Artículo de revista
A Delay-Free and Robust Object Tracking Approach for Robotics Applications
Fecha
2019Registro en:
Journal of Intelligent and Robotic Systems: Theory and Applications, Volumen 95, Issue 1, 2019, Pages 99-117
15730409
09210296
10.1007/s10846-018-0840-6
Autor
Pairo, Wilma
Loncomilla, Patricio
del Solar, Javier Ruiz
Institución
Resumen
© 2018, Springer Science+Business Media B.V., part of Springer Nature.In many robotic applications there is the need for detecting and tracking moving and/or static objects while the robot moves, in order to interact with them. High quality detection methods require considerable computational time when the number of objects to be detected is high, or when operating within dynamic, real-world environments. Then, when an object detection result is available, it is referred to a previous frame and not to the current one. A method for obtaining delay-free detections is introduced in this present article. It consists of projecting a delayed detection onto the current frame by using a set of feature tracks generated by using the KLT (Kanade-Lucas-Tomasi) tracker. The proposed method is shown to improve detection accuracy when the tracked object is moving with respect to the camera. In addition, the method is able to detect and manage false detections and occlusions using statistical classifi