Vehicle re-identification (ReID) aims to retrieve the target vehicle in an extensive image gallery through its appearances from various views in the cross-camera scenario. It has gradually become a core technology of intelligent transportation system. Most existing vehicle re-identification models adopt the joint learning of global and local features. However, they directly use the extracted global features, resulting in insufficient feature expression. Moreover, local features are primarily obtained through advanced annotation and complex attention mechanisms, which require additional costs. To solve this issue, a multi-feature learning model with enhanced local attention for vehicle re-identification (MFELA) is proposed in this paper. The model consists of global and local branches. The global branch utilizes both middle and high-level semantic features of ResNet50 to enhance the global representation capability. In addition, multi-scale pooling operations are used to obtain multi-scale information. While the local branch utilizes the proposed Region Batch Dropblock (RBD), which encourages the model to learn discriminative features for different local regions and simultaneously drops corresponding same areas randomly in a batch during training to enhance the attention to local regions. Then features from both branches are combined to provide a more comprehensive and distinctive feature representation. Extensive experiments on VeRi-776 and VehicleID datasets prove that our method has excellent performance.

Vehicle re-identification (Re-ID) is one of the promising applications in the field of computer vision. Existing vehicle Re-ID methods mainly focus on global appearance features or pre-defined local region features, which have difficulties in handling inter-class similarities and intra-class differences among vehicles in various traffic scenarios. This paper proposes a novel end-to-end three-branch embedding network (TBE-Net) with feature complementary learning and part-aware ability. The proposed TBE-Net integrates complementary features, global appearance, and local region features into a unified framework for subtle feature learning, thereby obtaining more integral and diverse vehicle features to re-identify the vehicle from similar ones. The local region feature branch in the proposed TBE-Net contains an attention module that highlights the major differences among local regions by adaptively assigning large weights to the critical local regions and small weights to insignificant local regions, thereby enhancing the perception sensitivity of the network to subtle discrepancies. The complementary branch in the proposed TBE-Net exploits different pooling operations to obtain more comprehensive structural features and multi-granularity features as a supplement to the global appearance and local region features. The abundant features help accommodate the ever-changing critical local regions in vehicles’ images due to the sensors’ settings, such as the position and shooting angle of surveillance cameras. The extensive experiments on VehicleID and VeRi-776 datasets show that the proposed TBE-Net outperforms the state-of-the-art methods.

Though experimental results have shown a strong correlation between contextual features and the driver’s fatigue state, contextual features have been applied only offline to evaluate a driver’s fatigue state. This paper identifies three of the most effective contextual features, i.e., continuous driving time, sleep duration time, and current time, to facilitate the real-time (online) recognition of fatigue state. By applying gray relational analysis, the three contextual features, together with the most effective facial and vehicle behavior features, are introduced in a two-level fusion structure to improve fatigue driving recognition. In the first level of fusion, labeled the feature-level fusion, three separate multiclass support vector machine (MCSVM) classifiers are used for the three feature sources, i.e., contextual features, driver’s facial features, and vehicle behavior features, to fuse information. These three MCSVM classifiers output probabilities as inputs for the three real-time dynamic basic probability assignments (BPAs) at the second level of fusion, labeled decision-level fusion. These BPAs, and the fusion result of the previous time step, are fused in the decision-level fusion based on the Dempster–Shafer evidence theory. This includes modifying the BPAs to accommodate the decision conflict among the different feature sources. Field experiments show that the proposed recognition method can outperform the single-fatigue-feature method and the single-source fusion-based method.

The image classification based on cloud computing suffers from difficult deployment as the network depth and data volume increase. Due to the depth of the model and the convolution process of each layer will produce a great amount of calculation, the GPU and storage performance of the device are extremely demanding, and the GPU and storage devices equipped on the embedded and mobile terminals cannot support large models. So it is necessary to compress the model so that the model can be deployed on these devices. Meanwhile, traditional compression based methods often miss many global features during the compression process, resulting in low classification accuracy. To solve the problem, this paper proposes a lightweight neural network model based on dilated convolution and depthwise separable convolution with twenty-nine layers for image classification. The proposed model employs the dilated convolution to expand the receptive field during the convolution process while maintaining the number of convolution parameters, which can extract more high-level global semantic features to improve the classification accuracy. Also, the depthwise separable convolution is applied to reduce the network parameters and computational complexity in convolution operations, which reduces the size of the network. The proposed model introduces three hyperparameters: width multiplier, image resolution, and dilated rate, to compress the network on the premise of ensuring accuracy. The experimental results show that compared with GoogleNet, the network proposed in this paper improves the classification accuracy by nearly 1%, and the number of parameters is reduced by 3.7 million.