R3-[Q-Shape] - Research: Shape-Based Tracking of Objects


  • Utilizes specially adapted, state-of-the-art shape matching
  • No odometry information required
  • Reliable & efficient
  • Can easily be extended to handle dynamics
  • Well-suited for object tracking and mapping applications


Principle of Operation

Shape Extraction

Shapes are extracted from laser range finder (LRF) data. In mapping/localization applications, shape extracted from a LRF can also be matched against shape extracted from a map.

Optimal Similarity of Polylines

Similarity of shape features is determined by an optimal matching focussing on similar sections. This allows a robust detection of similarity even below the noise level of sensor data.

Shape Matching

Complete shapes of two consecutive scans are matched against each other; this yields the desired correspondence of objects.


The presented shape matching is utilized in a sample application of tracking shape features. Data obtained from the Bremen autonomous wheelchair on an indoor course at the University of Bremen.

The demonstration presents the processed data scan by scan. The robot's position is marked by a red triangle, individual shapes are depicted in different colors. Whenever a new shape feature emerges, it is emphasized. For viewing convenience, scans are aligned as determined by the matching. The grid shown is 1 meter in size.

Shape based, polygonal map can be constructed from the matching. Herefore, methods for aligning and merging of polylines are required.

The authors would like to thank Dr.-Ing. Thomas Röfer for making the data available to us.