Multi-type road marking recognition using adaboost detection and extreme learning machine classification

This paper presents a multi-type road marking recognition system by using a monocular camera on a moving platform. The system can detect various road markings. Firstly, an Inverse Perspective Mapping (IPM) transformation is introduced to suppress the perspective effect in the image, and the image slices which potentially belong to road markings are extracted based on high brightness slice filtering. Secondly, the prior knowledge of road making is applied to generate candidate road marking regions. Afterwards, a coarse-to-fine marking recognition method is presented. In the coarse recognition, an Adaboost classifier with Haar-like feature is adopted to fast eliminate non-marking candidates regions. In the fine recognition, an ELM classifier with BW-HOG feature is designed to recognize the types of markings. Finally, we introduce a spatial-temporal fusion method to further enhance the recognition accuracy and reliability of the system. Experimental results demonstrate the effectiveness of the proposed system.