Filtering of meaningful features of fuzzy hough transform

Abstract

Hough Transform and its modifications are used to find straight lines or simple geometric figures on images. But this cannot prevent us from detecting not interesting objects completely. The paper introduces a novel method of filtering or fusion of straight lines after performing Fuzzy Hough Transform. The analysis of mutual arrangements of lines on the image is given. Possible distances between lines are described. A novel distance based on intersection of lines and borders of an image is introduced. Methods of line grouping and fusion based on some criteria selection are shown. Testing on real and test images of pipes was performed. Developed methods shows good robustness (less than 5% of errors) and performance (0.2–0.4 s. per image).