Tunnel Detection for Automotive Radar using Object Elevation Trends and Gaussian Filtering

This paper presents detection of tunnels based on object elevation property and gaussian filtering. Tunnel detection is essential to avoid false events in Adaptive Cruise Control (ACC) and Automatic Emergency Braking (AEB) functionalities. The new feature presented in this paper is one of the methods for tunnel detection. This method concentrates on the elevation history of objects in a specified region of interest. The feature helps to detect less reflective tunnels, curvy tunnels, tunnels with varying patterns of ceiling lights and some short tunnels. The use of gaussian filtering is done to filter out elevated objects from side guardrails and trees to avoid any false tunnel activation. The fusion of this new feature with the existing algorithm helps in improving the tunnel true positive rate by ~3.5% and reducing the tunnel false positive rate by ~4%.