Aiming at the problems of low detection accuracy and inaccurate positioning accuracy of light-weight network in traffic sign recognition task, an improved light-weight traffic sign recognition algorithm based on YOLOv4-Tiny was proposed. By improving the K-means clustering algorithm, the anchor with appropriate size is generated for the traffic sign data set to improve the detection recall rate and target positioning accuracy. The strategy of large-scale feature map optimization is proposed, which enriches the feature level of the network by using the low-level information, strengthens the representation of the feature information of the small target, and improves the detection accuracy of the long-range small target. In view of the problem of missed detection of high overlapping targets in the post-processing stage of the model, the paper proposes an improved NMS algorithm to screen the prediction box, avoid deleting the prediction results of different targets, and further improve the detection accuracy and recall rate of the target. Experimental results show that, compared with the original YOLOv4-Tiny algorithm, the improved algorithm in traffic sign recognition task based on TT100K dataset, mAP and recall are improved by 5.73% and 7.29% respectively, and FPS value is maintained at about 87 f/s, which meets the accuracy and real-time requirements of traffic sign recognition task.

The detection and recognition of traffic signs in complex environments has received extensive attention, and the correct detection of small targets and occluded targets are two key issues. This paper proposes a context-aware and attention-driven weighted fusion network for traffic sign detection. Specifically, the design of context module not only enhances the diversity of global features, but also reduces the sensitivity of convolution to occluded objects. In addition, the attention-driven weighted fusion feature pyramid is designed to efficiently fuse deep semantic information and shallow representation information, enhance the transfer of foreground positioning information, eliminate the problems of semantic spacing and noise interference in the process of feature fusion, and enhance the network’s representation of small and occluded traffic signs. The experimental results on the traffic sign dataset TT100K show that the method is effective and superior.

This paper addresses the problem of multiple signal classification‐ (MUSIC‐) based direction of arrival (DOA) and polarization estimation and proposes a new dimensionality reduction MUSIC (DR‐MUSIC) algorithm. Uni‐vector‐sensor MUSIC algorithm provides estimation for DOA and polarization; accordingly, a four‐dimensional peak search is required, which hence incurs vast amount of computation. In the proposed DR‐MUSIC method, the signal steering vector is expressed in the product form of arrival angle function matrix and polarization function vector. The MUSIC joint spectrum is converted to the form of Rayleigh‐Ritz ratio by using the feature where the 2‐norm of polarization function vector is constant. A four‐dimensional MUSIC search reduced the dimension to two two‐dimensional searches and the amount of computation is greatly decreased. The theoretical analysis and simulation results have verified the effectiveness of the proposed algorithm.