Downloads: 109 | Views: 202
Research Paper | Robotics Science | Nigeria | Volume 5 Issue 2, February 2016 | Popularity: 6.2 / 10
Determination of Target Number of Peaks for Automatic Peak Detection Threshold Determination
Gideon Kanji Damaryam
Abstract: Describes a process to automatically detect peaks from the accumulator array of transform space after points from an image have been transformed using the straight line Hough transform. The process works by specifying a target for the number of peaks to be returned from the transform of each image, rather than specify a threshold for detection of peaks as is traditionally done. This was necessary because no threshold was found that efficiently yielded the highest number of peaks that could be found, which matched lines manually judged to be navigationally important lines in the images. This paper presents how that target number of peaks was determined. A few typical images were processed to the point of determining gradients and intercepts of important straight lines from peaks found, and specifying their end-points, hence fully specifying the important lines, using various target numbers of peaks. Navigationally important lines in the images were also manually chosen. Optimal target numbers of peaks were chosen as the ones which matched the highest number of important lines manually chosen, while minimising unnecessary processing of peaks and use or storage space.
Keywords: automatic threshold determination, Hough transform, peak detection
Edition: Volume 5 Issue 2, February 2016
Pages: 189 - 196