The Hough Transform for Vertical Object Recognition in
The Hough Transform for Vertical Object Recognition in 3 D Images Generated from Airborne Lidar Data Christopher Parrish ECE 533 Project December 2006
Airborne Lidar GPS Reference Station Airport Obstruction Surveying
Hough transformbased approach for detecting vertical objects of cylindrical shape: Lidar Point Cloud Voxelize 3 D Grayscale Intensity Image 3 D Sobel operator 3 D Grayscale Edge Image Threshold segmentation 3 D Binary Edge Image Hough Transform to identify vertical cylinders Vertical objects of interest
2 D Color Image Laser Point Cloud 3 D Grayscale Image
Computing Binary Edge Image: Gradient of a 3 D image, f(x, y, z): Magnitude of the gradient: 3 D Sobel operator (three 3 x 3 x 3 filters expressed here as sets of three 2 D matrices) Thresholded (binary) edge image
3 D Binary Edge Images
HT Cylinder Detection Algorithm: Assume cylinders are vertical (axes parallel to mapping frame Z axis) => # of parameters reduced from 5 to 3. Representation: (X-s)2+(Y-t)2 = r 2 Input = 3 D binary edge image Quantize 3 D parameter space. q Initialize all accumulator cells to zero. q For each nonzero voxel in 3 D binary edge image, step through all values of s and t. At each location: q Solve for r q Round r to its nearest accumulator cell value q Increment counter for that (s, t, r) accumulator cell. q Find entry in 3 D accumulator array with highest # of votes.
Cylinders Detected Using Hough Transform:
Comparison of radii & axes locations of HT-detected cylinders with field-surveyed data:
- Slides: 9
