MOTION DETECTION BY THE RANDOMIZED HOUGH TRANSFORM PS version here

The Randomized Hough Transform (RHT) is originally a method for finding global features from an image. These features can be curve segments like line segments, circles, ellipses etc. However, the RHT can also be applied to motion estimation. The new algorithm, called Motion Detection Using the Randomized Hough Transform (MDRHT), can be used for solving machine vision problems like tracking of moving objects.

The MDRHT is an efficient and simple 2-D motion detection method. Moving objects are assumed to be rigid and a number of moving objects to be unknown. Objects can be partially non-rigid or distorted.

There are moving objects in the gray-level image sequence shown below. Consecutive image frames are examined frame by frame. Using edge points of gray-level pictures, translational motion of moving objects is estimated. Motion is detected first, and only after that objects are segmented for tracking and recognition. The result of applying the MDRHT in first two frames is shown in the last picture. Two moving objects are found.

For further information see the following papers on the MDRHT " Motion Detection using Randomized Hough Transform," " Randomized Hough Transform Applied to Translational and Rotational Motion Analysis"," " Motion Estimation and the Randomized Hough Transform (RHT): New Methods with Gradient Information," " Detecting Multiple Moving Objects by the Randomized Hough Transform," " Motion Detection Using the Randomized Hough Transform (RHT): Exploiting Gradient Information and Detecting Multiple Moving Objects," and " Motion Analysis by the RHT Method in Computer Vision," and on the RHT " Comparisons of Probabilistic and Non-probabilistic Hough Transforms" and " Probabilistic and Non-probabilistic Hough Transforms: Overview and Comparisons."

Nature of work: Basic Research     Realization: Heikki Kälviäinen, 1990-1994
                                                Lei Xu, Erkki Oja, 
Partners: Lappeenranta University  Finance: Academy of Finland
          of Technology