The museum mode of Intelligent and Essential Video Analytics is an optimized tracking mode to detect whenever someone tries to reach towards and touch a valuable asset like a museum exhibit.
It can also be used in any other application where approaching a sterile zone from a more crowded environment shall be detected.
Museum Mode FW 6.30:
This article describes what the Intelligent Video Analytics and Essential Video Analytics tracking mode “museum mode” is, where to use it, how to configure and what the best practices are.
What is “museum mode”? For perimeter protection, Intelligent Video Analytics and Essential Video Analytics detect and track objects. Alarms are typically generated for well-separated objects based on alarm fields or line crossings a line based on the object’s center.
Sometimes, however, an alarm is needed as soon as any part of the object, like a hand, comes into the alarm area. Combine this with a sterile zone directly besides an area where lots of people move around, and then this is when you need the museum mode.
The name is derived from one of the applications: Alarming whenever a museum visitor comes too close to an exhibit.
Typical „museum mode“ application: A sterile zone right in front and including the picture is next to an area where many people are expected. All areas outside the alarm field are masked out:
- Protect paintings and exhibits - Protect barriers against climbing - Protect sterile zones right next to crowded places - … and whenever you wonder why a person reaching with a hand into an alarm field does not trigger the alarm
Common product platform (CPP) Bosch cameras can be clustered by their common product platform. As different platforms offer a different amount of processing power, this can make differences in the video content analysis (VCA) performance. For an overview of the different product platforms and the cameras belonging to them, see the tech note on VCA Capabilities per Device.
Limitations General: - Geolocations of tracked objects not available - Difference in performance and configuration for Intelligent Video Analytics on CPP4 & Essential Video Analytics on the one hand and Intelligent Video Analytics on CPP6/7 on the other. Best performance available with Intelligent Video Analytics on CPP6/7.
Intelligent Video Analytics: - Not available on the following Bosch IVA cameras with FW 6.10 – FW 6.20:
o DINION IP starlight 8000 MP o DINION IP ultra 8000 MP o FLEXIDOME IP panoramic 7000 MP
- For FW versions smaller than 6.10, configuration of the museum mode is available but hidden. - Idle / removed objects not available on CPP6/7 - Object color filter not available on CPP6/7
- Shaking camera will not be compensated for
Configuration: Intelligent Video Analytics on CPP 6/7 with FW ≥ 6.30 (1) Set the alarm area for the sterile zone to be protected. (2) Mask out all other areas. (3) Decrease the “debounce time” of the alarm area to get the alarm as fast as possible (4) Set the tracking mode to “museum mode” (5) In case of false alerts, use sensitivity, noise suppression and min object size as filters
Calibration is not required for or used by the museum mode.
Configuration: Intelligent Video Analytics on CPP4 with FW ≥ 6.10 or Essential Analytics with FW ≥ 6.30 (1) Set the alarm area for the sterile zone to be protected. (2) Decrease the “debounce time” of the alarm area to get the alarm as fast as possible (3) Set the intersection trigger to “edge of box”. Thus every part of the object reaching into the alarm field will trigger the alarm. Note that false alarms may occur if the alarm field is sloped, as the object bounding box is used for the evaluation instead of the object shape. To display the object bounding box, right-click on the video image and select “Show”->”Object Bounding Boxes” (4) Set the tracking mode to “museum mode”. This means that the raw segmentation outline will be used in each frame. Though the raw segmentation looks very ragged, this is necessary as the hand of a person is very small in comparison to the person itself, and needs to be preserved in the object outline to correctly alarm on it. Furthermore, close objects will immediately be merged together, as any occlusion handling will be confused by too many objects anyway, and the shapes are estimated / smoothed during occlusion handling which again may cut the hand from the person outline. (5) Disable “noise suppression” for better sensitivity & faster detections. (6) Raise the stopped object debounce time so standing persons will not be regarded as idle and lost to the tracking. In addition to, or even replacing the activation of the museum tracking mode, the area outside the alarm field can be masked out. This reduces the needed processing power and neatly avoids the potential crowded situations outside the sterile zone. If objects reach into the alarm area and out again very fast, then is better to add some distance between the alarm field and the VCA mask to support earlier detection of the objects.
Calibration of the camera is not required for or used by the museum mode.
Intelligent Video Analytics on CPP4 & Essential Video Analytics museum mode: Ragged segmentation outlines but preservation of the hand. Alarming on edge of bounding box:
The field intersection trigger “ edge of box ” means the field alarms whenever any part of the object’s bounding box overlaps with the alarm field:
The video menu where the display of the object bounding boxes can be enabled.
... View more
The tech note in the attachment describes how to set up map displays of objects detected and tracked by Intelligent Video Analytics, Essential Video Analytics or Intelligent Tracking with FW 6.4x and Video Security Client 2.1.
... View more
Object classification takes an object detected and tracked by either Essential or Intelligent Video Analytics, evaluates its geometric and speed properties, and decides whether the object is an upright person, a bike, a car or a truck. The technical note in the attachment provides an overview of object classification, describes how to configure it with FW 6.30, and what the limitations are.
... View more
Never lose track of an object of interest! Intelligent Tracking automatically zooms in to and follows a selected Intelligent Video Analytics object as far as possible with the camera.
For AUTODOME / MIC IP 7000, this means automatically steering the camera to zoom in to the object and follow it.
For a FLEXIDOME IP panoramic 7000, Intelligent Tracking is available in the Virtual PTZ video and steers the dewarped cut-out.
For DINION / FLEXIDOME IP 7000 / 8000 cameras, Intelligent Tracking can steer a region of interest (ROI) stream. This article describes how to set Intelligent Tracking up with FW 6.30 and lists the differences based on the camera types.
AUTODOME / MIC IP 7000: - Automatically get close-up shots of suspects for viewing and recording - Automatically track the suspect even outside the current preposition to see and document what he is doing - Ease of use: Steer the camera for the operator
FLEXIDOME IP panoramic 7000: Ease of use: Steer the virtual PTZ camera for the operator to obtain the optimal and already dewarped video of the object of interest. By recording the full resolution in parallel, no detail is lost.
DINION / FLEXIDOME IP 7000 / 8000: Oftentimes, operators watch several cameras at once on a screen, leaving SD resolution or less for each single camera video. By using Intelligent Tracking to automatically zoom in, a close up of the suspect can automatically be shown in this small resolution, to allow the operator to faster decide on incidents. Furthermore, the motion of the ROI will aid in focusing the attention of the operator. By recording the full resolution in parallel, no detail is lost.
Intelligent Tracking Overview
Intelligent Video Analytics is needed to start the Intelligent Tracking - MOTION+ and Intelligent Video Analytics FLOW cannot start the Intelligent Tracking - Intelligent Tracking may be confused if more than one object moves through the field of view. - Intelligent Tracking may be confused by motion in the background like wind in trees or moving water. If possible, mask out these areas.
AUTODOME / MIC 7000:
o If a suspect is followed with Intelligent Tracking, another suspect may pass through undetected through the areas currently not covered by the cameras field of view. o If the camera needs to significantly zoom in on a faraway suspect, the object may be lost while zooming. o Motion towards the camera, especially with a low perspective, may appear too small to properly follow. o Will not work on water.
Visualization of Intelligent Tracking status
Intelligent Tracking active, waiting for object to follow
Blue object outline
Intelligent Tracking active and following object with blue outline
Blue rhomb at last object position
Intelligent Tracking active, object is lost but still waiting for reappearance
Intelligent Tracking active, object is lost
Comparison: Intelligent tracking on moving cameras (AUTDOME / MIC) and fixed cameras (DINION / FLEXIDOME)
For an explanation of Intelligent Video Analytics technology, see the corresponding tech notes, e.g. Intrusion Detection FW 6.30 and Counting FW 6.30. On AUTODOME / MIC, where the whole camera pan / tilts / zooms, the field of view of the camera changes during Intelligent Tracking. Therefore the base technology “background subtraction” from Intelligent Video Analytics cannot be used.
T he object detection and following is based on clustering motion of the same direction:
... View more