Fast and reliable recognition of supplementary traffic signs

Abstract

Supplementary traffic signs are used to alter the meaning of other traffic signs. Assistance systems that recognize traffic signs therefore must also recognize supplementary signs to evaluate their influence on the meaning of detected traffic signs. We propose an algorithm which is able to detect supplementary signs in the vicinity of other signs using a novel rectangle segmentation algorithm. Support vector machines are used for the classification and rejection of other objects. The combination of both components permits to recognize a supplementary sign in less than 40 ms. First quantitative results for a test set with four different supplementary sign types show a very good classification accuracy of more than 96 %.

Recognition of supplementary signs for correct interpretation of traffic signs

“…Nienhüser et al [9] a work handling more types of subei Nässe" ("when raining") d to train a one-against-all Vector Machine (SVM) classifier for each type, which makes difficult to extend their approach for handling Liu et al [10] also include detection (based on gradients and Hough voting) and recognition (using a tree of binary SVMs applied to Fourier and wavelet descriptors) in their system…”

“…A solution is to use the pixel values, provided that rescaling applied, as was done in [8] and [9]. any slight offset or "bounding" error in the detection stage can translate into a total change of the resulting descriptor.…”

Recognition of supplementary signs for correct interpretation of traffic signs

“…We focus on global descriptors which describe the entire image contrary to local ones, such as interest points, localizing information around discriminative regions. A solution is to use the pixel values to a fixed size is first applied, as was done in [8] and [9] However the drawback is that "bounding" error in the detection stage can translate into a total change of the resulting descriptor.…”

“…For this "offset robustness" issue, it seems better to use descriptors that integrate information on large enough (one of the most frequent in France). Nienhüser et al [9] a work handling more types of subei Nässe" ("when raining") d to train a one-against-all Vector Machine (SVM) classifier for each type, which makes difficult to extend their approach for handling Liu et al [10] also include detection (based on gradients and Hough voting) and recognition (using a tree of binary SVMs applied to Fourier and wavelet descriptors) in their system…”

Relevance estimation of traffic elements using Markov logic networks

“…Previous work put an emphasis on the recognition of such elements [6], [9], [12], [14], [17]. Based on that work we focus on Dennis Nienhüser, Thomas Gumpp and J. Marius Zöllner are with FZI Forschungszentrum Informatik, Intelligent Systems and Production Engineering, Haid-und-Neu-Straße 10-14, 76131 Karlsruhe, Germany.…”