Lab 7

Lab 7#

This lab is devoted to metrology, by detecting features on the image L.png.

  • Recall what the following detectors can do: Canny, Harris, Hough, Moravec, Prewitt, Roberts, Sobel.

Sobel and Canny detectors#

  • Display the image gradients with the Sobel detector (skimage.filters.sobel).

  • Apply a threshold on the obtained image to perform an edge detection. What is the influence of the threshold on the result?

  • Find the optimal threshold, i.e. the one which gives most of the edges of the object while keeping precise locations of the edges.

  • What is the relationship between Sobel and Canny detectors?

  • Display the edges detected by Canny detector (skimage.feature.canny).

  • Discuss the influence of the main parameters of the method, namely the size of the Gaussian filter and the two thresholds.

  • It is sometimes interesting to compare the studied methods in terms of computation time. The computation time can be obtained by calculating the difference between the time \(t_2\) measured after executing the method and the time \(t_1\) measured before its execution. For that, you can use the function time.time which gives the number of seconds elapsed since January 1, 1970. What is the fastest method?

Harris detector#

  • Apply the Harris detector (with skimage.feature.corner_harris and skimage.feature.corner_peaks).

  • Criticize the result: have all the corners been detected? Are there any false alarms (i.e. detections that do not correspond to corners)? How to explain these errors?

Hough transform#

  • The Hough transform does not apply directly to the original image: what image should you use?

  • Represent the Hough transform of the image with skimage.transform.hough_line.

  • The function skimage.transform.hough_line_peaks extracts the parameters of the most important lines from the Hough transform. Use this function to display the six most important lines of the image.

Contents