This is the second part of a two-part project in which an algorithm system had been set up for detecting and tracking targets in the naval surveillance.
Abstract
This is the second part of a two-part project in which an algorithm system had been set up for detecting and tracking targets in the naval surveillance. The system is intended to facilitate the transformation of the unmanned surface vehicle, the “Protector”, into an autonomic vehicle.
As a pre-process step, before the detection and tracking of the target, we presented in the previous project an algorithm for detecting and aligning the horizon in the panoramic video obtained from the “Protector’s” cameras.
In this work, we present an algorithm for detecting and tracking targets in the naval surveillance. The input to the algorithm is the frames with the aligned horizon (i.e. the output of the first project) and its output is those frames with markings around the targets in them.
The algorithm has been tested on a wide range of datasets including the panoramic video obtained from the “Protector” and videos taken from the internet. The results show a significant outperformance compared to results achieved in previous works that dealt with the detection and tracking issue in a naval surveillance.In this work we managed to deal with situations that previous works couldn’t dealt with such as tracking after static targets(buoy for an example) and tracking after several targets at the same time.
Finding the horizon
Results
