Do You See What I See?
Pitfalls of Video Analysis in Traffic Collision Cases
This article examines the differences between what a motorist or driver sees and what an in-vehicle video camera records. What a motorist describes after a traffic accident can be rather incomplete in details or incongruent compared with what a forensic video analysis uncovers.
Digital dashboard cameras, or dash cams as they are more commonly known, are present in almost every vehicle. Motorists install these in-vehicle video cameras to record their on-road journeys, and to safeguard the security of their parked vehicles, in hopes that they may provide useful visual evidence in the event of accident, theft or vandalism. A video clip may also be used as objective evidence by a traffic accident reconstructionist to determine how an accident occurred, and whether it could have been avoided.
A dash cam is usually mounted in an unobstructed position between the windscreen and the rear-view mirror to record visible events happening in its field of view. This is usually directly in front of the vehicle. Typical dash cams record 25 or 30 frames (or images) per second (fps). Some dash cams can even achieve 60 fps. Dash cams today usually have a resolution of 1920 x 1080 pixels, which is adequate for after-the-fact forensic reconstructions.
It is a common belief that a video footage is irrefutable evidence – a “silent witness”, a second pair of eyes and a proxy for the driver’s perception. However, does it really mirror the accident through the eyes of the motorist? The short answer is “No”. Although on the surface, there are similarities between cameras and the human eye, such as the use of an aperture, lens and light-sensitive sensors or receptors, the human vision is not like a photograph or a video stream. Due to the differences between human vision and video recordings, there are pitfalls that need to be avoided when interpreting and applying video evidence in the analysis of traffic collisions.
The Forensic Experts Group would like to share our insights on some of these pitfalls, based on the traffic collision cases we have encountered in recent years.
Video Recording is Passive Whereas Driving Entails Sustained Attention and Appropriate Responses to Dynamic Traffic Situations
Video recording is passive, automatic, non-selective and continuous. A dash-cam functions independently of changing traffic conditions and potential road conflicts. Each video frame is an all-inclusive, all-at-once capture of the entire field of view of the camera, with no discrimination of details. Video cameras often capture details that elude the driver’s eyes.
In contrast, driving is a complex human task in a spatio-temporal context. It is a dynamic, interactive and multi-tasking activity that requires sustained human effort and attention. Driving imposes concurrent visual, manual and cognitive workloads on the motorist. He faces competing demands on the road, and unexpected traffic conflicts.
Human vision during driving requires the timely extraction of salient visual information in a constantly changing traffic situation. The driver’s visual attention has to be judiciously divided between many targets: a motorist cannot look at all directions and objects at the same time; nor should his vision be fixated too long in a single direction. While searching for salient objects or activities, he may miss out on critical details. The driver is constantly deciding which on-road object is important.
Whether a motorist perceives warning cues of an impending traffic conflict may depend on whether he was looking at the pertinent area at the crucial moment. Without timely perception and comprehension at the decisive moment, he would not be able to respond appropriately to avert the accident.
The Dash Cam has a Different Field of View and Perspective from the Driver
All that is visible in the video footage may not necessarily be visible or apparent to the motorist’s perception in real-time. In comparison with dash cams which typically record a 120 to 170 degrees sector with uniform resolution across the field of view, the driver’s binocular field of view is much narrower, and resolution in the human eye is not uniform over the different regions of the retina. Although the fovea is capable of high-resolution colour vision, it is a small region on the retina that detects only the central 3 to 5 degrees sector of the visual field. Extending beyond the fovea to only 20 to 30 degrees sector on either side of an observer’s line of gaze is the near peripheral part of the retina that detects contrast and motion but with poor resolution, and in black and white (ie, monochrome). To compensate for these limitations, eye movements are necessary for the eye to track and focus on specific objects in the visual field.
The perspectives of the dash cam and the driver are also slightly different due to their different locations. The dash cam is often centrally mounted near the rear-view mirror, and has a less obstructed view than the driver’s eyes which are located further back and to one side of the vehicle cabin. The driver’s front side view is partially occluded (ie, visually obstructed) by the A-pillars.
A Video Camera can Reliably Store and Retrieve a Far Larger Amount of Visual Data
A dash cam captures and stores much more visual information than the driver. Current Secure Digital Memory cards (SD cards) have memory sizes ranging from 32 to 512 gigabytes. A key advantage of the dash cam is the permanence and stability of electronically stored visual data. So long as the stored file is extracted and transferred onto another device soon after the accident, the data can be preserved indefinitely and reproduced accurately without alteration.
Unlike dash cams, a driver’s ability to capture and process information in a visual scene is rather limited. Human recollection of what was perceived just before and during an accident is unstable and can be forgotten, contaminated or distorted by trauma, post-accident information, reasoning and false memories.
Video Processing Time is Different from Human Perception Time in Nature and in Speed
A video camera functioning at 30 fps records a full image every 0.033 second. Although the camera captures and digitises visual information very rapidly, it does not perform driving tasks such as object recognition, interpretation or decision-making with regard to traffic conditions. It must be emphasised that video recording and forensic video analysis are two distinct and separate operations; the latter is performed by the reconstructionist after the accident.
As opposed to only recording by the video camera, the driver’s vision entails processing by the brain to perceive and interpret the visual information. Human vision is inseparably integrated with processing and recognition by the brain. The eyes take in visual information and transmit it to the brain which then makes sense of the information. Mental processing time is much slower than video capture time.
The amount of available visual information in traffic situations is often beyond the processing capacity of the brain. For instance, the blink of the human eye takes about 0.25 second. During this brief interval, a 30 fps video camera would have recorded seven images of events, and a 50 km/h car would have travelled 3.5 metres. A gap of 0.5 metres in space can make a difference between collision and a near miss. Hence, in the blink of an eye, the driver could very well miss critical details captured by a dash cam, or avoid a traffic conflict.
Video Time-Stamps are Accurate, Unlike Human Time Estimates
Since accident situations typically develop and occur within a short time interval – usually a few seconds, the time factor is of the essence to forestall any road accident. The video recording with its precise and detailed time recordings will enable the precise determination of time and durations, correlation of time to locations and events, and calculation of speeds.
By contrast, human observation and memory do not come with any precise time marker. Without the aid of external measuring devices, humans are often unreliable and subjective in estimating distance, time and speed.
Scientific Analysis of Traffic Collisions
Traffic accident reconstruction and collision avoidance analysis are complex forensic examinations to be undertaken by suitably qualified experts. There is no doubt that video recordings can be useful evidence in traffic accidents. Owing to the accessible, visual nature of video evidence, laypersons may be tempted to analyse and give their take on the cause of an accident and whether it could have been avoided. Unfortunately, they are often unaware of the various factors and scientific principles that need to be considered, as well as the pitfalls of video analysis. The traffic accident reconstructionist has to interpret the findings holistically, taking into consideration within realistic boundaries, the numerous factors at play and the limitations of the scientific findings.
A Video Recording can be Played Back Repeatedly Whereas the Driver is Constrained to a One-Time, Real-Time Perception of the Situation
Dash cam video files play back actual events exactly as captured; the human eye is unable to do this. During video analysis, the reconstructionist can repeatedly play a video recording in slow motion, back and forth, and frame by frame, to identify events or details of interest.
In contrast, the motorist’s perception and reaction to the traffic situation must be on the fly. He experiences the entire critical incident only once in real-time, briefly, in a matter of seconds. Details that the motorist recollects of the incident would be limited to his observation and memory in that short time interval. Therefore, something seemingly obvious to someone with the benefit of viewing the video many times may not have been noticed or correctly interpreted by the motorist at all.
The Driver Approaches a Traffic Hazard with Foresight, Whereas the Post-Accident Observer Analyses the Traffic Collision with Hindsight
A motorist’s primary focus is the prevailing traffic condition and the road ahead. He must maintain situational awareness of ambient traffic, exercise foresight, and oftentimes make correct choices despite incomplete information and cues. Foresight is recognising the current spatio-temporal context and predicting future events in the absence of outcome information. Even a careful driver may not anticipate an accident; thus is not alerted or primed to look out for relevant details in the seconds preceding a collision.
As opposed to the motorist, the reconstructionist or post-accident observer knows that the accident occurred at a particular time and location. He will examine the video footage and physical evidence for all relevant information on circumstances leading to the accident.
Hindsight is the ability of a post-incident observer to perceive the significance, causes and nature of events after they have occurred. It is realising the consequences of events after they have been analysed, although they were not understood or predictable when the accident occurred. The inferences of the observer can be unwittingly influenced by outcome knowledge. This hindsight bias tends to make the path to failure obvious and avoidable to the observer, although it was not foreseeable to the motorist at the time of the accident. Hence, although findings from video analysis are of great use in collision reconstruction, they must always be cautiously applied to collision avoidance analysis.
Collision Analysis Reports
A reconstructionist can uncover from the video recording a large amount of information on vehicle location, time, path, distance, speed, events and manoeuvres. However, much of this information may not be evident to the motorist in the brief moments leading to the traffic collision.
The responsible reconstructionist must constantly bear in mind that his expert opinion has an impact on the case outcome. In report-writing and court testimony, he has to recognise and account for the salient differences between what the video records and what the motorist could possibly see in reality.
When undertaken by a qualified forensic scientist, video analysis is indisputably a powerful tool for traffic accident reconstruction as it enables the accurate determination of time and duration, correlation of time to locations and events, and calculation of speeds.
On the other hand, collision avoidance analysis is particularly susceptible to hindsight bias developed from the repeated viewing and detailed analysis of a video clip. Applying incorrect inferences on what the motorist could have seen can result in errors in collision avoidance analysis. It is dangerous to assume that a dash cam recording is close to what a driver would or should see in the moments leading to a traffic collision. There may be cases where it is unrealistic and unreasonable to infer from a video clip that the motorist ought to have perceived warning cues and therefore was remiss in failing to avert a collision. The driver’s judgment and decision at the time of an accident was based on less certain and less complete knowledge than that derived from analysis of the video recording. Hence, caution is imperative in collision avoidance analysis.
The traffic collision reconstructionist has an overriding duty to the Court. It is incumbent on the reconstructionist to consider within realistic boundaries the limitations of his analysis and interpretation, and the impact of numerous factors at play in the accident.
The Forensics Experts Group
E-mail: [email protected]
The Forensic Experts Group is Singapore’s first one-stop private and independent provider of forensic consultancy, analysis, research, training and education for legal and law enforcement agencies, forensic and tertiary institutions, and private organisations. It comprises a team of accomplished and innovative forensic scientists, who are combining 75 years of specialised knowledge, unique experience and skillsets to deliver high quality forensic services both locally and overseas. While leading the Forensic Chemistry & Physics Laboratory at the Health Sciences Authority, TFEG’s senior consultants developed traffic crash reconstruction capabilities from the year 2005 onwards. TFEG has contributed a series of forensic science related articles to the Law Gazette since 2015, and has published a book Forensic Science – Briefs for the Legal Practitioner in 2017. This book was written to aid legal practitioners in navigating key concepts, scientific terminologies, methodology, and challenges related to various forensic disciplines. It is our hope that the book will serve as a bridge and platform between the legal fraternity and the scientific community.