The Shadow Knows: Detecting Speed Bumps and Negative Obstacles at Night

Driver assistance systems using forward facing cameras are becoming mainstream. These cameras are typically rigidly mounted inside the driver cabin near the rearview mirror. This puts the cameras at a fixed position, significantly above the vehicle headlights. In this configuration the shadows produced by the headlights, when driving on dark roads, are visible in the camera image. This paper describes how to use these shadows to enhance obstacle detection. In particular it describes how the shadows can be used for detecting speed-bumps and negative obstacles at night from a moving vehicle. This information can be used to warn the driver or to prepare the vehicle for the anticipated change in road surface. This paper performs a detailed analysis of the particular image motion characteristics of the shadow when the car is moving forward. This information allows the discrimination between shadows and other dark patches on the road and then, through careful analysis of the shadow over time, both the depth and shape of the obstacle can be determined. The detection range from shadow analysis is significantly larger than what is possible from stereo or structure from motion (SFM) in dark night conditions.