Unobstructed space recognition with a monochrome camera

The objective of the proposed algorithm is to identify unobstructed space in front of a moving passenger car by means of a single monochrome camera. The approach is divided into two principal parts. First a model based road border recognition with focus on a sophisticated image feature extraction was developed to deal with different sources of distracting noise like bad lightning conditions, broken or absent markings, etc. We utilize the orientation of edges as measurement input for the following recursive parameter estimation. Here, a Kalman filter is used to estimate the model parameters like vehicle position and width and curvature of the road. The second part is more ambitious: The detection of obstacles within the road boundaries. That includes moving and stationary objects of different sizes. The simple but challenging fact which comes with a monocular camera system is the missing range information. In this contribution we present an extended approach to deal with this problem by combining odometry and optical flow. Both parts were tested with a passenger car under real road conditions. The obtained results provide a qualitative and quantitative statement of the capabilities of the system.