A. Nevin, A.S. Hodel, and D.M. Bevly (USA)
object registration, Hough transform, signal and image pro cessing, robotics
We investigate different methods to detect objects in a sequence of images with digital image processing tech niques, and propose a registration method that combines local statistics with gradients to identify and track objects from frame to frame. These methods are evaluated against both synthetic and captured video data. Common algo rithms such as edge detection, Hough transform, and nor malized cross correlation work well on synthetic data, but appear to be unreliable with data acquired from a digital camera. Our proposed registration method is shown to be effective with both synthetic data and acquired data.
Important Links:
Go Back
