However, this attention-worthy problem ensures the accuracy of target detection, the low delay of video transmission and the high quality of video streams in a CVIS with the characteristics of a dynamic network topology, small transmission radius and time-varying channel.Īs one of the current popular research fields, deep learning was first proposed by Hinton in 2006. These were divided into video transmission, video content distribution and video target detection, which can effectively grasp road conditions, improve driving safety and provide users with popular video streams to improve the QOE. Video applications over CVIS is one of the important parts of ITS. Vehicles equipped with wireless communication units and sensing units are used as mobile nodes, and their communication modes include vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I) and vehicle-to-roadside (V2R) modes. The proposal of CVIS can greatly alleviate the pressure of network overheads in ITS. Intelligent transportation systems (ITS) are expected to relieve traffic pressure and prevent traffic jams through video target detection, video-assisted driving and wireless resource management. In recent years, with the rise of the social vehicle ownership rate, the needs of traffic safety and quality of user experience (QOE) are increasing.
Finally, the challenges and development trends of deep learning in the field were explored and discussed. Then, we summarized the main methods of deep learning and deep reinforcement learning algorithms for video applications over CVIS, and made a comparative study of their performances. Firstly, the research status of traditional video application methods and deep learning methods over CVIS were introduced the existing video application methods based on deep learning were classified according to generative and discriminative deep architecture.
Therefore, the research value and significance of video applications over CVIS can be better reflected through deep learning. However, the in-depth structure of deep learning has the ability to deal with high-dimensional data sets, which shows better performance in video application problems. Dealing with large datasets of feedback from complex environments is a challenge when using traditional video application approaches. Video application is a research hotspot in cooperative vehicle-infrastructure systems (CVIS) which is greatly related to traffic safety and the quality of user experience.