Improved Matching Accuracy in Traffic Sign Recognition by Using Different Feature Subspaces
- Arihito Ihara, Hironobu Fujiyoshi, Masanari Takagi, Hiroaki Kumon, Yukimasa Tamatsu
- Machine Vision Applications, pp. 130–133, 2009.
Download: PDF (English)
This paper presents a traffic sign recognition method based on keypoint classification by AdaBoost using PCA-SIFT (principal component analysis scale invariant feature transform) features in different feature subspaces. A technique for recognizing traffic signs from an image taken with an in-vehicle camera has already been proposed to assist drivers. SIFT features are used for traffic sign recognition because they are robust to changes in scaling and rotation of traffic signs, but real-time processing is difficult because the computation cost of SIFT feature extraction and matching is high. In our method, two different feature subspaces are constructed from gradients in traffic sign images and those in general images. Detected keypoints are projected into both subspaces, and AdaBoost is used to classify whether they are on the traffic sign or not. Experimental results show that the computation cost for keypoint matching can be reduced to about half that of the conventional method.