Java implementation of the Hough transform and RANSAC algorithm

• The constructor reads the data points from a text file into a list
• showCanvas calls execute
  • for (maxIter iterations)
    • The execute method chooses 3 points from the data list at random. Their indices are stored in the concensusSet.
    • A circle that passes through all 3 points is chosen as the current circle model for (all points in the data set)
      • getoffset is responsible for calculating the difference between the points and the circumference of the circle
      • Points that are within threshold distance are added to the consenus set
    • The best model so far (as determined by the highest number of points in the consenus set) is kept track on during the iteration
  • The best model is returned
• showCanvas draws the data points, the best circle and the smallest width annulus on a canvas and displays the result to the user

• (+) showCanvas() Displays the data points and the calculated circles on a canvas.
• (+) execute() Runs the RANSAC algorithm.
• (-) getOffset() Calculate the distance between a point and the center of a circle
• (-) getCircle() Find the circle that passes through the three given points
• (-) getNPoints() Choose n data points at random

Points from points.r.data are being used.

• Cirlces (x, y, r, number of points)
  • 20, 100, 100, 100
• Noise: 900 points randomly selected between -200 and 200
• smallestRadius=92.17273473900397
• highestRadius=111.01668853035679
• 18.843953791352817

• The constructor reads the data points from a text file into a list
• A wrapped 3D array - the accumulator - is used to store information about potential circles. A circle is identified by pixel ranges for x, y and radius
• execute()
  • for (all points)
    • all circles that could have point on its circumference are calculated
    • for (all circles)
      • the cell in the accumulator is incremented for the given circle
• filterNeighbours removes circles that lie too close to each other. Circles with a high count are preferred.
• showCanvas draws the points and the identified circles

The smallest width annuli for the Hough circle suffers from some offset problems, possibly caused by rounding errors.

• (+) execute() Executes the Hough Transform algorithm.
• (+) showCanvas() Render view based on this.pixels.
• (-) getCircles() Finds every other circle that has @point.getY() and @point.getX() as its center

Points from points.ht.data are being used.

• Circles (x, y, r, number of points)
  • 0, 0, 50, 50
  • 100, 140, 110, 50
  • 0, 50, 50, 50
  • 20, 100, 100, 100
• Noise: 120 points randomly selected between -200 and 200