Minimum contrast ratio threshold for reliable camera-based ADAS lane recognition

Determine a specific minimum contrast ratio threshold of road markings, preferably quantified using the Michelson contrast ratio computed from camera imagery, that reliably predicts successful traffic lane recognition by camera-based advanced driver assistance systems whose algorithms simultaneously use edge detection and recognition of other roadway features across diverse visibility conditions.

Background

The study evaluates whether inadequate contrast ratio of road markings can indicate impending failures of an open-source camera-based advanced driver assistance system (ADAS) in lane recognition under varying conditions (day, night, rain, glare). The authors measured the Michelson contrast ratio of Type I (flat) and Type II (structured) road markings and compared these measurements to the ADAS’s lane detection performance.

While they found strong positive correlations between contrast ratio and ADAS detection under poor visibility, they were unable to identify a single minimum contrast ratio value that guarantees reliable operation. They attribute this to the complexity of camera-based ADAS algorithms, which rely simultaneously on edge detection and recognition of other roadway features, leaving the precise threshold undetermined.