(c) (d)
Figure 19. Error Results of Hough-kmeans and Improved Hough-kmeans
Algorithms.
Comparing the method proposed in this article with the probabilistic
Hough line detection method and the LSD line detection method, it can be
seen in Figure 21a that the probability Hough line detection has a high
false detection and missed detection rate, and some unrelated background
lines have been detected, with a large number of lines. The LSD line
detection method in Figure 21b also suffers from false detections, with
the detected lines being intermittent and a large number of interfering
lines in the background being detected. As shown in Figure 21c, the
original k-means line fitting algorithm has the advantages of low false
detection and missed detection rates. However, when the Hough transform
threshold is low, the detection error increases or even fails.The 21d
line fitting algorithm proposed in this article only displays the four
lines we need after clustering, and still has good detection performance
and high robustness even in the presence of significant interference in
the edge image.