Attention shake siamese network with auxiliary relocation branch for visual object tracking

A Siamese tracker with "dynamic-static" dual-template fusion and dynamic template adaptive update

In recent years, visual tracking algorithms based on Siamese networks have attracted attention for their desirable balance between speed and accuracy. The performance of such tracking methods relies heavily on target templates. Static templates cannot cope with the adverse effects of target appearance change. The dynamic template method, with a template update mechanism, can adapt to the change in target appearance well, but it also causes new problems, which may lead the template to be polluted by noise. Based on the DaSiamRPN and UpdateNet template update networks, a Siamese tracker with "dynamic-static" dual-template fusion and dynamic template adaptive update is proposed in this paper. The new method combines a static template and a dynamic template that is updated in real time for object tracking. An adaptive update strategy was adopted when updating the dynamic template, which can not only help adjust to the changes in the object appearance, but also suppress the adverse effects of noise interference and contamination of the template. The experimental results showed that the robustness and EAO of the proposed method were 23% and 9.0% higher than those of the basic algorithm on the VOT2016 dataset, respectively, and that the precision and success were increased by 0.8 and 0.4% on the OTB100 dataset, respectively. The most comprehensive real-time tracking performance was obtained for the above two large public datasets.

Introducing Depth Information Into Generative Target Tracking

Common visual features used in target tracking, including colour and grayscale, are prone to failure in a confusingly similar-looking background. As the technology of three-dimensional visual information acquisition has gradually gained ground in recent years, the conditions for the wide use of depth information in target tracking has been made available. This study focuses on discussing the possible ways to introduce depth information into the generative target tracking methods based on a kernel density estimation as well as the performance of different methods of introduction, thereby providing a reference for the use of depth information in actual target tracking systems. First, an analysis of the mean-shift technical framework, a typical algorithm used for generative target tracking, is described, and four methods of introducing the depth information are proposed, i.e., the thresholding of the data source, thresholding of the density distribution of the dataset applied, weighting of the data source, and weighting of the density distribution of the dataset. Details of an experimental study conducted to evaluate the validity, characteristics, and advantages of each method are then described. The experimental results showed that the four methods can improve the validity of the basic method to a certain extent and meet the requirements of real-time target tracking in a confusingly similar background. The method of weighting the density distribution of the dataset, into which depth information is introduced, is the prime choice in engineering practise because it delivers an excellent comprehensive performance and the highest level of accuracy, whereas methods such as the thresholding of both the data sources and the density distribution of the dataset are less time-consuming. The performance in comparison with that of a state-of-the-art tracker further verifies the practicality of the proposed approach. Finally, the research results also provide a reference for improvements in other target tracking methods in which depth information can be introduced.