Advanced video algorithms for the security industry
The Research & Development of artifical intelligence tools is done by the internal team with the collaboration of university searchers. VideoWave continuously develops new video analytic algorithms to be included in the OneTrack system.
Video Analytic Tools
- PTZ Video Tracking: Consist of detecting an object on a fixed camera and tracking and zooming on the object with a PTZ camera.
- Object detection and classification: Based on shape, size motion, speed of motion and other factors, objects detected can be classified under different categories, thus enabling more focused and accurate reaction and tracking.
- Rejection of nuisance objects: Many types of video phenomenon are easy to confuse with real significant objects, from a detection stand point. For example, trees , flags, water poddle or waves. Based on shape, size, texture and degree of instability, many of these “false objects” can be automatically rejected and thus many false alarms avoided.
- Object counting: This is to count the number of occurrences of a category of objects in a specific area.
- Directional detection: The detection of an object motion in a specific direction.
At the core of the technological innovation is the vision module. The task of the vision module is to locate significant objects in the image and to track them on several images. From the information extracted from the image, much as position, speed and category of each object (pedestrian or vehicle), the system can make decisions on the class of the objects and cations on like zooming on the pedestrian’s face or the license plate of the car.
Examples of detection, tracking and object recognition in a wide-angle video of a parking. The pedestrians paths are in yellow and the vehicules paths are in red.
Remote communication with balanced channels
In a typical remote surveillance configuration, a DSL link is used with a maximum upstream bit rate of approximately 700kbps. Within this total bit rate, up to 8 or 16 video channels must be transmitted. To transmit quality video for 16 channels over much a bandwidth is a major challenge.
Sophisticated algorithms have been developed to best manage these limitations in the context of the OneTrack product. A compression mode enable to encode in a fixed stream size per channel. In this case, the codec adjust dynamically the encoding compression on a frame by frame based in order to maintain the bitrate within a configured setting.
Moreover, additional strategies are used to “balance” the various videostreams of a OneTrack system such that the global bandwidth used is maintained within the set limit, and at the same time, the video of each channel in maintained at an optimal possible quality.
In order to do so, the system distributes portions of the bandwidth to the various channels according to the amount needed for each one to maintain good quality. This bandwidth allocation is extremely important since bandwidth requirement for a channel can change by a factor of 10 or mode depending on the action in the scene. In an extreme case of low bandwidth requirement, we have a static scene with no activity. Such a scene can be transmitted as an stream of a few kbps. In the opposite extreme, we have a PTZ camera in reaction. Such case requires typically as much as MJPEG encoding.
In the balanced mode bit rate, the OneTrack controller re-allocate cintinuously the bandwidth allocation of each channel based on the knowledge of the level of activity of each camera.
These characteristics are particularly fruitful in contexts of remote surveillance, specially with remote controls.