Breaking Down Video-Based Conflict Monitoring

Surrogate Safety Measures

The Highway Safety Manual tells us that the safety performance associated with a site refers to the number of crashes (by severity) expected to occur within the site boundaries during a unit of time. However, as mentioned above, we don’t want to wait for crashes to recognize where the safety issues are located. In some cases, the roadway or facility is not yet in service, has only been in service for a short time, or crash frequencies have not been collected (it typically requires years of crash data to uncover reliable patterns). Additionally, while people walking and biking are overrepresented in crash data, at the intersection level, crashes are still rare enough to make drawing statistically meaningful conclusions difficult. This is where surrogate safety measures come in.

A surrogate safety measure is an indirect way of measuring roadway safety performance. While not a one-to-one replacement for measuring crashes, surrogate measures can help us assess safety performance through measuring things that correlate to crashes. Surrogate safety measures include potential conflicts between users, collision avoidance behaviors (such as evasive actions or excessive deceleration), and road user behaviors (such as traffic violations, lane encroachments, or delayed reaction time). The degree to which a surrogate measure provides a reliable indication of safety performance can vary from measure to measure. The stronger the correlation between the surrogate measure and crash frequency or severity, the more reliable of a measure it is.

Conflicts as a Surrogate for Crashes

Conflict monitoring is built on the understanding that higher concentrations of conflicts indicate a greater likelihood of crashes if left unresolved. As the name suggests, video-based conflict monitoring uses cameras to record road user interactions at intersections, and software to flag potential conflicts, or near misses, within the footage.

The effectiveness of measuring conflicts as a surrogate for crashes has been supported by studies like PennDOT’s Smart Intersections study, in which engineers selected 15 intersections across Pennsylvania and manually reviewed 100 events at each intersection that were flagged by video-based software. They compared the events they confirmed to be conflicts against five years of crash data and found a strong correlation.

So, how are conflicts measured? Here are a few of the surrogate safety measures commonly used:

• Time to collision (TTC) predicts the time remaining before a crash would occur if the paths and speeds of two road users are maintained. A smaller value indicates a higher likelihood of a collision. This measure can be used for all road users.

• Post encroachment time (PET) measures the lapsed time from when a road user leaves a specific location, or conflict point, to when another user arrives at that location. A smaller value indicates a higher likelihood of a collision. While it can be used for all road users, PET has some limitations, which we’ll discuss later on.

• Delta-v represents the change in vehicle velocity in the event of a collision. This measure is used with other surrogates to add supplemental information about crash severity. A larger value indicates a more severe crash.

• Modified time to collision (MTTC) predicts the time to collision that includes consideration of the likely deceleration of the involved road users. A smaller value indicates a higher likelihood of a collision. MTTC is typically used to evaluate rear-end conflicts.

• Deceleration rate to avoid a crash (DRAC) is the minimum required deceleration rate a vehicle has to apply to avoid a crash. A larger value indicates a higher likelihood of a collision.

Video-based monitoring can also identify other safety issues like red light violations and speed violations, both of which are known to lead to collisions.

When to Use Conflict Monitoring

While video-based conflict monitoring can have a number of use cases, its two primary uses are:

1. Prioritizing improvements, especially around pedestrians and bicyclists or other low-frequency/high-severity crash types. If 70 intersections each have one pedestrian or bicyclist crash documented, how do you prioritize where to install countermeasures at first? Creating a database of comparable counts of conflicts will aid your prioritization.

2. Evaluating the effectiveness of countermeasures. Video analytics can be used to compare conflicts before and after a change was implemented to see if it made a difference. Measuring conflicts rather than crashes in before/after studies can dramatically shrink the timeline needed to complete the study, while also offering a larger dataset to work with.

Keys to Success with Video-Based Conflict Monitoring

When using any surrogate safety measure, you’ll still need to do some post-processing to get the most accurate results. Human review and judgment is required to verify the data that software can pull from video monitoring depicts true conflicts. For example, some encounters—like a pedestrian waiting for a car to pass and then crossing the street, or a permissive left-turn made after a through vehicle passes by—could have a PET of less than two seconds, a common threshold for potential conflicts, while being a legal, safe interaction. Human judgment must always be part of the equation.