Search

Vehicle Re-identification (Re-ID) has been broadly studied in the last decade; however, the different camera view angle leading to confused discrimination in the feature subspace for the vehicles of various poses, is still challenging for the Vehicle Re-ID models in the real world. To promote the Vehicle Re-ID models, this paper proposes to synthesize a large number of vehicle images in the target pose, whose idea is to project the vehicles of diverse poses into the unified target pose so as to enhance feature discrimination. Considering that the paired data of the same vehicles in different traffic surveillance cameras might be not available in the real world, we propose the first Pair-flexible Pose Guided Image Synthesis method for Vehicle Re-ID, named as VehicleGAN in this paper, which works for both supervised and unsupervised settings without the knowledge of geometric 3D models. Because of the feature distribution difference between real and synthetic data, simply training a traditional metric learning based Re-ID model with data-level fusion (i.e., data augmentation) is not satisfactory, therefore we propose a new Joint Metric Learning (JML) via effective feature-level fusion from both real and synthetic data. Intensive experimental results on the public VeRi-776 and VehicleID datasets prove the accuracy and effectiveness of our proposed VehicleGAN and JML.

Deep learning-based intelligent vehicle perception has been developing prominently in recent years to provide a reliable source for motion planning and decision making in autonomous driving. A large number of powerful deep learning-based methods can achieve excellent performance in solving various perception problems of autonomous driving. However, these deep learning methods still have several limitations, for example, the assumption that lab-training (source domain) and real-testing (target domain) data follow the same feature distribution may not be practical in the real world. There is often a dramatic domain gap between them in many real-world cases. As a solution to this challenge, deep transfer learning can handle situations excellently by transferring the knowledge from one domain to another. Deep transfer learning aims to improve task performance in a new domain by leveraging the knowledge of similar tasks learned in another domain before. Nevertheless, there are currently no survey papers on the topic of deep transfer learning for intelligent vehicle perception. To the best of our knowledge, this paper represents the first comprehensive survey on the topic of the deep transfer learning for intelligent vehicle perception. This paper discusses the domain gaps related to the differences of sensor, data, and model for the intelligent vehicle perception. The recent applications, challenges, future researches in intelligent vehicle perception are also explored., Comment: Updated some figures

The impact of droplets on a liquid pool is ubiquitous in nature and important in many industrial applications. A droplet impacting on a liquid pool can result in the pinch-off of a regular bubble or entrap a large bubble under certain impact conditions. In this study, the cavity deformation and the bubble entrapment during the impact of droplets on a liquid pool are studied by combined experimental measurements and numerical simulations. The time evolution of the free surface profile obtained in the numerical simulation is in good agreement with the experimental results. The cavity created by the droplet impact affects the pinch-off of regular bubbles and the entrapment of large bubbles. The regular bubble pinch-off is the direct consequence of the capillary wave propagating downward along the interface of the cavity and merging at the bottom of the cavity. In contrast, the large bubble entrapment is due to the merging of the liquid crowns at the mouth of the cavity. Gravity and environmental pressure play important roles in cavity deformation and bubble entrapment after droplet impact on liquid pools. The maximum depth of the cavity decreases as the gravitational effect becomes stronger. The phenomenon of regular bubble pinch-off may disappear as the gravity decreases. We find that the regular bubble pinch-off can transform into large bubble entrapment when increasing environmental pressure. The size of the large bubble entrapped decreases with increasing the environmental pressure due to the larger difference in the pressure and the vorticity around the liquid crown at increased environmental pressure. Finally, the regime map of bubble entrapment after the droplet impact is obtained., Comment: 13 pages, 13 figures

The impact of droplets at increased environmental pressure is important in many industrial applications. Previous studies mainly considered the impact process at standard or reduced environmental pressure, and the effect of high environmental pressure is unclear. In this study, we experimentally investigate the impact of ethanol droplets on dry smooth surfaces at increased environmental pressure. The effects of the environmental pressure on the splashing and rupture of the crown during the impact process are analyzed. The results show that surrounding gas with high environmental pressure can lead to the splashing of the crown in a 'thread rupture' mode and the sizes of the secondary droplets from the rim of the liquid crown increase with the environmental pressure. The threshold for the transition from spreading to splashing during the impact process is obtained based on the theory of aerodynamics analysis of the lamella. At increased environmental pressure, the threshold speed of the impact decreases with increasing the environmental pressure because the wedge of the lamella is prevented from moving forward and is driven to detach from the substrate by the air ahead, which has a higher density due to the higher environmental pressure., Comment: 10 pages, 9 figures