Object Detection and Tracking in Range

Sequences by Separation of Image Features Esther B. Meier, Frank Ade Abstract| We propose a new approach for a driver assistance system based on range image sequences to increase the safety on motorways. This control system is capable of automatically keeping the car at an adequate distance or warning the driver when he gets too close to other cars. The knowledge of the sensor geometry with respect to the road is used to separate image features into ground and road obstacles. We assume that the road is at but possibly inclined. To distinguish between obstacles and road pixels we use a separating plane that is slightly steeper than the current road model. Pixels lying above this plane are assumed to belong to obstacles, and pixels below the plane are assumed to belong to the road. To update the road model, we t a plane through all pixels marked as ground. The partition of the detected obstacles into their di erent objects is performed using connected component analysis. Components are then merged into objects according to proximity. For the tracking we employ Kalman ltering to provide most likely estimates of the state of each object. We show results of some experiments with simulated tra c scenes and toy tra c scenarios. Real data will be available later in the project. Keywords| range images, model-based segmentation, Kalman ltering, object tracking.