AN IMPROVED DtBs METHOD FOR AUTOMATIC TRAFFIC SIGN RECOGNITION

摘要

In this study, we developed an improved image-based traffic sign recognition approach based on the Distances to Boundary (DtBs) method. The proposed algorithm is capable of detecting and adjusting the in-plane and out-of-plane rotations of each candidate object. Consequently, this approach is translation-, scale-, and rotation-invariant. Based on the preliminary results from a real field experiment, it has been demonstrated that a significant improvement on the traffic sign detection result can be achieved after implementing the proposed approach.