Object detection and tracking for Real World Applications

Road detection and tracking for personal navigation

These images show the problem with the use of a single vanishing point for road detection and how multiple vanishing points may correct the problem.

 

Junction detection
Detection takes place in the Inverse Perspective Mapping (IPM) image. The IPM image tends to be like mosaic. Even though false edges in the original image appear irregular such as those of the grass and trees, they become rectangular in the IPM image due to the mosaic effect. This could produce false positives. Another source of errors is when, say the left boundary of a junction, may be vague and the detection algorithm takes the middle line as the left boundary so the junction shifts to the right.

Video based Navigation

Li Bai, Yan Wang, The University of Nottingham

 

The aim of the project is to develop a system that will allow the driver to view a virtual sign (yellow arrow) on the road, in front of the car, in the video captured by a camera, or through a HMD so the driver can travel to the destination by following the virtual sign. This new generation of video based navigation systems are for those who find traditional 2D map based navigation systems difficult to use - you can read a map, only can't find the way. 

 1. Vision - Where is the road?

 

2. GPS - Where am I?

2. GPS+INS - Inertial sensor to correct GPS errors

 

4. Map Matching – GPS to GIS

5. Route planning - Where to go next?