seconds is being used.

Using two individual hi-resolution cameras that only grab 1 image of the ball each, we can set the frame timing to be as fast as we want.

At a 1 milli-second frame time, the dual camera system has the equivalent of a 1000 fps camera and at 0.2 ms, the camera system will have

the equivalent of a 5000 frames per second camera.

Not bad for a low cost system we think.

The Quantum C will undoubtedly be our best selling system by far and it's price -

compared to the competition that are in the $10,000 to $20,000 range - should ensure that it dominates the market.

Data captured: Ball spin and spin axis, ball speed, LA and path, club face, club speed and path with swing video playback .

 

Development of new ball spin detection of balls without markings for the Quantum C is underway.

 

The real image size of the ball is huge when using the Quantum C 48mm lenses.

Image matching

The above shows two images of a ball after dimple edge detection and conversion to a binary image using two separate frames.

i.e. the edge detection filter process is run separately for each frame.

The second ball image is of the ball after it has been rotated by 45 degrees

An image matching and rotation process of the two frames is run until a close match is found.

i.e. we start out at 0 degrees and go through all 360 degrees, noting the number of pixels in the images that match for every 1 degree of rotation and store this matching value into a table.

We then just search through the 360 element table for the closest match - which will then tell us the amount of rotation the ball has made within the 1 ms frame time.

From this, the spin rate is calculated.

Using this method, stage 3 - dimple center detection - will not be required.


September 9 11:00 am

Ball spin detection without ball markings

Stage 2 - gray scale dimple edge conversion to binary dimple pattern completed.

The conversion itself was actually very simple.

Far more time was spent creating a smother gray scale dimple edge image which is now a composite of 12 segments - 8 around the sides and 4 in the center.

Stage 3 - dimple center detection - is going to be quite a challenge but I'm confident I'll have this completed by the weekend.

Then it's off to the races : image matching to determine rotation and thus spin rate.

White paper quote:

Although the small depression parts on a golf ball surface, which are called dimples, may appear to be regular, they actually have slight irregularity in their sizes and arrangement

in order to improve aerodynamic characteristics for, for example, better stability of flight or longer flight distance.

This is true for any real golf balls, as far as the authors know, because completely regular dimples offer poor aerodynamic performance.

Quantum C - Ball spin detection without ball markings from ceiling mounted cameras

The above shows what the UV map of the dimple pattern looks like.

While it doesn't look like anything recognizable to the eye,

it is in fact a unique pattern that should be detected in a second image of the ball after it has moved or rotated a number of degrees.


September 17 5:30 pm

Quantum C - Ball spin detection without ball markings from ceiling mounted cameras

The "spherical ball surface to a 2D plane surface map" has now been completed with auto positioning and scaling.

i.e. the hemispherical grid is auto positioned and scaled over the ball image.

Next step is to extract the pixel gray scale levels from all the 590 points of the hemispherical grid onto the 2D surface map.

Once this has been completed ( this weekend no doubt ),

we can start with the pixel matching process of the two ball images that will determine the amount of rotation and rotation direction

and thus the spin rate and spin axis of the ball.

It has yet to be determined if the ball image should be run through the dimple edge detection filter (see below) and/or converted to a binary image or left as is with its gray scale levels.

BTW while most consider a monochrome image to be a "Black and White" image photo, this is not technically correct.

A true black and white image contains only 2 levels - black or white - i.e. a binary image - (above right image)

while an image that most consider to be back and white is actually a gray scale image (above left image).


Measuring spin rate from an overhead camera system using spin dot balls

A little math is required to deduce the spin rate from the circle arc measured from the rotating and shifting key point pattern on the ball.

Knowing the circle arc length (i.e. the vertical shift of the key point pattern on the ball as viewed from above) and the radius of the ball, the arc central angle can be calculated.

This arc central angle is the rotation in degrees (or radians) for a given time between the two frames (1 ms) and from this, the spin rate can be calculated.

VisTrak Quantum data capturing method

1. The VisTrak camera detects when a ball is placed in the launch zone area

2. When a ball is detected, the stereo cameras are triggered to grab frames of the ball at the launch position.

3. The VisTrak camera is then placed in full speed mode (590 fps) while constantly checking to see if the ball moves forward.

4. As soon as the ball moves, (strike frame) a trigger signal is sent (via a GPIO line) to the stereo cameras

5. The stereo cameras are programmed to delay any trigger signal by 5 milli-seconds and 5.1 ms

6. The Stereo cameras then detect the height the ball is off the ground and calculates the launch angle and the spin rate based on the ball's rotation within 0.1 ms

7. The ball path - left or right - is detected by the VisTrak camera as is the club speed, face angle and club path

8. The ball speed is then calculated from the distance the ball has traveled from the ball strike frame in the VisTrak camera and the ball's position in the stereo cameras.

Notes:

While the camera's GPIO trigger signal is very fast (just a few nano seconds), there can be up to a 1.7 ms delay (1/590 frame rate) for the VisTrak camera to detect ball movement.

At a speed of 200 mph - and in a worst case - the ball may have already traveled some 6 inches before movement had been detected.

This distance has to be taken into account when measuring how far the ball has traveled within the 5 ms time frame in order to accurately measure ball speed.

After all these operations, a final check is made to see if the data is valid and that the ball's movement extended outside the launch zone.

i.e. it wasn't just a re-positioning of the ball on the hitting mat or from an inadvertant ball contact from a club waggle (or the ball just fell off the tee).


 

All camera products Club Tracking Camera Ball spin camera VisTrak systems LX Launch monitors Enclosures All products Software Business Installation and setup