offer - speed from video tape to refute LTI2020
Discussion
Hi
I made this offer in another thread but I think it would be interesting to do it so I'll make it more widely. The idea comes from the fact that many people believe that erroneous LTI20/20 readings have been used to achieve prosecutions but the authorities refuse to admit the device may have erred. In fact, the video recording taken during operation of the device (is this always done?) may contain proof of the laser's error.
Given a high resolution video of a vehicle from a pan/tilt camera position it is possible to remove background motion and recover the vehicle speed. All that is required is a calibration measure from the scene e.g. vehicle width.
If
- you are being prosecuted for speeding and
- genuinely believe the speed reading is erroneous and
- you have access to the video sequence in which the speed reading is captured and
- you are willing to fight it in court
.. I am happy to assess the video to recover vehicle speed throughout the sequence.
The assessment will be a replicable mathematical procedure and I am happy to produce documents explaining the process, and to testify in court if necessary.
The procedure will always work, however, depending on camera angle, vehicle distance and video resolution the error (which will be known) may be too large to refute the laser speed measurement.
I would also appreciate any comments on the utility of this general approach!
sincerely.
dave
I made this offer in another thread but I think it would be interesting to do it so I'll make it more widely. The idea comes from the fact that many people believe that erroneous LTI20/20 readings have been used to achieve prosecutions but the authorities refuse to admit the device may have erred. In fact, the video recording taken during operation of the device (is this always done?) may contain proof of the laser's error.
Given a high resolution video of a vehicle from a pan/tilt camera position it is possible to remove background motion and recover the vehicle speed. All that is required is a calibration measure from the scene e.g. vehicle width.
If
- you are being prosecuted for speeding and
- genuinely believe the speed reading is erroneous and
- you have access to the video sequence in which the speed reading is captured and
- you are willing to fight it in court
.. I am happy to assess the video to recover vehicle speed throughout the sequence.
The assessment will be a replicable mathematical procedure and I am happy to produce documents explaining the process, and to testify in court if necessary.
The procedure will always work, however, depending on camera angle, vehicle distance and video resolution the error (which will be known) may be too large to refute the laser speed measurement.
I would also appreciate any comments on the utility of this general approach!
sincerely.
dave
Interesting, and the theory is sound.
However...you will be asking magistrates to give your argument/opinion/evidence more weight than that of the well-established makers and operators of the LTI. More than that, I would suggest the magistrates will probably be under pressure to 'toe the line' (sad, but true...there's a LOT of people now with a vested interest in the current system).
Having said that, one of the grounds for appeal under UK law is 'point of fact'. If your evidence is mathematically sound, then that should fall into the category of fact. But I'd say you'll still need an open-minded judge somewhere along the line.
With all those caveats said, I hope it works, and I wish you luck!!!
However...you will be asking magistrates to give your argument/opinion/evidence more weight than that of the well-established makers and operators of the LTI. More than that, I would suggest the magistrates will probably be under pressure to 'toe the line' (sad, but true...there's a LOT of people now with a vested interest in the current system).
Having said that, one of the grounds for appeal under UK law is 'point of fact'. If your evidence is mathematically sound, then that should fall into the category of fact. But I'd say you'll still need an open-minded judge somewhere along the line.
With all those caveats said, I hope it works, and I wish you luck!!!
one small prob.
You 'll present your evidence - all mathematically sound and all that.
BUT, Mr CPS may say.......
'Is there room for error in your method of calculating speed?'
'Erm, yes'
'And would this error have the effect of increasing or decreasing your speed results?'
'erm...........decreasing your honour'
'So what you are saying is, your method could produce a result that could be LOWER than the actualy speed of the vehicle?'
'.......yes'
oh dear.
That's even before zoom, observation angles, cosine effect, lens geometry, and even heat haze have been factored in.
You 'll present your evidence - all mathematically sound and all that.
BUT, Mr CPS may say.......
'Is there room for error in your method of calculating speed?'
'Erm, yes'
'And would this error have the effect of increasing or decreasing your speed results?'
'erm...........decreasing your honour'
'So what you are saying is, your method could produce a result that could be LOWER than the actualy speed of the vehicle?'
'.......yes'
oh dear.
That's even before zoom, observation angles, cosine effect, lens geometry, and even heat haze have been factored in.
bryan35 said:
one small prob.
You 'll present your evidence - all mathematically sound and all that.
BUT, Mr CPS may say.......
'Is there room for error in your method of calculating speed?'
'Erm, yes'
'And would this error have the effect of increasing or decreasing your speed results?'
'erm...........decreasing your honour'
'So what you are saying is, your method could produce a result that could be LOWER than the actualy speed of the vehicle?'
'.......yes'
oh dear.
That's even before zoom, observation angles, cosine effect, lens geometry, and even heat haze have been factored in.
You 'll present your evidence - all mathematically sound and all that.
BUT, Mr CPS may say.......
'Is there room for error in your method of calculating speed?'
'Erm, yes'
'And would this error have the effect of increasing or decreasing your speed results?'
'erm...........decreasing your honour'
'So what you are saying is, your method could produce a result that could be LOWER than the actualy speed of the vehicle?'
'.......yes'
oh dear.
That's even before zoom, observation angles, cosine effect, lens geometry, and even heat haze have been factored in.
I agree. The standard CPS response is to show the calibration certificates for the device and show the type approval documents for the device signed by the Home Secretary. Propaganda maybe, but very strong propaganda in the eyes of the magistrates eager to seek the most legal and likely scenario. This will then be supported by the operator promising that they used the device properly where no written evidence is there to prove otherwise and where the CPS cannot refute the glaring errors as the operator has slipped up or there is written or recorded evidence of failures then the ACPO Code of Practice then is renamed 'only guidelines' and the CPS prove that the operator has been supposedly trained properly.
Sorry to be less optomistic but any other estimate or calculation of speed would be useful but the defendant would still have to get around the above. Believe me I have tried!
hi guys
The method of calculation I am thinking about would give upper and lower bounds on possible speed (known error).
e.g.
visually-recovered est. speed 35 mph would come with something like
minimum possible speed: 30 mph
maximum possible speed: 40 mph
Now say laser estimation was: 55 mph
That'd give you evidence that the laser is wrong, even allowing for error due to processing.
I think it could quite easily be argued as a point of fact. You'd only have to get a few image processing experts / mathematicians etc to go through the procedure and confirm its veracity.
This is all assuming some serious laser error e.g. via reflectance/slip giving a wildly inaccurate result, and suitable video evidence.
regards
dave
The method of calculation I am thinking about would give upper and lower bounds on possible speed (known error).
e.g.
visually-recovered est. speed 35 mph would come with something like
minimum possible speed: 30 mph
maximum possible speed: 40 mph
Now say laser estimation was: 55 mph
That'd give you evidence that the laser is wrong, even allowing for error due to processing.
I think it could quite easily be argued as a point of fact. You'd only have to get a few image processing experts / mathematicians etc to go through the procedure and confirm its veracity.
This is all assuming some serious laser error e.g. via reflectance/slip giving a wildly inaccurate result, and suitable video evidence.
regards
dave
havoc said:
davidra said:
This is all assuming some serious laser error e.g. via reflectance/slip giving a wildly inaccurate result, and suitable video evidence.
regards
dave
regards
dave
And in these cases the prosecution will likely realise it and drop the case before it gets to court.
Dr Mike Clark has some evidence available from a previous case that was so bleedin obvious it was a joke!!! I'm sure he'd love to have the video work done at that one.
There may be another way to do this, although this is just an idea I have in my head and may not be workable in practice.
In theory you could use the framerate of the video to accurately measure the time taken for the car to pass two visible points on the road. These points would have to be clearly visible on the tape and not too far from the car so that it could be clearly shown to pass over them. Something like catseyes, marks on the road surface, broken white lines or drain covers would probably be ok but lamp posts or other roadside items may be too far from the line of the car.
The video will operate at a known framerate, probably 24 or 25 frames per second, and providing you can accurately measure the actual points covered you can use the number of frames taken to pass between the two points to calculate the average speed. Obviously this might be tricky on a motorway or dual carriageway but you may be able to get the details from the local authority , via a summons if necessary.
Obviously this wont work where there are no obvious marks on the road or the video is very short, but it may help in some cases.
Opinions, anyone ?
In theory you could use the framerate of the video to accurately measure the time taken for the car to pass two visible points on the road. These points would have to be clearly visible on the tape and not too far from the car so that it could be clearly shown to pass over them. Something like catseyes, marks on the road surface, broken white lines or drain covers would probably be ok but lamp posts or other roadside items may be too far from the line of the car.
The video will operate at a known framerate, probably 24 or 25 frames per second, and providing you can accurately measure the actual points covered you can use the number of frames taken to pass between the two points to calculate the average speed. Obviously this might be tricky on a motorway or dual carriageway but you may be able to get the details from the local authority , via a summons if necessary.
Obviously this wont work where there are no obvious marks on the road or the video is very short, but it may help in some cases.
Opinions, anyone ?
davidra said:
hi guys
The method of calculation I am thinking about would give upper and lower bounds on possible speed (known error).
e.g.
visually-recovered est. speed 35 mph would come with something like
minimum possible speed: 30 mph
maximum possible speed: 40 mph
Now say laser estimation was: 55 mph
That'd give you evidence that the laser is wrong, even allowing for error due to processing.
I think it could quite easily be argued as a point of fact. You'd only have to get a few image processing experts / mathematicians etc to go through the procedure and confirm its veracity.
This is all assuming some serious laser error e.g. via reflectance/slip giving a wildly inaccurate result, and suitable video evidence.
regards
dave
The method of calculation I am thinking about would give upper and lower bounds on possible speed (known error).
e.g.
visually-recovered est. speed 35 mph would come with something like
minimum possible speed: 30 mph
maximum possible speed: 40 mph
Now say laser estimation was: 55 mph
That'd give you evidence that the laser is wrong, even allowing for error due to processing.
I think it could quite easily be argued as a point of fact. You'd only have to get a few image processing experts / mathematicians etc to go through the procedure and confirm its veracity.
This is all assuming some serious laser error e.g. via reflectance/slip giving a wildly inaccurate result, and suitable video evidence.
regards
dave
Hi Dave,
I'm very interested in what you have to say.
See my new contribution to Peter Wards impending case.
Those on here who talk about how the courts would view it seem to miss the point that all that is need for an acquittal is reasonable doubt, not absolute proof.
If the magistrates miss this, as they frequently do, then it has to go to appeal, and I have found a much greater level of succes in the Appeal Court.
E-mail me if you want to put a few ideas down that we could work on.
njwc said:
There may be another way to do this, although this is just an idea I have in my head and may not be workable in practice.
In theory you could use the framerate of the video to accurately measure the time taken for the car to pass two visible points on the road. These points would have to be clearly visible on the tape and not too far from the car so that it could be clearly shown to pass over them. Something like catseyes, marks on the road surface, broken white lines or drain covers would probably be ok but lamp posts or other roadside items may be too far from the line of the car.
The video will operate at a known framerate, probably 24 or 25 frames per second, and providing you can accurately measure the actual points covered you can use the number of frames taken to pass between the two points to calculate the average speed. Obviously this might be tricky on a motorway or dual carriageway but you may be able to get the details from the local authority , via a summons if necessary.
Obviously this wont work where there are no obvious marks on the road or the video is very short, but it may help in some cases.
Opinions, anyone ?
In theory you could use the framerate of the video to accurately measure the time taken for the car to pass two visible points on the road. These points would have to be clearly visible on the tape and not too far from the car so that it could be clearly shown to pass over them. Something like catseyes, marks on the road surface, broken white lines or drain covers would probably be ok but lamp posts or other roadside items may be too far from the line of the car.
The video will operate at a known framerate, probably 24 or 25 frames per second, and providing you can accurately measure the actual points covered you can use the number of frames taken to pass between the two points to calculate the average speed. Obviously this might be tricky on a motorway or dual carriageway but you may be able to get the details from the local authority , via a summons if necessary.
Obviously this wont work where there are no obvious marks on the road or the video is very short, but it may help in some cases.
Opinions, anyone ?
I agree with this. I do believe that some cases have been tried with this evidence. I think it was someone who was contesting a case where a trafpol car was videoing evidence att he same time. Whereas the officers used the onboard 'computer' which need the operator to flick a switch on and off through a set distance, this was tested by the timecode on the video as the car went past markers by the side of the road.
The way it could be used in this occasion would be by getting a metre rolling tool and measure out the distance by the side of the road between two reference points on the video. You could then calculate an average speed of the car between two points. This would be remarkably simple. All you would have to do is state on oath how you had measured the distance. You could also show the video and show the timecoding and even show the magistrates how to work it out themselves.
I think that would be a lot more convincing than the evidence from a most likely misused dodgyscope.
Justin we're basically talking about the same principle, it's just a matter of how many points you can measure to refine the result. In fact you can use multiple points to corroborate each other.
The video framerate is suitable for timing with an error <= 1 frame delay because the lag is never cumulative. However, if a video has been digitally captured then depending on encoding used the period between frames might not be a truthful time signal.
my email is
david . rawlinson "at" eng.monash.edu.au
cheers
dave
The video framerate is suitable for timing with an error <= 1 frame delay because the lag is never cumulative. However, if a video has been digitally captured then depending on encoding used the period between frames might not be a truthful time signal.
my email is
david . rawlinson "at" eng.monash.edu.au
cheers
dave
Mrs LongQ was entrapped by an LTI20:20 some years ago - just before the whole scam really began to grab everyone's imagination. She was adamamnt that she was doing 30 in a 30. The LTI recorded 40.
We went to see the video. At 300 meter range it was difficult to judge likely speeds. We had two pics though - some distance apart so that they coould get an image of the number plate (and try and catch a few more wayward drivers in between) so I knew the elapsed time and the frame count of the video. I also know with about a meter where the camera was and the second picture gave me excellent triangulation information related to a nearby building.
So I got the large scale road and building plans, did some calcs and could prove that the average speed over the distance (no braking or intentional speed change occured I was told) was 31mph plus or minus 1mph according to my estimates.
What I could NOT prove, and I don't think the video would have had the resolution of aything to assist, was that the speed at point of measure (at the bottom of a down slope which, foot off throttle, would add 3 to 4 mph to the car's speed) could not have been 40mph. Even if a 100 meters later the speedo was showing 30 anyone could have claimed that a slight braking effect would have reduced the speed in fractions of a second at some point during that period and contributed to the overall average calculated.
So in the end any court case would have to have relied upon the magistrate accepting the evidence of the driver that they had not slowed intentionally or used the brakes ... yep, right. Even then I would have thought it would be tricky. Having read stories since than about the way the cases go it would seem a bit pointless to pursue it.
However, if the technique (and presumably the information available from the video images) can be used effectively related to the 1/3rd second during which the measurement is made, irrespective of the distance over which the laser/caera combination is working, then there may be a chance it could be accepted. 1/3rd second at a 50 per second frame rate suggests about 16 frames to play with with a possible 2 frame overlap discrepancy for the start and stop points (?) so an error potential of 1/8th or 12.5% assuming reasonably consistent correlation between the laser and the video device.
Now comes the question of optics.
In the recent post with the image taken at over 700 meters the detail it pretty poor which , I would imagine, does not help the cause at all. On the other hand when Gwent/South Wales were operating on various parts of the M4 a few years ago they used longer lenses, so the image was good BUT the compression of depth of view that such lenses deliver also makes accurate distance comparisons very tricky.
So as David says if the error is relatively great there is a potential to cast doubt upon it and that SHOULD be enough. If it a small discrepancy it may not be much help.
However, cast enough doubt on the large discrepancies and the smaller ones can also be brought into question without necessarily proving any errors.
Or is that wishful thinking?
Whichever way it goes it has to be worth following through.
We went to see the video. At 300 meter range it was difficult to judge likely speeds. We had two pics though - some distance apart so that they coould get an image of the number plate (and try and catch a few more wayward drivers in between) so I knew the elapsed time and the frame count of the video. I also know with about a meter where the camera was and the second picture gave me excellent triangulation information related to a nearby building.
So I got the large scale road and building plans, did some calcs and could prove that the average speed over the distance (no braking or intentional speed change occured I was told) was 31mph plus or minus 1mph according to my estimates.
What I could NOT prove, and I don't think the video would have had the resolution of aything to assist, was that the speed at point of measure (at the bottom of a down slope which, foot off throttle, would add 3 to 4 mph to the car's speed) could not have been 40mph. Even if a 100 meters later the speedo was showing 30 anyone could have claimed that a slight braking effect would have reduced the speed in fractions of a second at some point during that period and contributed to the overall average calculated.
So in the end any court case would have to have relied upon the magistrate accepting the evidence of the driver that they had not slowed intentionally or used the brakes ... yep, right. Even then I would have thought it would be tricky. Having read stories since than about the way the cases go it would seem a bit pointless to pursue it.
However, if the technique (and presumably the information available from the video images) can be used effectively related to the 1/3rd second during which the measurement is made, irrespective of the distance over which the laser/caera combination is working, then there may be a chance it could be accepted. 1/3rd second at a 50 per second frame rate suggests about 16 frames to play with with a possible 2 frame overlap discrepancy for the start and stop points (?) so an error potential of 1/8th or 12.5% assuming reasonably consistent correlation between the laser and the video device.
Now comes the question of optics.
In the recent post with the image taken at over 700 meters the detail it pretty poor which , I would imagine, does not help the cause at all. On the other hand when Gwent/South Wales were operating on various parts of the M4 a few years ago they used longer lenses, so the image was good BUT the compression of depth of view that such lenses deliver also makes accurate distance comparisons very tricky.
So as David says if the error is relatively great there is a potential to cast doubt upon it and that SHOULD be enough. If it a small discrepancy it may not be much help.
However, cast enough doubt on the large discrepancies and the smaller ones can also be brought into question without necessarily proving any errors.
Or is that wishful thinking?
Whichever way it goes it has to be worth following through.
My thinking on judging speed in the critical 1/3 second during which speed is laser-measured:
If it can be shown that car cannot have braked or accelerated quickly enough in the preceeding and subsequent moments, a longer time period may be used for speed estimation and still refute the laser. Maximum rate of acceleration/dec. would give us a few seconds probably.
Plus acc/dec. can be plotted throughout a sequence, backed up by corroborative factors from video such as car attitude (car braking hard will be nose-down).
dave
If it can be shown that car cannot have braked or accelerated quickly enough in the preceeding and subsequent moments, a longer time period may be used for speed estimation and still refute the laser. Maximum rate of acceleration/dec. would give us a few seconds probably.
Plus acc/dec. can be plotted throughout a sequence, backed up by corroborative factors from video such as car attitude (car braking hard will be nose-down).
dave
davidra said:
My thinking on judging speed in the critical 1/3 second during which speed is laser-measured:
If it can be shown that car cannot have braked or accelerated quickly enough in the preceeding and subsequent moments, a longer time period may be used for speed estimation and still refute the laser. Maximum rate of acceleration/dec. would give us a few seconds probably.
Plus acc/dec. can be plotted throughout a sequence, backed up by corroborative factors from video such as car attitude (car braking hard will be nose-down).
dave
If it can be shown that car cannot have braked or accelerated quickly enough in the preceeding and subsequent moments, a longer time period may be used for speed estimation and still refute the laser. Maximum rate of acceleration/dec. would give us a few seconds probably.
Plus acc/dec. can be plotted throughout a sequence, backed up by corroborative factors from video such as car attitude (car braking hard will be nose-down).
dave
OK but, in reverse order, I don't think that a dab on the brakes would be very evident on the video unless it was quite extreme, even at 300 meters. At 700+ meters with a 'normal' lens on the video you would nt be able to identify the difference between a brake application and a typical pothole or bump. (Might be different with a telephoto lens.) Even if close analysis might show a differnece I doubt the court would get excited about it (i.e. the LACK of any evidence ....). Many cars show little nose down attitude anyway these days, especially the more 'interesting' ones. Bear in mind that wheb we watched the video we saw about an hour's worth as the operator was having problems with the playback system. Only one car, in all the video time we saw, showed clear evidence of braking and that one was accelerating for an overtake at about 55mph iirc when the driver seemed to spot the camera presence.
When I looked at what would be required to slow from 40 to, say, 30 and achieve a 31mph average over the 220 meters (from memory) between the measurement point and the second frame (in Mrs. LongQ's case) I was surprised at the number that came up. Basically quick dab on the brakes for not more than a second or a gentle application over a couple of seconds would be enough. Lift off for about 3 or 4 seconds, given a manual box, might also be enough. (Of course the probability that the vehicle had only momentarily hit anywhere near the claimed speed, and that that was entirely due to the down slope, would be the converse argument which, no doubt, would also be ignored by any court. I wonder if they would accept a defence based on ecology and saving the planet?)
So my conclusion was that using any evidence, other than some type of high frequency recording system like a Race Logic device, would of necessity have to specifically cast doubt upon the data from the period of measurement and no period either side of it. However in a more discretionary regime a longer period and reliance on average speeds rather than an extreme peak might acceptable.
davidra said:
Justin we're basically talking about the same principle, it's just a matter of how many points you can measure to refine the result. In fact you can use multiple points to corroborate each other.
The video framerate is suitable for timing with an error <= 1 frame delay because the lag is never cumulative. However, if a video has been digitally captured then depending on encoding used the period between frames might not be a truthful time signal.
my email is
david . rawlinson "at" eng.monash.edu.au
cheers
dave
The video framerate is suitable for timing with an error <= 1 frame delay because the lag is never cumulative. However, if a video has been digitally captured then depending on encoding used the period between frames might not be a truthful time signal.
my email is
david . rawlinson "at" eng.monash.edu.au
cheers
dave
Agreed.
Another thought I had was on the video, the third of a second that the device used to create a speed calculation will be reasonably obvious. The ACPO code of practice states that the device should be aimed at the area of the number plate of the car. The third of a second (which may be actually 0.3 best to check first) will equate to 8 frames assuming that the frame rate is 24 frames per second. It would be very interesting to produce 8 zoomed in photos with the crosshairs on 8 different points of the car, especially if the points follow a 'sweep' along the vehicle. If there is a sweep along the vehicle then it would be technically impossible for the device to produce an exact reading.
You would then explain what slip error means, if you are lucky you could bring the Daily Mail report on slip error and give it to the Mags, show that slip error can only have occured. In addition to this if you can get a waypoint to measure the distance travelled *within the third of a second that the device also mismeasured* then you have a watertight case.
Methinks getting the 8 shots would be enough though. Back this up by the fact that this is the reason why the ACPO Code says not to use the device at this distance.
justinp1 said:
The third of a second (which may be actually 0.3 best to check first) will equate to 8 frames assuming that the frame rate is 24 frames per second. It would be very interesting to produce 8 zoomed in photos with the crosshairs on 8 different points of the car, especially if the points follow a 'sweep' along the vehicle. If there is a sweep along the vehicle then it would be technically impossible for the device to produce an exact reading.
All the video clips I have seen show a 50 frame per second count.
LongQ said:
justinp1 said:
The third of a second (which may be actually 0.3 best to check first) will equate to 8 frames assuming that the frame rate is 24 frames per second. It would be very interesting to produce 8 zoomed in photos with the crosshairs on 8 different points of the car, especially if the points follow a 'sweep' along the vehicle. If there is a sweep along the vehicle then it would be technically impossible for the device to produce an exact reading.
All the video clips I have seen show a 50 frame per second count.
That sounds like a timecode as oppose to a frame rate. I think the frame rate of a standard video player is either 24 or 25 frames per second, that would be what you would use to get the stills and the timecoding on each one is the proof of where they have come from.
Gassing Station | Speed, Plod & the Law | Top of Page | What's New | My Stuff