Data Fusion

Tracking and Multi-Sensor Fusion

‘Tracks’ are estimates over time of the state of something such as temperature or the speed of an object. Among other techniques, a Kalman filter provides a means of estimating tracks from indirectly related or indistinct ‘noisy’ data.

‘Multi-sensor fusion’, ‘information fusion’ and ‘data fusion’ are terms that are often used interchangeably to describe the combination of data from multiple sources to provide information that is better than could be obtained from the individual sources alone.

Conekt applies tracking and/or multi-sensor fusion techniques to data to realise:

Reduced measurement uncertainty (including improved accuracy, precision, reliability, lower false/missing detection rate) and less measurement noise
Increased measurement coverage (eg through combining sensors with different measurement ranges)
Increased measurement robustness as sensors compensate for each other (eg a radar sensor compensating for a video sensor the performance of which is degraded by poor light or foggy conditions)
Extra measures, such as through exploiting synergies (eg combining data from a sensor that measures distance to an object with one that measures its relative speed to calculate the arrival time)

Conekt has supplied multi-sensor fusion and tracking algorithms using commercial off-the-shelf (COTS) hardware through to bespoke real-time embedded solutions. These embedded solutions often utilise higher specification processors and/or FPGAs (Field Programmable Gate Arrays).

Tracking and multi-sensor fusion techniques are being applied to:

Driver Assistance Systems (DAS) eg multi-sensor fusion for lane change support functions
Intelligent Transport Systems (ITS) where radar vehicle tracking is fused with image information
Simultaneous Localisation and Mapping (SLAM) of environments using image data and motion measurements

For more information about any element of multi-sensor fusion, please contact us.