Use Case Description
A company owns a fleet of vehicles that are driven by its staff. Dash cameras are installed in each vehicle to protect their assets and employees by recording admissible evidence for insurance purposes in the event of an accident. Collisions can be detected from a characteristic spike in accelerometer data that is embedded within the video files. Video files are downloaded automatically from each vehicle when they return to the company depot.
Pub-Sub Component For Collision Monitoring
A subscriber requires prompt notification of any collision events identified from accelerometer data in the video files. Such events should be associated with one or more video files and an interval on the associated video timeline in order to determine details of the event that are captured in the footage. Relevant details include:
- Was this event a collision?
- What was the colliding object?
- Who or what was responsible for the collision?
- What damage may have been caused by the collision?
A time sequence filter use case could identify collision events, though the associated processing overhead may overload a lightweight pub-sub system. An alternative approach would be to include pre- and post-processing steps as follows:
- Download a video file to the archive.
- Pre-process the video file to characterise accelerometer spikes.
- Publish a collision notification for each spike, including:
- Video file name(s)
- Video timeline interval
- Collision severity - spike height
- Collision duration - spike width
- Other details such as vehicle registration, driver’s name, location, time, speed.
- Receive a collision notification.
- Triage published details to determine whether this collision should be investigated.
- Query the archive to access relevant video and timed metadata content for inspection and analysis.
This approach aims to keep the pub-sub messages succinct but relevant by only using timed metadata, and to shift all the heavyweight video file access elsewhere. However, being able to triage notifications is important so that the appropriate response can be actioned quickly. For example:
- Hitting a pothole may trigger a notification that only requires inspection of the vehicle's tyres.
- Being hit by another vehicle while stationary requires an insurance claim against the other driver.
- Hitting another vehicle at speed may require police investigation and a physical examination of the driver.
Use Case Description
A company owns a fleet of vehicles that are driven by its staff. Dash cameras are installed in each vehicle to protect their assets and employees by recording admissible evidence for insurance purposes in the event of an accident. Collisions can be detected from a characteristic spike in accelerometer data that is embedded within the video files. Video files are downloaded automatically from each vehicle when they return to the company depot.
Pub-Sub Component For Collision Monitoring
A subscriber requires prompt notification of any collision events identified from accelerometer data in the video files. Such events should be associated with one or more video files and an interval on the associated video timeline in order to determine details of the event that are captured in the footage. Relevant details include:
A time sequence filter use case could identify collision events, though the associated processing overhead may overload a lightweight pub-sub system. An alternative approach would be to include pre- and post-processing steps as follows:
This approach aims to keep the pub-sub messages succinct but relevant by only using timed metadata, and to shift all the heavyweight video file access elsewhere. However, being able to triage notifications is important so that the appropriate response can be actioned quickly. For example: