Fast Object Tracking Through the Use of Artificial Neural Networks
A key function for an autonomous robot is recognition and tracking of pertinent objects observed through a camera. Real-time interpretation of camera images is critical to a robot’s interaction with the physical world. This talk presents preliminary results in using artificial neural networks (ANN) to examine the pixels of an image. While processing all pixels through ANNs would jeopardize the real-time processing requirements, the accuracy gained facilitates use of algorithms that only need to examine a fraction of the pixels composing an image in order to recognize and track the objects of interest. The test environment and problem statement, description of a methodology that employed ANNs to address the problem, and details of the algorithms that interpret images will be presented. The results show a high degree of success within the domain of the test environment.