Salient Object Detection in Traffic Scenes Based on RGB-D Fusion

In autonomous driving technology, the detection of salient objects in traffic scenes plays a crucial role in ensuring vehicle safety. This paper introduces a novel detection approach that integrates RGB imagery with depth information to overcome the limitations inherent in current research on salient object detection within traffic contexts. Our method not only emulates the distribution of human visual attention but also prioritizes the identification of fully salient objects, thereby providing the target recognition component of intelligent driving systems with more precise data. We engineered a bidirectional information transmission network (BIT-Net) that adeptly merges data from both modalities to facilitate the accurate detection of salient objects across various scales. Extensive experiments conducted on several well-established benchmark datasets demonstrate that our approach outperforms existing methods in terms of accuracy and recall, among other key performance metrics. This research not only advances the field of salient object detection but also provides a robust foundation for the practical implementation of intelligent driving assistance systems.