Browse Prior Art Database

Multiple Moving Objects Detection Using General Hough Transform

IP.com Disclosure Number: IPCOM000062508D
Original Publication Date: 1986-Dec-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Biland, HP: AUTHOR

Abstract

For detecting the speed and direction of objects in images, the following procedure is proposed: a) Take three images I0, I1, and I2 of an image sequence captured at times t0, t0+Wt, and t0+2Wt. Assuming uniform (non-accelerated) movements only, Wt should be sufficiently small, e.g., 40 milliseconds, such that for every object the displacement vector between I0 and I1 is approximately equal to the displacement vector between I1 and I2 . Preprocessing: b) Calculate difference images I01 = I0 - I1 and I12 = I1 - I2 . c) Perform threshold operation on I01 and I12 to obtain binary images T01 and T12 of the moving regions. d) Apply a gradient operator (Sobel or similar) to T01 and T12 to represent regions by boundaries G01 and G12 . Main Process: e) Apply a general Hough transform to G01 and G12 .

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 58% of the total text.

Page 1 of 1

Multiple Moving Objects Detection Using General Hough Transform

For detecting the speed and direction of objects in images, the following procedure is proposed: a) Take three images I0, I1, and I2 of an image sequence captured at times t0, t0+Wt, and t0+2Wt. Assuming uniform (non-accelerated) movements only, Wt should be sufficiently small, e.g., 40 milliseconds, such that for every object the displacement vector between I0 and I1 is approximately equal to the displacement vector between I1 and I2 . Preprocessing: b) Calculate difference images I01 = I0 - I1 and I12 = I1 - I2 .
c) Perform threshold operation on I01 and I12 to obtain binary images T01 and T12 of the moving regions. d) Apply a gradient operator (Sobel or similar) to T01 and T12 to represent regions by boundaries G01 and G12 . Main Process: e) Apply a general Hough transform to G01 and G12 . This is equivalent to performing an autoconvolution except the fact that one of the images G01 and G12 is considered to be the template (the shape that is looked for) while the other is the image where the template has to be found in. Postprocessing: f) Find the local maxima in the accumulator array. This can be done either by applying a cluster-finding algorithm or by doing a hierarchical (pyramidal) thresholding operation, lowering the threshold value step by step ignoring the regions growing around already found maxima. g) After having found peaks, discard pairs of peak values that are symmetrical with respec...