We introduce the channel and spatial reliability concepts to DCF tracking and provide a learning algorithm for its efficient and seamless integration in the filter update and the tracking process. The spatial reliability map adjusts the filter support to the part of the object suitable for tracking. This both allows to enlarge the search region and improves tracking of non-rectangular objects. Reliability scores reflect channel-wise quality of the learned filters and are used as feature weighting coefficients in localization. Using only hand-crafted features, the tracker achieves state-of-the-art performance on VOT 2016, VOT 2015 and OTB100. The tracker implemented in Matlab runs close to real-time on a CPU. The method was ported to C++ after the publication and now runs confidently in realtime. Preliminary version of the tracker was published as a paper at CVPR 2017, a top conference in computer vision.
COBISS.SI-ID: 1537691075
In this paper we address the problem of developing on-line visual tracking algorithms. We present a specialized communication protocol that serves as a bridge between a tracker implementation and utilizing application. It decouples development of algorithms and application, encouraging re-usability. The primary use case is algorithm evaluation where the protocol facilitates complex evaluation scenarios specific for various robotic applications like robotic boats. We present a reference implementation of the protocol that makes it easy to use in several popular programming languages and discuss where the protocol is already used and some usage scenarios that we envision for the future.
COBISS.SI-ID: 1537470147
The paper describes the method of tracing obstacles, based on the 3D point cloud, which is based on the 3D segmentation of the scene to the space above water, a space under water, and a narrow belt under and above the water surface. The water surface is is detected by fitting the plane model. However, the results are compared with the design constraints of the robot vessel and data on the 3D orientation of the vessel as obtained by an inertial sensor. In this way, we can reliably determine whether the algorithms have failed in determining the water surface - in this case we rely on otherwise less accurate data from the inertial sensor. Clusters of points above the surface are detected as obstacles - this concept is not completely new, but it does not work well; so we upgraded it with the original contribution, where specific cloud histogram characteristics ("fingerprints") are calculated from the point clouds of individual obstacles, and their distances in feature space are taken into account when tracking detections between successive time moments. It turns out that the characteristics of false positive detections change rapidly, and the characteristics of the real obstacles change slowly. The result is a significant reduction in the number of false detections, from 1: 6 to 1:10.
COBISS.SI-ID: 1537721027