In this work, we obtain a vehicle semantic understanding using their image features and a rule-based system. These features provide the vehicle spatial and temporal information. Vehicle spatial feature is obtained using an ACF network. The vehicle temporal information is obtained using a novel semantic segmentation framework. The statuses of the neighboring vehicles are categorized as «Carfollow » and «Car-avoid» also. We validate our proposed frame-work with multiple acquired sequences. Our experimental results show that the proposed framework can estimate the status of the different vehicles in the urban road environment in near real-time.
Инноватика-2021 : сборник материалов XVII Международной школы-конференции студентов, аспирантов и молодых ученых, 22-23 апреля 2021 г., г. Томск, Россия. Томск, 2021. С. 164-167