One of the steps towards designing innovative driving assistance system installed in a car is perfectly estimating the velocity of the driving car in any situated moment.
Abstract
One of the steps towards designing innovative driving assistance system installed in a car is perfectly estimating the velocity of the driving car in any situated moment. This system should be operated in real time, thus it’s vital to provide a credible algorithm with excellent performances. This system received as an input a video movie, which is taken from a digital video camera, which installed in the car and steered to the driving direction. The aim of this project is to evaluate the velocity of the driving car by analyzing video movie analyze.
Basic Approach
In this project, we have examined several methods of estimating the driving car velocity by calculating the traffic sign height in pixels and by evaluating the distance of the fragmented separating lines on the road. At first, we have developed a method, which is locked onto single object, and afterwards we have tried another method, which is locked onto several objects or regions simultaneously. Finally, we have developed a motion vectors model in order to map the movement domain and by this way extracting the car velocity.
The Simulation
Identifying the traffic sign height
Following after the object
Image processing stages:
- The original frame
- Finding the main line on the road
- Image screening (after binarization)
- Leaving the edge points after performing positive vertical gradient
- Leaving the three close edge points of the fragmented separating line
Motion vectors method
Tools
The Project was programmed in Matlab 6.5, on a PC platform.
Summary and Conclusions
The traffic sign method is unreliable since the search after the object depends on its environment.
The separating lines method by image processing is quite reliable even for problematic frames, but it’s designed only for a specific case of fragmented separating lines.
The vector motion method by locking a specific object during several frames is unreliable because of low success probability.
The vector motion method by vectors summing is more effective and reliable since it has a better success probability and we can add some probabilistic changes.
Acknowledgment
We would like to thank our devoted supervisor Mr. Ehud Orian for his professional guidance of developing this project. We are also grateful to the lab engineer Mr. Johanan Erez and to the VISL lab staff for their support and help. We are also grateful to the Ollendorff Minerva Center Fund for supporting this project.