Configuration
Management of your devices
To access the device configuration page, simply click on the respective device in the "Devices" list view.
Tipp
Use a logical naming scheme for your devices to keep the configuration organized.
Camera settings
Clicking on one of the displayed cameras opens its settings. Once you've configured the camera connection, you'll see a live image.
Camera Configuration
Data transmission
Your devices send events to an MQTT broker whenever a trigger condition is met.
Custom MQTT lets you send events to an MQTT broker that you provide. The event format definition is available here.
- Use the
ssl://prefix for the broker address to enable transport encryption. - Message compression can save up to 50% of the bandwidth used to send events. Ensure that both the broker and the consuming application support decompressing zlib/deflate payloads.
Models & Event Triggers
In the configuration, you can select the model that suits your use case and configure event triggers. For guidance, consult the corresponding section in the solution areas.
Configuration of the model and event triggers
Models
Below you will find a brief description of each model. Different models are recommended depending on the solution area.
Traffic & Parking
- Traffic & Parking (standard): Recognition of 12+1 object classes in the context of traffic and parking.
- Traffic & Parking (Accuracy+): Recognition of 12+1 object classes in the context of traffic and parking. Optimized for higher accuracy at roughly half the processing speed (measured in FPS) compared to “Traffic & Parking (Standard)”.
See the class definition for a detailed description of the classes.
People: Full Body
The entire body of a person is recognized.
Note: When recognizing and tracking people, we do not perform facial recognition at any time. No sensitive personal information is processed.
People: Head
The head of a person is detected. This is advantageous if the entire body is not visible and the distance to the objects is less than 5 meters.
Note: When recognizing and tracking people, we do not perform facial recognition at any time. No sensitive personal information is processed.
Event triggers
Event triggers create an event when a defined condition is met.
Crossing Line (CL)
A Crossing Line (CL) triggers an event as soon as the center point of an object crosses the line.
The CL also logs the direction of travel in which the object crossed the line, distinguishing between In and Out directions. You can swap the In and Out directions at any time. Furthermore, a custom name can be assigned to both directions.
By default, a CL counts each object only once. If every crossing is to be counted, there is an option to activate events for repeated CL crossings, with the caveat that counts are only considered if there's a five-second interval between them.
Crossing-Line options
Speed estimation can be activated as a trigger option for a CL. Configure the physical distance in meters between the lines. For optimal results, use a straight section without curves or inclines.
Region of Interest (RoI)
A Region-of-Interest (ROI) consists of four points that form an area. An ROI triggers an event after a defined time interval or when the occupancy changes from occupied to unoccupied (or vice versa).
| Options | Description |
|---|---|
| ROI type |
|
| Trigger action |
|
Origin-Destination Zone (OD)
An event is triggered when an object's trajectory has moved across at least two OD zones, with the first and last zones of the event being used.
The first zone an object passes through is called the Origin zone, and the last zone the object passes through is called the Destination zone. The event is triggered when the object's sighting ends, for example, when the object moves beyond the edge of the image.
Focus Zone
A Focus Zone defines an image area where object detection for tracks is considered. Object detection is active even outside the Focus Zone, but objects detected outside it aren't used for creating tracks. This means objects are pixelated throughout the entire image, regardless of the Focus Zone.
Classification and license plate recognition (if activated) only occur for objects detected within the defined Focus Zone.
Using Focus Zones allows for targeted system optimization:
- Improved Tracks: The Focus Zone can limit where tracks are formed. For instance, Focus Zones can be defined to exclude the image's border areas, preventing unwanted merging of tracks.
- Improved Classification: The Focus Zone can restrict the area where classification takes place. For example, areas where vehicles are too small for successful classification can be excluded by Focus Zones.
- Increased Performance: Concentrating on essential image sections reduces the computational effort for classification and license plate recognition (if activated) for objects. This positively impacts the accuracy of object tracking, classification, and potentially license plate recognition.
Calibration
Live Calibration
For optimal system calibration during commissioning, Live Calibration provides a real-time visualization of detected and classified objects. In addition, it also displays the triggering of events and rules – including the associated BMC contacts.
Features
- Object Detection & Classification
- Objects detected by the selected model are highlighted with bounding boxes.
- The center point of each object is tracked and visualized as a trajectory.
- To ensure privacy, objects are pixelated.
- Overlays are color-coded according to the object class (e.g., Green = Car, Violet = Truck, Pink = Bicycle). A legend on the right-hand side explains the applied colors and classifications.
- Rules & BMC Contacts
- Configured rules are displayed in an overview in the upper left corner of the screen.
- The following information is shown:
- Rule name
- Time of the last trigger
- Optionally, the associated BMC contact
- Status display:
- Gray = Rule inactive
- Red = Rule active
Track Calibration
The Track Calibration is an important tool for optimal positioning of the Event Triggers: It displays a live camera image and overlays it with the last 100 tracks, where each track visualizes the trajectory of a detected object.
Track Calibration helps to determine:
- How objects and their tracks were classified.
- If tracks are continuous or interrupted, for example by obstructions.
- From what point objects are large enough for classification.
ANPR Calibration
The ANPR calibration is used for visual verification and fine-tuning of automatic license plate recognition. ANPR calibration enables:
- Optimization of device mounting and alignment for optimal automatic license plate recognition.
- Optimal positioning of Crossing-Lines for optimal results.
- Diagnosis of the data quality of license plate acquisition.
Tip
Set the crossing line at the end of the green tracks in the direction of travel.
ANPR overlay description
The most recently recognized license plates (top-left in the image):
- The Crossing-Line name, direction of travel, and time of the last five recognized license plates are displayed.
- The recognized license plate text and image cutout are shown, allowing for manual verification of the results.
Color-coded trajectories in the focus area (center of the image):
- Real-time camera image with color-coded trajectories for vehicle movements and ANPR recognition quality.
- 🟢 Best ANPR: Optimal recognition – license plates were read correctly.
- 🟠 Inconsistent ANPR: Ambiguous or unstable recognition.
- 🔴 License plate not detected: Vehicle detected, but no license plate found.
- 🔵 Object too small for ANPR: Vehicle too far away or too small, ANPR not possible.
- ⚫ Class not relevant for ANPR: Object detected, but the vehicle class is not relevant for ANPR (e.g., pedestrians, bicycles, scooters, trams).
Focus Area: A rectangular detection area with an orange border, within which ANPR recognition occurs.