Image: Road Safety Support
Experts have expressed concern over recent publicity about a new smartphone app that claims to gather evidence of speeding.
The SpeedCam Anywhere app is designed to allow pedestrians to use their smartphones to gather evidence of speeding vehicles.
The app takes a short video clip of a passing vehicle which is uploaded to an Artificial Intelligence (AI) server. The app suppliers say the server can ‘measure the vehicle speed from the video, check vehicle speeds against speed limits, and provide a report showing evidence of the vehicle speed, as well as generating road safety reports and identifying speeding hotspots’.
But Road Safety Support, which provides expert forensic video analysis to multiple UK police forces, has expressed concern over the app’s capabilities.
Steve Callaghan, a forensic engineer who is head of Road Safety Support’s ISO 17025 accredited speed calibration laboratory, said: “We think that this concept is based on what is, in part, a sound principle, however we feel that before any approval is given to it or it is used by the police, that a proof-of-concept test is performed to demonstrate the system and its capabilities.
“Currently, there is an apparent gap between the technology and the law. The speed calculation needs to be proven on every occasion because of the way the application assumes that the unknown mobile telephone it is used on is working correctly.
“From my understanding of the measurement technique and the intended deployment on multiple phones, Home Office Type Approval would be challenging.”
Road Safety Support aims to reduce deaths and injuries on the roads through the effective use of accurate and reliable road traffic enforcement technology.
The company’s ISO 17025 accredited speed calibration laboratory, thought to be the only one of its kind worldwide, has been approved by UKAS to test and calibrate speed cameras, speedometers, telematics and other speed measurement devices.
The vast majority of UK police forces, highway authorities and safer roads partnerships, as well as some international forces, are members of Road Safety Support and receive specialist technical and legal advice on a subscription basis.
Meredydd Hughes, chairman of Road Safety Support, said: “New technologies such as Artificial Intelligence offer great potential for increasing road safety through wider enforcement, as well as for more active traffic management generally.
“For the public to accept that prosecutions are fair, they need to know that the equipment is accurate and being used correctly.
“The Home Office Type Approval (HOTA) process acts as this guarantee to the public. Now, in the same way that e-scooters are providing a challenge to the existing rules and regulations of road use, the introduction of new technology is challenging enforcement standards.
“Additionally, the growing use of cameras by ordinary members of the public to record offences and refer them to their local police is great news, but the police need more capacity and the skills to deal with this material effectively and ensure that prosecutions are based on accurate evidence.”
I look forward to RSS being in a position to understand better the breakthrough that this new app represents. Almost all speed detectors have involved a sensor in a device. They are made in relatively small numbers due to their cost which in turn suppresses the market demand.
The Speedcam Anywhere app does not use a radar or laser sensor in the smartphone, but relies on its video camera. This produces a video with a timestamp on every frame. We understand that historically when analysing videos forensic experts have relied upon a single point on the car and measured points on the ground as the basis for calculating the speed using (distance between those point) divided by (difference in timestamps).
This app takes a smart approach by using the fixed distance between the wheels and AI identifying the frames where the centre of the wheel passes a fixed point on the road. This enables the speed to be calculated by (distance between wheels) divided by (difference in timestamps).
It also eliminates the cosine effect in laser and radar devices where the angle from user to vehicle differs from the direction of travel.
Within the calculation there are only three variables that could introduce inaccuracy.
1. The wheelbase for the vehicle. Fixed by manufacturer.
2. The selection of frames chosen when centre of wheel passes a point. Verifiable by review of what is a high definition video.
3. The timestamps on the video frames. Embedded in video where phone implicitly requires a very accurate clock and would not work without such a clock.
And if you look at the calculations you can see that the propensity for inaccuracy is minimal.
The wheelbase of a car is of the order of 2000-3000 mm and Fixed to a high tolerance by manufacturer. Lets say 2mm. That an inaccuracy of < +/- 0.1%
The frames selected are chosen by AI, but verifiable by observing the video frame by frame. Lets say this is wrong by 50mm. Then this introduces an inaccuracy of < +/- 2.5%.
With regard to the accuracy of the timestamp this is expected to be very high. It is not the actual time which is important but the difference between the timestamps. It is important to ask questions regarding how the timestamp is generated and the developers are confident that the timestamps are fully accurate. These timestamps are coded into the video frames and cannot be manipulated by editing.
So in essence can be sure of accuracy with a variation of < 3% which is fully verifiable in both the wheelbase and the frames selected. Hence for recording a 40mph car then that is +/- 1.2mph.
Note that none of the analysis takes place on the phone or in the app. All the app does is to gather the video evidence and upload it to cloud-based servers for analysis using Artificial Intelligence to identify the key values in what is quite a simple (v=s/t) calculation. You can read more about the actual process at https://www.20splenty.org/how_sca_works
20's Plenty and the developers are confident that for Section 59 offences then a measurement of a car at 40mph in a 30mph limit is more than the required "reasonable grounds" for an officer to believe an offence has been committed. Hence this does not require HOTA.
Smartphone Anywhere therefore not only provides a method for the public to upload evidence of speeding to dashcam portals, but also provide any police officer with a smartphone to issue a Section 59 order. This therefore empowers police and should be welcomed as an additional capability in addressing road crime.
I look forward to RSS supporting this new capability in forensic analysis of speeding and eventually granting approval for speeding offences. In the meantime police should be encouraged to use if for Section 59 enforcement.
More information can be found at http://www.20splenty.org/speedcamanywhere
Rod King, Lymm
+2