An unmanned ship sea area target tracking method based on a TLD algorithm

Abstract

The invention discloses an unmanned ship sea area target tracking method based on a TLD algorithm. According to the tracking method, a TLD algorithm framework is used as the basis of a target trackingalgorithm, the information of a target is executed and maintained through a detection module, a tracking module and a learning module, and an online template is stored, so that the repositioning function of the unmanned ship under the condition that the target in a sea area is lost is realized. The target candidate box generation algorithm uses a Selective Search algorithm based on region namingto generate candidate boxes with fewer numbers and more accurate proportions; the tracking module uses a tracking method based on correlation filtering, so that the tracking effect is improved, and the tracking speed is ensured. The accuracy and robustness of the tracking system are effectively improved under the condition that the illumination and the shape of the sea area target of the unmannedship change greatly.

Description

technical field [0001] The invention relates to the technical field of unmanned ship local obstacle avoidance target tracking technology, in particular to a TLD algorithm-based unmanned ship sea area target tracking method. Background technique [0002] With the development of artificial intelligence technology, network communication, and sensor technology, various technologies in the field of modern unmanned ship research are also becoming more and more perfect. The application fields of fully autonomous unmanned ships continue to expand, with broad application prospects and huge commercial potential. Specifically, it can be applied to the transportation of goods and resources in sea areas, sea surface patrol search and rescue, military surveillance and reconnaissance, marine surveying and mapping and hydrological monitoring, etc. [0003] The TLD (Tracking-Learning-Detection) algorithm was proposed by Zdenka Kalal. Compared with ordinary tracking algorithms, the TLD algor...

AI Technical Summary

Problems solved by technology

The sliding window method is simple and fast, but generates a large number of windows to pass into the detection stage

The detection module consists of three parts cascaded detection, and the large number of candidate frames leads to an increase in the screening time of each layer of the detector

The tracking module of the TLD algorithm uses a simple median optical flow tracking algorithm, and only uniformly selects points as feature points for tracking. The tracking effect is not ideal when the unmanned ship is reflected by the sea surface and the illumination changes greatly.

Images