Recognition




Low Level processing (feature extraction)

Goal: extract meaningful features from optical image (to detect object borders)

  1. Edge detection (Canny filter)
  2. Straight-line extraction (Burns method)
  3. Line filtering and recovery


Edge detection by Canny filtering

Goal: detect points belonging to object edges

Example

Image after Perona-Malik filtering
Edges extracted by Canny filtering



Straight-line extracton by Burns approach

Goal: detect border lines

Example

Edges extracted by Canny filtering
Line types extracted by the Burns method



Line filtering and recovery

Goal: eliminate spurious segments and recovery of partially missing segments

Example

Lines extracted by the Burns method
Recovered lines



High Level processing

  1. Line Properties extraction
  2. Relational Graph Construction
  3. Matching with model database


Line properties extraction

Pair of lines attributes:

Angles attributes:

Proximity attributes:


Relational Graph Construction

Nodes

Arcs

Example


Interaction with Environment Database




Back to previous page