Calibrating your DVRs

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Thu Aug 28, 2014 8:21 pm

Isn't the problem with what you are doing that you a are taking grid lines from a particular camera angle "calibration" and then putting them over the top of a camera scene taken from a different camera angle?

You can angle your camera any way you want as long as its exactly the same as when you calibrate it to when you are riding with it. The way to calibrate is to mark out your centreline, handlebar line, several distances out to 1m, then 1m steps back from your rear wheel with chalk. You can then take the image and draw lines on top of the chalk ones. It also lets you do the calibration scene merging into Photoshop line scene merging into traffic video I posted earlier. Doing it this way shows how non-arbitrary the lines are and that they came from real world measurements.

Even if your lines from calculations are correct (I assume they are), it lacks the direct impact that a shot of chalk markings has which seems like much more solid "proof" to the police officer who really wants every excuse he can have to ignore your lines. It'd be harder to ignore or dismiss chalked ones, than ones that have been drawn using mathemagical trickery. Police officers have enough trouble spelling ;)

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Mon Sep 01, 2014 1:57 pm

InTheWoods wrote:Isn't the problem with what you are doing that you a are taking grid lines from a particular camera angle "calibration" and then putting them over the top of a camera scene taken from a different camera angle?

You can angle your camera any way you want as long as its exactly the same as when you calibrate it to when you are riding with it. The way to calibrate is to mark out your centreline, handlebar line, several distances out to 1m, then 1m steps back from your rear wheel with chalk. You can then take the image and draw lines on top of the chalk ones. It also lets you do the calibration scene merging into Photoshop line scene merging into traffic video I posted earlier. Doing it this way shows how non-arbitrary the lines are and that they came from real world measurements.

Even if your lines from calculations are correct (I assume they are), it lacks the direct impact that a shot of chalk markings has which seems like much more solid "proof" to the police officer who really wants every excuse he can have to ignore your lines. It'd be harder to ignore or dismiss chalked ones, than ones that have been drawn using mathemagical trickery. Police officers have enough trouble spelling ;)

Hi in The Woods
I have taken all the actual pics from the same bit of video shot with the Sony 10 camera. The camera is on the bike for every shot; that car pic is from a video shot with that cam and is a still off the video. The grid is shot with the camera on the bike as well. I have gone over all this previously, but the info is spread over about 4 threads now. :)
The grid is transferred to that line drawing from shots of the grid layed out on the ground with the same camera still fixed to the bike.

The grid is layed out with 8 metre tapes and all you need do is select a colour, click whilst holding shift and click on the other end. Repeat on all tapes and you have you (chalk) outline., true to life.

This is the shot of the grid, all I have done is draw lines over the laid out tapes, concealed every thing else and am left with a grid that I can make any colour. All shot with the same camera and at the same lens setting, the camera never left the bike.
That is why any grid has to be done with the owners own particular camera.I provide all this in pictures to the police with each complaint, in an easily understood manner to show how I calibrated my camera. No mathemagical trickery at all.
What I have done is make a grid from a real life photograph. No point in doing anything else.
But it is not for every one, and I am only trying to make the process understood and the effects of a camera, and how, if pointing the wrong way just what the camera presents to the eye.
Lets just say that the different results over time will aid us all in the result and final out comes of reports to the police. But even then every thing is in the eye of the beholder and the police mans initiative or lack there off.
I agree it is a bit hard to fathom if you don't see the process done. Being a non clear scientist myself it is all just an attempt to get the right combination of proof. I take every bit of info on board and I may be barking up the wrong tree as far as simplicity is concerned, I realise that.

Image
Vivente World Randonneur complete with panniers

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Mon Sep 01, 2014 2:07 pm

Oxford wrote:After seeing some of what has been posted, I have decided to wind back the detail and just show what I had back in the earlier versions of just showing the markers around the edge. I really only want to show what an infringer is doing as they pass me. Not so far back that the evidentiary nature of the content can be called into question. After all it is only of concern what happens at the time they pass you, not 5 or more metres back. I think it is becoming a little too complicated and you have to remember, if someone decides to contest in court and reasonable doubt can be created, the charge is thrown out.


Yep I agree that the simpler the better. Why I have persisted with the 8 metre grid is for experimental purposes to help align the camera and to see the effects of side ways and vertical manipulations.
seeing some fly6 shots pointing way left or right needed some explaining, I thought.
But at incident level it makes no difference to be only 2.4 metres out once the camera is set up.
Members have shown some grids that go to a point on the horizon over short distances though, and I was trying to get as realistic a grid as possible.Explaining it all is a bit hard, and I am learning the ropes on what we need as well.
Vivente World Randonneur complete with panniers

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Mon Sep 01, 2014 3:15 pm

outnabike wrote: This is the shot of the grid, all I have done is draw lines over the laid out tapes, concealed every thing else and am left with a grid that I can make any colour. All shot with the same camera and at the same lens setting, the camera never left the bike.

Ah ok, I wasn't getting that gist.

My point was only that it doesn't matter which way your camera is pointing as long as it is in the same position&direction for normal use as when you do the chalk/tape lines for the calibration shot. I thought you were saying if you didn't get the camera pointed exactly straight back and level the calibration was no good.

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Mon Sep 01, 2014 5:35 pm

InTheWoods wrote:
outnabike wrote: This is the shot of the grid, all I have done is draw lines over the laid out tapes, concealed every thing else and am left with a grid that I can make any colour. All shot with the same camera and at the same lens setting, the camera never left the bike.

Ah ok, I wasn't getting that gist.

My point was only that it doesn't matter which way your camera is pointing as long as it is in the same position&direction for normal use as when you do the chalk/tape lines for the calibration shot. I thought you were saying if you didn't get the camera pointed exactly straight back and level the calibration was no good.


I am saying its no good if it is more than a couple of degrees sidways to the left or right. What happens is , if a car comes close on your right to overtake you, and your camera is twisted toward the car , it will always appear closer. It will give an indication of closeness out at two metres but lose perspective when right beside you. Its the same as standing by a wall and filming the wall then swing into the wall beside you. Ofcourse it will appear closer.

The up or down element is not as bad but still makes the view for-shortened and appears to distort the grid. I have indicated the point of the disappearing perspective in green.
The camera cant see below ground but makes the 8 metre grid look shorter and compacts it. It is virtually the opposite effect to a tele photo lens.
Image

Here is the effect in reality, it distorts the area of how the car is coming at the camera. In reality the camera lens view is the centre of its radial view.
so instead of the end of the grid being at the 8 metre blue mark it is now for -shortened and the grid appears out of line with the centre road markings (in yellow). It is not a real view of the situation. I am no lawyer but I reckon in a dispute this could well be very important.
It will be interesting to see how the police view all this in a court case is what I am saying. All this is shown above in the samples but needs careful observation.

Image
Vivente World Randonneur complete with panniers

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Mon Sep 01, 2014 6:05 pm

outnabike wrote:I am saying its no good if it is more than a couple of degrees sidways to the left or right. What happens is , if a car comes close on your right to overtake you, and your camera is twisted toward the car , it will always appear closer. It will give an indication of closeness out at two metres but lose perspective when right beside you. Its the same as standing by a wall and filming the wall then swing into the wall beside you. Ofcourse it will appear closer.


Here I totally disagree with you. We aren't dealing with how close things "look" or appearances. The only thing that matters is where the tyres touch the road surface relative to the calibration lines superimposed over the top, with those lines taken from real world chalk or tape markings measured out on a flat surface. All of your angles and trig are making it way more complicated than it has to be.

Lets say I point my camera 45 degrees to the right. I mark out the chalk lines on my driveway, take the calibration shot, and draw the lines in photoshop where the chalk is. Now I go for a ride (with the camera still 45 degrees to the right) and a car tyre passes me driving in a steady straight line past me on the 50cm line. Presuming the road was flat like my driveway and the road is straight, how is the tyre not passing 50cm away from the tip of my handlebars? It was precisely where I measured 50cm to be. Are you saying that because the camera isn't pointed straight back, that the 50cm chalk line in the images it takes are not actually 50cm away even though that is the physical reality?

Its irrelevant what angle the camera is at, as long as its at the same angle as when I did the calibration. The only way to make it wrong is to have remounted the camera at a different angle (say 30 degrees) but still use the calibration lines I got when it was mounted at 45 degrees.

Once you have the calibration shot of the chalk lines, distortion is irrelevant because the chalk lines show, *in the distorted world of the camera*, how far things (chalk lines, car tyres) are away from you on the ground. When you look at a video of a close pass when the camera was mounted identically, you are looking at it through the *same distorted camera view* so the chalk lines and car tyre positions can be compared.

Here is the effect in reality, it distorts the area of how the car is coming at the camera. In reality the camera lens view is the centre of its radial view.
so instead of the end of the grid being at the 8 metre blue mark it is now for -shortened and the grid appears out of line with the centre road markings (in yellow). It is not a real view of the situation. I am no lawyer but I reckon in a dispute this could well be very important.


You are confusing me here. Your image shows distortion/inaccuracy with the red lines clearly being wrong, so what did you change to make them wrong? You said a) that your camera was always mounted in the same spot and that b) the calibration lines are from the camera in the same spot. So how did the red lines in your shot get wrong? Its impossible that they give a different perspective if you don't change anything. Either you drew them in a different spot, or you shifted the camera when you went on the road, or you used a different calibration shot. Otherwise the red lines would be in the right spot.

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Mon Sep 01, 2014 6:42 pm

Please look up at the 4 pics that I provided as an example. I will say no more about it.
Just found this and it may help.
viewtopic.php?f=12&t=58039&start=7775

http://s238.photobucket.com/user/diesel ... d.jpg.html
Last edited by outnabike on Mon Sep 01, 2014 6:55 pm, edited 1 time in total.
Vivente World Randonneur complete with panniers

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Mon Sep 01, 2014 6:54 pm

outnabike wrote:Please look up at the 4 pics that I provided as an example. I will say no more about it.


All the 2nd, 3rd and 4th pictures show is that when you make a mistake when mounting your camera, you get inaccurate results. You have 3 totally different calibration line shots applied to one image (the car photo is identical in all three). A bit captain obvious?

Lets focus on the 2nd picture.
1. The calibration lines are taken with the camera pointing 2 degrees down.
2. The shot of the car is incorrectly taken with the camera level.
There should be no surprise you get a bad result here, so I fully agree with you on this.

The correct way to have done this was to
1. Take the calibration lines with the camera pointing 2 degrees down
2. Take the shot of the car with the camera also pointing 2 degrees down.
This would give you an accurate result.

All you have shown is if you point your camera in a different direction to normal when you do your calibration shot, your results will be out? Isn't that obvious?

Try mounting the camera 2 degrees down on your bike, then compare it to your 2 degrees down calibration shot and you'll find that it is accurate once more.

So once again, you can point your camera any angle you like, so long as it is identical for your calibration shot AND normal use. The easiest way to do this (I thought this was obvious) was mount the camera on your bike in its normal spot, take the calibration shot, and then continue to mount the camera in exactly the same normal way. If you start messing with the angle of the camera after taking the calibration shot, well, I totally agree that is bad.

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Mon Sep 01, 2014 9:09 pm

Its occurred to me I could make things a bit clearer, because on the whole we actually are in agreement.

1. I agree with you that your first shot is a good calibration
2. I agree with you that your second shot is an inaccuratecalibration.
3. I agree with you that your third shot is an inaccurate calibration.
4. I agree with you that your fourth shot is an inaccurate calibration.

We only disagree on the reason why 2 3 4 are inaccurate.

You say its because the camera is not centred. While this is true, I am saying it is not the real reason. The real root reason is that it is because the camera angle you used for the car photo (centred) is different to the angle used for the calibration (not centred). The camera can be pointed at any angle you like as long as its the same for calibration and recording use.

Its like calibrating a thermometer. You wouldn't calibrate it and then add more mercury. When you calibrate something you don't change anything afterwards, else you have to re-calibrate it for the new conditions.

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Mon Sep 01, 2014 10:37 pm

Thanks mate I know what you mean.
All I was trying to show here is how small percentages in movement have a certain result. Not necessarily all bad news though. Back in those posts I pointed the camera down to show the divergence between the centre line of the road and the grid (that bit in yellow on the last pic) to show the difference. No one can doubt their grid will be the same with the same camera if they do not change the cameras position.
But it is a more effective proof of a close shave the straighter out the back the camera is.

Offcourse my efforts in explanation need to be improved. In an effort to demonstrate I made a drawing to show a plan view of what I mean.
It is the sideways position that is most important and until you actually try to set a camera up it is a bit hard to see what I am trying to say.

Image
Vivente World Randonneur complete with panniers

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Mon Sep 01, 2014 11:23 pm

Agreed and if you tilt the bike it also affects it.

When I played around with different angles it showed that with the calibration lines closest ie 25cm and 50cm at the bottom of the frame weren't thrown out that much, but the more distant ones were. Another reason not to bother the police for passes greater than 75cm clearance.

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Tue Sep 02, 2014 9:36 am

InTheWoods wrote:Agreed and if you tilt the bike it also affects it.

When I played around with different angles it showed that with the calibration lines closest ie 25cm and 50cm at the bottom of the frame weren't thrown out that much, but the more distant ones were. Another reason not to bother the police for passes greater than 75cm clearance.


Yes, In rifle shooting if you lay a scope left or right from vertical it is called cant and will result in a miss. If you were pedalling hard uphill (out of the saddle) and took still shots during two heavy (Left and right) pedal strokes there would be a big difference in the apparent closeness of a vehicle right beside you. That is why you get sea sick viewing footage of a rider doing this. The bike may cant 15 degrees off vertical and swing back. In reality at 30 frames per second though we can establish reality and I have deliberately not gone down that path, as its not really a concern.

If there was one thing that would assist a cyclist to set their camera up it would be a lazer pointer built into the camera. This was popular as a rough alignment tool with some cameras at inception, mainly due to them being round in design.
You get a completely different concept of lining up a camera if you draw the centre line 2 meters longer than the proposed grid and place the bike, camera attached on that centre line. If you also have a vertical centre line against the wall at the far end. Your footage will soon enough show where to point the camera.
The reason that I have gone for an eight metre long grid is simply for accuracy of overall view. the longer the grid viewed on a road way the more we see its accuracy to the roadway we are riding on. Ther is no problem to shorten the grid in actual use if you want off course.
I am also not giving away the concept of say one vertical line either as it can show the body overhang of the offending vehicle.

This earlier grid shot gives not only an idea of the wheel placement but also an idea of where the mirror is going to be as it passes the bike. If the camera was pointing 5 degrees left in this instant it would show the car actually in top of the rider.
Image
Vivente World Randonneur complete with panniers

User avatar
InTheWoods
Posts: 1840
Joined: Sun Jul 31, 2011 2:34 pm
Location: Brisbane

Re: Calibrating your DVRs

Postby InTheWoods » Tue Sep 02, 2014 12:22 pm

The vertical lines are interesting but I don't think they can prove anything without much more effort. To identify the mirror's location in space from your photo, you would need to know how high the mirror is on that car, how far back the car tyre is from you, how far it is from the car tyre to the mirror, and how far the mirror sticks out past where the tyre contacts the ground. A stick the same height as the mirror could be placed at varying clearances at the same distance away from you as the mirror is in the photo could then show how close it was. Effectively you would have to take a lot of datapoints when doing the calibration so you could draw a 3d grid - one on the ground, one at the height of the mirror. Would look cool, but its simpler just to measure the wheel clearance. Maybe I'll do it just for fun anyway one day. The bike would have to be locked in a stand while you move around with the stick to get the data points.

Btw in that shot of what is obviously a close pass, your camera is not mounted quite the same as when you did the calibration (is this a calibration from your real scene or a calculated one based on camera focal length), or you are angled to your left to avoid said car because you saw it was coming at you. You can see the centreline is pointing at the other side of the road (a path we can assume you didn't follow :)) and the other lines are not parallel to the rest of the scene (the infinity point of the lines is near the car's rear bumper, but the infinity point of the scene itself appears to be near the blue(?) car(?) in the far distance).

User avatar
outnabike
Posts: 2068
Joined: Tue Jan 08, 2013 2:53 pm
Location: Melbourne Vic

Re: Calibrating your DVRs

Postby outnabike » Tue Sep 02, 2014 8:41 pm

Yep its a different grid, and is only 500 wide at the centre . My first attempt. The lines are messy allright but it is easy to nominate a size on just one line say at a metre high and it is not hard to extrapolate a distance for the car as the grid is 2.4 precisely from front to back. But you have to get a Measurement from the front of the camera to its actual view of the beginning of the cameras field of view of the grid.
I may yet divide the grid longitudinally into 300 mm sections along one line. Thanks for taking the time to talk on this, It's a big help and gets all sorts of ideas started. :)
Vivente World Randonneur complete with panniers

diggler
Posts: 624
Joined: Sun Aug 01, 2010 11:23 pm

Re: Calibrating your DVRs

Postby diggler » Tue Feb 17, 2015 5:40 pm

Sorry for not reading all of the thread, but has there ever been a prosecution under the 1m rule based on video evidence?

My guess is a cop confronted with a video and electronic lines would throw his hands in the air and cry too hard. I think a cop would prosecute if there was actual contact. Pretty unambiguously within 1m. Anything less than contact, then we are in the realm of why am I bothering if nobody got hit and can you really tell how far the car is from the bike?
Motorists hate cyclists and cyclists hate the motorists and the pedestrians hate the bikers and everybody hates the trucks.

madmacca
Posts: 356
Joined: Mon Aug 03, 2015 5:13 pm

Re: Calibrating your DVRs

Postby madmacca » Fri Feb 12, 2016 11:05 am

diggler wrote:Sorry for not reading all of the thread, but has there ever been a prosecution under the 1m rule based on video evidence?

My guess is a cop confronted with a video and electronic lines would throw his hands in the air and cry too hard. I think a cop would prosecute if there was actual contact. Pretty unambiguously within 1m. Anything less than contact, then we are in the realm of why am I bothering if nobody got hit and can you really tell how far the car is from the bike?


Yeah, I'm inclined to agree. Regardless of the offence, police don't want to prosecute unless they are reasonably sure of a conviction. And a defence lawyer could drive a truckload of doubt between the calibration marks of user-supplied footage, because of the uncertainty of the rigour of someone's calibration efforts. (this is a comment on defence lawyers, not outnabike's calibration efforts).

Where things could be more interesting was if there were factory calibration marks (possibly software defined?) in the Fly6 itself. One test case with expert witnesses from Fly6, then police would be more confident that it would stand up in court.

diggler
Posts: 624
Joined: Sun Aug 01, 2010 11:23 pm

Re: Calibrating your DVRs

Postby diggler » Sat Jun 24, 2017 2:06 am

I've usually been down on videos in the past but I've had a change of heart. I have a horizontal flag which extends 50cm beyond the tip of my handlebars. Some !(AT)#$wit hit the end of my flag today so they were within 50cm of me let alone 1m. Unfortunately I didn't see the rego and I had no footage. A video camera is therefore useful IN CONJUNCTION with a horizontal flag to police the 1m rule.
Motorists hate cyclists and cyclists hate the motorists and the pedestrians hate the bikers and everybody hates the trucks.

Who is online

Users browsing this forum: No registered users