CSS Menu Buttons Css3Menu.com

Stereo Parallel camera setup Stereo Converging camera setup Camera calibration Carry adjustments Putting and chipping IBS image processing Technical help

CX Converging Stereo Camera Setup

! Note: this setup procedure is only for the individual CX stereo camera configuration !

CX Surround systems require no calibration

CX Stereo ball tracking

Camera mounting

The two cameras ( V and H cams) are mounted around 4 to 6 ft ahead of the player's normal ball striking position and aproximately 8ft apart - centered.

Note that the cameras are mounted sideways in order to increase the left and right FOV.

Aiming the cameras to the center line

Place a ball or any clearly visible white object on the floor between the two cameras.

Starting with camera 1, select video stream mode and aim the camera so that the white ball appears exactly in the center of the image.

There are vertical and horizontal center lines visible in the image to help with this.

Do the same for camera 2.

CX Surround's case width increased to 96" (8ft).

The CX Surround now only uses a converging stereoscopic camera setup.

In order to ensure a +/- 25 degree ball path detection range, the casing's width has had to be increased from 4.5ft to 8ft.

BTW: Foresights Sports GC Hawk is around 7.6ft in width we have been told so we're about the same here.

A big advantage over the CX Stereo (using individual cameras) is that we can pre-calibrate the cameras if they are fixed in a single case.

CX Surround 3D stereoscopic ball and club tracking

The CX Surround has all cameras and lights built into one long (8ft) case

CX2 Surround : $ 3,299

2 camera ball tracking - calculated club and ball spin data

CX3 Surround : $ 4,299

3 camera ball and measured club tracking - calculated spin and spin axis

CX4 Surround : $ 5,299

4 camera ball and club tracking - measured spin and spin axis

Note: the CX4 Surround features a smaller separate case

containing both the Ccam club tracking and Bcam ball spin cameras

mounted directly over the player's hitting position

Calibration

Note that the CX Surround's camera are pre-calibrated

so this calibration procedure is only required for the CX Stereo which is using individual mounted cameras

Mounting the cameras sideways means the direction of play is then in the vertical direction

i.e. from down to up in the camera images instead of from right to left.

You will thus have to move the "Ball Launch position" to the bottom center of the camera image.

The above shows the converging camera calibration table that is automatically loaded when "Converging cameras" option is selected

Each row in the table consists of 5 values that are automatically calculated:

Note that if both cameras are perfectly aimed at the same center line then disparity at the converging point at floor level will be zero.

This will probably prove to be difficult to achieve but at least the disparity number at floor level should be a low number.

The above table shows a disparity (Disp) of 24 at floor level.

.

A graph plot of disparity to golf ball height shows a non-linear exponential curve.


Calibration

Note: please use Control Panel version 8.8.0.5 or greater for this procedure.

Calibration procedure

Should re-calibration be required then the following calibration procedure should be made:

Calibration setup

In order to calibrate, we need to raise a 20cm white strip mounted on a black card from the floor up to around 170cm in 10cm steps on the center line.

You can use any method you like to raise the strip off the ground in the 10cm steps.

Attaching the strip to a metric tape measure that is suspended from the center line of the enclosure and extended to the floor is one method.

The above image shows a white strip card directly attached to a tape measure that can simply be moved up or down to the desired height.

You could also use an extend-able camera tripod.

Note that the calibration table consists of 17 row entries at set 10cm distances from 0 cm to 170 cm.

i.e. 0, 10, 20 ... to 170 cm

17 Calibration steps

  • To start the calibration entry process, click the "Show calibration table" option in the one of the two
  • camera windows (stereo left or stereo right) and press the * key.
  • This allows you to make entries into the calibration table.
  • Using the up down keyboard arrow keys (ensure that no other parameter setting is active)
  • select the first table entry
  • Tthis will be height 0 i.e. floor level
  • Step 1: Place the card at the floor level directly under the cameras center point
  • i.e. between the two cameras.
  • Ensure that the white strip on the card is pointing directly forward toward the screen
  • Step 2: The user triggers both cameras (by pressing the Enter key) to capture images from both cameras
  • so that a basic height, a disparity value, a height and a scaling factor for the known length of the white strip
  • is calculated.
  • Note that the white tape strip represents a ball trace with 2 3D points at either end of the strip
  • so ensure that system detects both ends of the tape strip in both cameras.
  • Check both camera frames with every step to verify that both ends of the white strip are detected
  • as it is quite possible that the tape measure or parts of the tripod may be being picked up inadvertently.
  • Use the blue border lines to confine the valid FOV so that only the white strip is detected.
  • Step 3. The user preses the "Ins" key to store the values.
  • --
  • Step 4. Now move the card up to the first or next 10 cm marking on the tape - using either the camera tripod or lab stand -
  • and select the next row height step (arrow up key) in the calibration table.
  • i.e. if you started at 0 cm height, then next entry is the 10 cm high entry
  • Step 5: The user repeats these steps 12 to 17 times for a total height of 120 cm to 170m or as high as possible before
  • the card's tape strip goes out of the FOV of the cameras..
  • When raising the black card with white strip on it, make sure to keep the card and strip level.
  • With every step, ensure that both ends of the white strip on the card are detected in both cameras
  • (i.e. the green cross hairs are exactly at both ends of the white strip in cameras 1 and 2)
  • Also ensure that the tape measure or any part of the Tripod or Lab Stand is not being picked up.
  • Either use the blue valid FOV borders block these out of the valid FOV of the cameras or cover the stands with a black material.

The above image shows the white strip connected to the tape measure with the slider.

Using the calibration slider on the tape measure, a complete calibration can be done in 10 to 15 minutes.

When testing, set both RAW Path and RAW Launch angle to ON

Note that the above image is from the parallel setup page where the strip orientation is from right to left.

The converging camera strip oprientation is vertical i.e from down to up.

You can view the complete calibration table by selecting the "Show calibration table" option.

Virtual checker board

Instead of using a large cumbersome checker board to calibrate the stereo cameras,

this method uses a virtual checker board that automatically scales and positions itself depending on the position and image size of the white tube

Note that the CX Surround uses super low distortion lenses so lens bend doesn't have to be taken into account

  • Note that even though the calibration is only made in 10cm steps, the calibration
  • values are weighted for every ball position in 1 cm steps in 3D space.
  • Example:
  • The real ball height off the ground is say 105 cm with a stereo camera disparity value of 296
  • The calibration table has height correction factor values of 0.28 & 0.14 for 100 cm and
  • 110 cm based on the disparities 290 & 302.
  • The disparity percentage difference is then 296 / (290 + 302) = 0.5.
  • So the real height correction is actually halfway between 0.28 and 0.14
  • i.e. 0.28 + 0.14 x 0.5 = 0.21
  • Tests with a laser distance measurement instrument confirm that the weighted correction value
  • is well within the 1cm tolerance.
  • Disparity
  • Disparity ( the offset between 2 points in 2 images from 2 cameras) is used to calculate the
  • distance the point or object is away from the camera.
  • It is used in the formula D = bf / d
  • where D = Distance of point in real world = (base offset) * (focal length of camera) / (disparity)
  • Scaling factors
  • The calibration table also consists of a scaling factor for each height.
  • It is calculated by taking the real length of the white tube (a constant 20cm) and dividing it by the
  • length of the tube measured in pixels
  • (which - of course - varies depending on the distance the tube or object is away from the camera)
  • The scaling factor is used to measure the distance the ball deviates from the center line left and right (ball path)
  • and the distance the ball has traveled forward.
  • Scaling factors convert distances measured in pixels to real world distances measured in cm or inches.
  • CX Surround - Raw path and LA vs Fixed point path and LA.

Although the real distance (d) the ball travels (and thus the ball speed) within the camera exposure time is always measured using RAW LA and Path

(i.e. as measured from the both ends of the ball trace)

it will be more accurate to measure LA and path using fixed launch positions for chipping as by the time the ball is in the FOV of the cameras with slow shots

the ball may well be descending and thus launch angles will be less or even negative if measured from trace ends instead of a known launch position height.

Note that when using a fixed launch position (i.e. when both the RAW LA and RAW path options are switched off)

then the fixed "Ball launch position" has to be set in both the left and right cameras separately

as the left and right cameras will see the launch position in a different places.

Switch RAW Path and RAW LA on/off in camera 2 of the Control Panel


Pros and Cons of Stereoscopic vs H cam V cam setups

  • Stereoscopic Pros
  • 1. Increased ball speed detection accuracy
  • Unless calibrated correctly , the V an H cam setup is prone to ball speed measurement errors and inconsistencies when
  • balls are hit left or right and at varying launch angles
  • With the stereoscopic setup, measured ball speed is consistent and accurate no matter in what direction the ball is
  • going in (left or right) and at what launch angle.
  • Note that V and H cam calibration is quite a long and involved process compared to the simple 12 or so step stereo calibration process.
  • 2. Both cameras and lighting are mounted out of the way on the ceiling
  • The V and H cam setups have additional lighting and a camera mounted at floor level which can be inconvenient
  • 3. No light ghosting issues
  • With the stereoscopic setup the lighting is not in view of the cameras so there are no light ghosting issues.
  • Whereas lighting in the V an H cam setup is both on the floor and ceiling - as are the cameras - and thus ghosting light can get into the camera lenses
  • and cause big problems with the image processing
  • Stereoscopic Cons
  • 1. Camera images of the ball make no sense to the viewer
  • Unlike the V and H cam setups where the user can easily see the launch angle and path of the ball,
  • the ball images in a stereoscopic system show no direct relationship to LA and path to the average user.

  • Converging camera test shots

The above screen shots are from customer Kevin showing the left and right camera images os a full shot using the new Converging stereo setup.

Kevin and Keith's comments

I decided to start with little brook manor as you suggested. I know it well since it is one of the R-Motion courses. -I get consistently good traces on full swings as well as putting.

-Putting seemed pretty good! I am no expert and I am out of practice but I felt pretty confident with the distances. -Short chips are a problem with spin. I

like to play bump and runs and the ball would carry the proper distance but instead of running they would backspin. I turned down chipping within 17yards down to 3000 from 5000.

I may turn it down more. -Most full swing shots I would say were great! Fat shots were short and pure shots went what I would say is my norm. I had a few baffling shots. I hit my 5 iron in the 170-180 range

. I had a few fly over 200yards (uphill!). I also hit my 4hybrid about 215. Most were in that range when struck well. I had one that went 265 which is in no way possible for me

I checked the cameras and both traces picked up properly. (I should have taken screen shots) -I believe I corrected my wedge issue with the clubhead getting into the FOV.

I am positioning the ball just in front of the rear line sensor which moved me back just enough. I am a little concerned how this may effect club face and direction if it may move the ball left or right.

Kevin

I figured out a way to get a proper LA with my set up. I shifted the image up so that I can move my launch position back.

I also added a 12x12 fiberbuilt fairway piece ahead of my hitting area. I chip from there and find that the ball flight and carry is pretty accurate on slow chips.

Because I moved the launch position back, full shots are within Ĺ degree of the raw angle shots. So, good news!!!!! You donít have to add anything to the control panel.

Chips that carry 4 yards in the basement, carry 4 yards in the game ( after reducing my LW and SW to 90%)

Flop shots have to be taken from further back as the club head sometimes comes into play. I think that I have everything working as well as my X,Y system.

Short chips are working well now.

Keith

 

  • The above video is from customer Keith showing putting using the new Converging stereo setup.

  • Martin, Yesterday I was having trouble with IBS when it switched to putting. The screen filled with white specs. Today I tried it with the putting camera gain turned all the way down to zero. Problem solved. I suppose that this is the case due to the difference in shutter speed. I did a computer reboot and found that I didn’t have any no reads during the 13 holes that I played today. Maybe suggest that a reboot might be helpful after a CP update. IBS was turned on and it works great. I will be switching out to a full putting green shortly and I am hoping that I won’t have any issues. I have tested the LS cam with pieces of my stance mat from FB. I can adjusted it the it works fine. Chipping and putting was working great. Off the tee, I thought that the ball flight was too straight…. Maybe too forgiving. So I set up for a big draw. I managed to get 27 degrees of right SA……so my system can pick up mishits for sure. Then I set up for a fade but with less SS. The ball started right about 2 degrees and had left SA of 9.7 degrees. So the ball started right and faded back to the left nicely. I am really happy with the new stereo set up. It seems like Kevin is dialing in his settings as well. And the best part…. There is nothing for you to fix at this time.

    Keith

    11/23/18

    Hey Martin and Keith, I just wanted to say thanks and I am in love!!! I believe I have made almost all of my adjustments and the PX5 is awesome! Keith kept telling me to use IBS and I finally switched over tonight.... WOW! The images and reads are great. Attached is two 6 iron pics to show my IBS images. I played a full round tonight. No reads were at a minimum. I had one putt, one tee shot (which I launched from CP), but then I did have to close TGC and GSA and reopen when I was trying to take a half wedge shot that wasn't reading after multiple attempts. After the restart I believe I had one more no read. Keith- I started Q school tonight. My TGC username is Swiggs. It looks like you check on some of the scoring so I wanted to let you know I had one rewind near the very beginning. I hit my hybrid (on 5 wood setting) and it travelled 45 yards too far. I went into the CP and noticed spin was at 7000 and speed was 117%. I brought down to 100% and 3000ish spin and used the CP to relaunch after. ( I don't want to be the new guy who cheats and still gets last place! ha!)

    Kevin

    11/23/18

    Hey Martin !
    I have my cameras officially set up 13.5 feet apart and it is awesome! I did leave the IRs in the original spots about 3 feet apart. Launch angles are much more accurate. Only a small difference between LA and Raw LA. I used an online right angle calculator and the only thing I did was take it down 1* from what LA was reading.
    I did have to slow my shutter speed to 7100. And I am now making small adjustments because every once in a while my wedge will enter the FOV again (which I had perfected in 5.5 foot setup.) So just a few small adjustments angling my line scan a hair further back and sliding the FOV a tiny bit.
    My BIG breakthrough was with the PX mat! I decided since the shaft’s shadow was picking up and always reading wide open to put my ball just off center of tee holder and also angle my halogen light towards me. This way the shadow of the shaft never is able to be sensed no matter how much shaft lean.
    So my report is I am loving it!!! Feels even more accurate. Club readings, ball readings, putting have all been awesome as of my initial few hours with the new setup.

    Kevin

    01/28/19

     

 


 

 

 

 

All camera products Club Tracking Camera Ball spin camera Line Scan camera DIY Kits - Components Enclosures All products PX Systems Software Business Installation and setup