Pipeline modules¶
Built-in pipeline modules shipped with RD Analytics. Configure them on a scan while status is DRAFT:
| Where in the UI | Module role |
|---|---|
| AI processing settings | Detectors, classifiers, tracker, stabilizer, motion, and other processing slots |
| Geometry editor / Draw settings | Scene modules (counting lines, areas, speed, density, zone dwell) |
| Output settings | What the pipeline writes (SystemDB, video, SQL, webhook) |
NN modules also appear under System → AI-Models. A model must be ACTIVE before its detector/classifier options show in AI processing. Optional modules need a matching license feature.
System → Modules is only for uploading custom partner packages — not this built-in catalog.
Labels below match the English UI. Exact options can vary slightly by release; fields are schema-driven from the server.
AI processing modules¶
Vehicle detection¶
Detects vehicles and related objects on each frame. Choose one model size in the Detector type picker for the Vehicle detection slot.
| Option (type) | What it does | License |
|---|---|---|
Vehicle Detector S (vehicle_detector_s) |
Small model — faster, lower GPU load | — |
Vehicle Detector L (vehicle_detector_l) |
Large model — better accuracy | — |
Vehicle Detector XL (vehicle_detector_xl) |
Extra-large model — highest accuracy / cost | — |
Vehicle Drone M (vehicle_drone_m) |
Balanced detector for drone footage | vehicle_drone |
Vehicle Drone L (vehicle_drone_l) |
Max-accuracy detector for drone footage | vehicle_drone |
| Parameter | Purpose |
|---|---|
| Enabled | Soft on/off for the slot (default on; shown in the slot header) |
| Min confidence (%) | Ignore detections below this score (default 10). Range 5–80 for S/L/XL; 1–80 for drone models |
Shown by default in AI processing settings.
Tracker¶
Associates detections across frames (bounding-box IoU) and keeps stable track IDs. Required — always present; enable toggle is hidden.
| Parameter | Purpose |
|---|---|
| Patience | Missed-frame budget for established tracks (about five successful hits). Internal limit is patience × 2 frames (slider 5–100, default 40 → about 80 frames). Unconfirmed tracks drop sooner. Higher values keep tracks through longer gaps |
| Tracking point | Which point on the box scene geometry uses (line crossings, areas, speed, density, zone dwell): Bottom (default) or Center. Track association itself uses IoU, not this point |
Active area¶
Required region of interest (ROI). Later steps use this polygon; tracks predicted outside it are dropped. Required geometry — draw it on the canvas.
| Parameter | Purpose |
|---|---|
| Active area | Polygon ROI (geometry) |
Stabilizer¶
Digital image stabilization for camera shake. Runs before detection. Uses the region only as a feature lock mask (default mask covers most of the frame). Disable when the camera is rigidly mounted. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Turn stabilizer on or off (default off) |
| Stabilizer | Polygon mask for lock features (geometry) |
Motion detector¶
Background-subtraction motion detector. Useful at night when neural detectors struggle. Emits unclassified object detections. Opt-in via Add. Runs automatically after vehicle detection (fixed pipeline order — you do not reorder slots).
| Parameter | Purpose |
|---|---|
| Enabled | Turn motion detection on or off (default off) |
| Sensitivity | Standard (default) — bright foreground only (for example headlights at night); High — all motion. Caps at two blobs per frame and drops motion boxes that overlap NN detections |
Vehicle classes¶
Classifies tracks that currently have a detection in {person, motorbike, car, truck, bus, bicycle} into fine-grained classes (for example car, van, pickup, truck, bus, bicycle, person, and related labels). License: vehicle_classes. Shown by default when licensed.
| Option (type) | What it does |
|---|---|
Vehicle Classes S (vehicle_classes_s) |
Small classifier |
Vehicle Classes M (vehicle_classes_m) |
Medium classifier |
| Parameter | Purpose |
|---|---|
| Enabled | Soft on/off for the slot (default on) |
No other per-scan fields — pick the classifier type in the slot. Class names feed class groups and reports.
Vehicle make/model¶
Recognizes make and model for car, truck, and bus tracks that have a current detection. License: vehicle_cmm. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Turn make/model recognition on or off (default off) |
| Min confidence (%) | Minimum score to accept a make/model result (10–80, default 30) |
License plate detector¶
Detects plate geometry inside an ROI and attaches plate quads to vehicle tracks (skips person / bicycle / shadow). License: anpr. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Turn plate detection on or off (default off) |
| LP detection area | Polygon ROI where plates are searched (geometry) |
| Plate normal size | Reference plate size rectangle used to scale the search ROI (geometry) |
| Min confidence (%) | Minimum plate detection score (1–100, default 30) |
| Focus on black plates | Bias detection toward dark plates (default off) |
License plate recognition¶
Reads plate text on confirmed vehicle tracks that already have a plate contour. License: anpr. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Turn recognition on or off (default off) |
System-wide (not per scan): region profile — JSON OCR/region ruleset on System → AI-Models → plate recognition model Parameters (edit or upload). Default is a general profile; shipped profiles include regional files (for example US, UK, EU, BR, CL, CO, ID, SA).
Multi-camera tracking¶
Tracks vehicles across cameras using crossing lines and re-ID embeddings. License: multicam_track. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Turn multi-camera tracking on or off (default off) |
| Line to other cameras | Directed line for traffic leaving toward other cameras (geometry) |
| Flip arrow | Reverse the “to others” direction |
| Line from other cameras | Directed line for traffic arriving from other cameras (geometry) |
| Flip arrow | Reverse the “from others” direction |
Scene / geometry modules¶
Add and edit these under Draw settings. You can create multiple instances of each type (for example several counting lines). Containment and crossings use each track’s tracking point (see Tracker).
Blind area¶
Excludes detections inside user-defined polygons (for example trees, UI overlays, or irrelevant regions).
| Parameter | Purpose |
|---|---|
| Blind area | Exclusion polygon (geometry) |
| Name | Label for this instance (default Blind area) |
Counting line¶
Counts objects that cross a line, with direction.
| Parameter | Purpose |
|---|---|
| Counting line | Line with arrow showing count direction (geometry) |
| Name | Line label (default A) — used in reports and the Lines widget |
| Flip arrow | Reverse counting direction |
| Skip acute crossings | Ignore shallow/acute angle crossings that are often noise (default off) |
Speed measurement¶
Measures speed for tracks that cross both reference lines, using the configured real-world distance. Results are stored in speed bins for reporting.
| Parameter | Purpose |
|---|---|
| Line 1 / Line 2 | Reference lines (geometry) |
| Name | Section label (default Speed section) |
| Distance between lines (m) | Real-world distance used for speed (0.1–100, default 10) |
| Speed bin (km/h) | Histogram bin size: 5, 10 (default), or 20 km/h |
| Sampling rate | Aggregate interval: 30 sec, 1 min (default), or 5 min |
Density¶
Counts tracks whose tracking point lies inside a polygon, on a fixed sampling interval.
| Parameter | Purpose |
|---|---|
| Density area | Polygon (geometry) |
| Name | Area label (default Density) |
| Sampling rate | 30 sec, 1 min (default), or 5 min |
| Display counts | Show live counts on the preview (default on) |
Zone dwell¶
Measures how long tracked objects stay inside a polygon and emits exit events.
| Parameter | Purpose |
|---|---|
| Zone dwell | Polygon (geometry) — not an organizational Zone |
| Name | Zone label (default Zone Dwell) |
| Minimum dwell time (sec) | Ignore stays shorter than this (0–3600, default 1) |
| Sampling rate | 30 sec, 1 min (default), or 5 min |
Output modules¶
Configure under Output settings. Outputs run on a processing run (not preview-only). Empty Objects filter means all classes.
SystemDB¶
Persists metrics, tracks, and track events for reports. Shown by default.
| Parameter | Purpose |
|---|---|
| Enabled | Write to the system database (default on) |
| Recording mode | Metadata only — events/metrics without images; Object snapshot (default) — crop/snapshot of the object; Full frame — store full frames |
Video Output¶
Records processed video with scan drawings and detection overlays at a fixed 30 FPS output rate. License: video_output. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Record output video (default off) |
| Playback speed | Timebase of the file: Original (default), 2×, 3×, or 5× |
| Resolution | Original (default), Half (1/2), or Quarter (1/4) |
SQL DB¶
Streams recognition events to an external MySQL database. License: sql_db. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Send events to MySQL (default off) |
| Host / Port | Database server (default 127.0.0.1:3306) |
| Username / Password | DB credentials (username default root; password empty by default) |
| Database / Table | Target database (empty by default) and table (default output_data) |
| Fields map | Optional extra columns: one source > column TYPE mapping per line (# starts a comment). Empty = standard columns only |
| Read/write timeout (sec) | Per-operation timeout (1–1000, default 15) |
| Stop after timeout (sec) | Stop the scan after prolonged DB failures (0 = do not stop; default 120) |
| Exclude events | Event names/types to skip |
| Objects filter | Limit which object classes are sent (empty = all) |
Webhook¶
Posts recognition events as JSON to an HTTP(S) endpoint. License: webhook. Opt-in via Add.
| Parameter | Purpose |
|---|---|
| Enabled | Post events (default off) |
| Scheme | HTTP (default) or HTTPS |
| Host / Port / Endpoint | Target URL parts (defaults 127.0.0.1, port 80, endpoint /endpoint-1) |
| Auth header name / Auth token | Optional auth header (default header name Authorization; token empty) |
| Read/write timeout (sec) | Request timeout (default 15) |
| Stop after timeout (sec) | Stop the scan after prolonged failures (default 0 = do not stop) |
| Include image URL | Attach image URL in the payload when available (default off) |
| Include track | Include track details in the payload (default off) |
| Event name | Optional exact event-name filter (empty = all events). Does not rename events |
| Objects filter | Limit which object classes are sent (empty = all) |
License features (quick map)¶
| Feature flag | Modules |
|---|---|
| (none) | Vehicle Detector S/L/XL, Tracker, Active area, Stabilizer, Motion detector, Blind area, Counting line, Speed measurement, Density, Zone dwell, SystemDB |
vehicle_drone |
Vehicle Drone M/L |
vehicle_classes |
Vehicle Classes S/M |
vehicle_cmm |
Vehicle Make/Model |
anpr |
License plate detector + recognition |
multicam_track |
Multi-camera tracking |
video_output |
Video Output |
sql_db |
SQL DB |
webhook |
Webhook |
Related pages¶
- AI processing settings
- Geometry editor
- Output settings
- AI-Models
- Licensing
- Modules — custom uploads only
- Class groups