Video Synopsis

Video Synopsis condenses hours of recorded footage into a short interactive summary by extracting moving objects and replaying them simultaneously on a static background. An operator can review an entire day’s activity in minutes rather than hours.

How It Works

The synopsis pipeline processes recorded footage in three stages:

1. Background extraction — Background subtraction algorithms build a clean static background from the footage.
2. Object tracking — A centroid-based tracker identifies and follows individual moving objects across frames, assigning each a unique track ID.
3. Track assembly — Each tracked object’s trajectory is extracted as an independent clip that can be composited onto the background at any point in time.

The result is a compressed timeline where all detected movements are visible simultaneously or can be filtered interactively.

Note

Synopsis uses efficient background subtraction and object tracking — no GPU is required. It runs efficiently on CPU, making it available on recording servers without dedicated graphics hardware.

Interactive Player

The synopsis player uses an HTML5 Canvas renderer that composites tracked objects onto the background in real time. Unlike traditional video playback, the player provides full interactive control over which objects are displayed.

Player Controls

Control	Description
Play/Pause	Start or stop the synopsis playback
Speed	Adjust playback speed (0.5x to 8x)
Scrub	Drag the timeline to any point in the synopsis
Object click	Click any moving object to isolate it and view its full track
Filter panel	Show/hide objects based on analyzer criteria

Object Interaction

Clicking an object in the synopsis view:

• Highlights its complete trajectory path on the background
• Shows metadata (class, duration, speed, direction, zones visited)
• Provides a “Jump to Recording” link to view the original footage at the exact timestamp
• Isolates the object track for focused replay

Analyzers

Synopsis includes 13 built-in analyzers that classify each tracked object. These analyzers run automatically during synopsis generation and their results power the forensic search filters.

Analyzer	Description
Speed	Estimated speed of the object (slow, walking, running, vehicle speed)
Direction	Primary direction of travel (N, NE, E, SE, S, SW, W, NW)
Color	Dominant color of the object (extracted from the object crop)
Size	Relative size classification (small, medium, large)
Dwell Time	How long the object remained in the scene
Path Length	Total distance traveled within the frame
Zone Entry	Which detection zones the object entered
Zone Exit	Which detection zones the object exited
Start Position	Where in the frame the object first appeared
End Position	Where in the frame the object was last seen
Aspect Ratio	Shape classification (tall/narrow vs. wide/short)
Activity Level	Amount of movement (stationary, slow-moving, active)
Time of Day	When the object appeared (morning, afternoon, evening, night)

Forensic Search

The forensic search interface allows filtering synopsis results using analyzer criteria. This turns hours of footage into a targeted investigation tool.

Filter Examples

“Show all people walking northward between 14:00 and 16:00”

• Direction: N, NE, NW
• Size: Medium (person-sized)
• Time of Day: Afternoon
• Speed: Walking

“Find vehicles that stopped in the loading zone for more than 5 minutes”

• Zone Entry: Loading Zone
• Dwell Time: > 5 minutes
• Size: Large

“Red objects moving east”

• Color: Red
• Direction: E, NE, SE

Combining Filters

Multiple analyzer filters are combined with AND logic. Each added filter narrows the result set. The synopsis player updates in real time to show only objects matching the active filters, with non-matching objects hidden.

Generating a Synopsis

To create a video synopsis:

1. Navigate to Operations > Synopsis.
2. Select a camera from the camera list.
3. Choose a date and time range (maximum 24 hours per synopsis).
4. Click Generate Synopsis.

The synopsis worker processes the footage in the background. Processing time depends on the duration and activity level:

Footage Duration	Typical Processing Time
1 hour	2-5 minutes
4 hours	8-15 minutes
8 hours	15-30 minutes
24 hours	30-60 minutes

Processing progress is shown in real time. Once complete, the interactive player loads automatically.

Synopsis Storage

Generated synopsis data is stored alongside recordings:

/recordings/{server_id}/{camera_id}/synopsis/
  ├── {timestamp}_background.jpg    # Static background image
  ├── {timestamp}_tracks.json       # Track data with analyzer results
  └── {timestamp}_objects/          # Individual object crops per frame

Synopsis data is subject to the same retention policies as recordings. When the source footage is deleted by retention cleanup, the corresponding synopsis data is also removed.

Performance

Synopsis processing is CPU-bound and runs as a separate subprocess to avoid impacting recording or detection workloads.

Resource	Usage
CPU	1-2 cores during processing
RAM	500 MB - 1 GB per active synopsis job
GPU	Not required
Disk I/O	Reads source segments, writes synopsis assets

Multiple synopsis jobs can run concurrently, limited by available CPU cores. The system queues excess requests and processes them in order.

API Reference

Method	Endpoint	Description
POST	/api/synopsis/generate	Start a synopsis generation job
GET	/api/synopsis/jobs	List active and completed synopsis jobs
GET	/api/synopsis/jobs/{id}	Get job status and progress
GET	/api/synopsis/{camera_id}/{timestamp}	Get synopsis data (background + tracks)
DELETE	/api/synopsis/{camera_id}/{timestamp}	Delete a generated synopsis