ViSense CrowdDynamics v1.5.0
This document provides an overview of the changes for ViSense CrowdDynamics v1.5.0, released on January 19th, 2017.
The following sections describe the most notable changes for this release (aside from bugfixes and minor improvements). For the complete changelog, see below.
New pedestrian detection subsystem
As ViNotion, we are continuously improving our core video content analysis technology. Our research team has made improvements to the pedestrian detection algorithms, resulting in an improved detection subsystem for ViSense CrowdDynamics. This significantly improves ViSense analysis performance with up to 5% more accurate pedestrians detection. The system also provides more support for different viewing angles. More specifically, the difference between the counting directions is more balanced to create better insights in crowd dynamics (front-view vs. back-views).
ViNotion will always continue to improve their core video analysis subsystem, so that the analysis performance of ViSense in various scenarios increases, without updating any physical system components.
New: revert settings to default values
The most requested feature for ViSense is the possibility to reset the settings to their default values, without reverting the system to the previous firmware or performing a factory reset. From ViSense CrowdDynamics 1.5.0 onwards, you can revert the settings to the default values corresponding to the recent firmware version. Camera configuration, calibrations, counting rules and other settings will be defaulted, while keeping your passwords and network settings.
On the web-based user interface, you can revert the settings to default inside the ‘Update’ menu in the admin settings panel. Please see the picture on the next page for details:
Improving speed of the counting statistics
The core value of ViSense is to provide insight in the dynamics of the movement of people, by supplying 24/7 counting statistics over a long period. Previously, when viewing the counting statistics graph over a long period of time, the loading of the graph would take longer the more data you requested. By implementing new counting data aggregation, the counting graph loading times have been significantly decreased. So now, your requested statistics information appears more quickly on your screen.
Special message for system Integrators: cumulative-count API
The implementation of the counting data aggregation is supplied for system integrators with a cumulative-count API. The cumulative-count API provides server-side count accumulation while keeping the flexibility of the regular count data API. If you want to extract the accumulated counting data from the API, you can connect to the following websockets:
ws://xxx.xxx.xxx.xxx/api/v1/data/main/count/cumulative/preset0
ws://xxx.xxx.xxx.xxx/api/v1/data/main/count/cumulative/preset1
You can set the requested time-frame by using the parameters TimeStart and TimeStop, and define the interval between count packages by setting parameter CountInterval. Please note that the TimeStart and TimeStop parameters are set differently with respect to the existing count API. Please see this section of the API documentation for details, or view the documentation on your ViSense box at:
http://xxx.xxx.xxx.xxx/getting-started.html
Suggestions?
Of course we are always busy improving our current system and documentation. We are always interested in suggestions to improve ViSense, thus if you have any remarks or tips, don’t hesitate to contact us.
Complete changelog
New features
- Improved pedestrian detection subsystem.
- Revert settings-to-default option.
- Faster loading of counting statistics.
- Cumulative-count API.
Improvements
- Add JPEG API documentation (incl. example HTML page).
- Add extra visualization options in the administrator menu.
Bug fixes
- Default count threshold set to 20% in parameter template.
- Density description metadata added to documentation.
- Improved error handling in the log.