Additional Traffic Sign Detection and Content Recognition

Visual perception of a vehicle's surroundings is becoming an increasingly important aspect of modern advanced driver assistance systems developed on the road towards automated driving. One of its fundamental, yet challenging tasks is the recognition of traffic signs. In this work, a subproblem of standard main traffic sign recognition is considered, which is essential to ensure their correct interpretation: the detection and recognition of additional traffic signs. These signs may further specify the meaning or alter the validity of main signs, such as speed limits. We present the first complete pipeline for additional sign content recognition, including a novel sign detection approach. Each component and the complete system are extensively evaluated on a database of mostly German additional traffic signs. Furthermore, the problem of directional road sign detection in complete image frames is investigated.