Complex Traffic Scene Image Classification Based on Sparse Optimization Boundary Semantics Deep Learning

With the rapid development of intelligent traffic information monitoring technology, accurate identification vehicles, pedestrians and other objects on road has become particularly important. Therefore, in order to improve recognition classification accuracy image complex scenes, this paper proposes a segmentation method semantic redefine using boundary region. First, we use SegNet model obtain rough features vehicle object, then simple linear iterative clustering (SLIC) algorithm over segmented area image, which can determine each pixel super area, optimize target small areas image. Finally, edge recovery ability condition random field (CRF) is used refine boundary. The experimental results show that compared with FCN-8s SegNet, proposed improves by 2.33% 0.57%, respectively. And Unet, performs better when dealing multi-target segmentation.