Vehicle semantic detection on the highway via the moving platform
Wentao Lv, Xiaocheng Yang, Jiaqi Adam Huang, Long Wu, Weiqiang Xu · 2017
This paper focuses on the vehicle detection on the highway from a moving platform. We present a novel detection algorithm based on the semantic information. First, a semantic screening strategy is used to restrict the search area. This is helpful for improving efficiency of the method. Then, the superpixel segmentation algorithm is applied to extract the suspected targets. Based on each isolated region, a series of features, including geometric features and the histogram of orient gradient (HoG) feature, are calculated. Finally, these features are fed into a classifier to recognize the real vehicles. Since the semantic contents are used for distinguishing the real targets from other interferences, our method has a more robustness for the false information. The experimental results based on real scenes demonstrate the effectiveness of our method.