Vehicle trajectory prediction on highways using bird eye view representations and deep learning
نویسندگان
چکیده
Abstract This work presents a novel method for predicting vehicle trajectories in highway scenarios using efficient bird’s eye view representations and convolutional neural networks. Vehicle positions, motion histories, road configuration, interactions are easily included the prediction model basic visual representations. The U-net has been selected as kernel to generate future of scene an image-to-image regression approach. A implemented extract positions from generated graphical achieve subpixel resolution. trained evaluated PREVENTION dataset, on-board sensor dataset. Different network configurations have evaluated. study found that with 6 depth levels linear terminal layer Gaussian representation vehicles is best performing configuration. use lane markings was produce no improvement performance. average error 0.47 0.38 meters final 0.76 0.53 longitudinal lateral coordinates, respectively, predicted trajectory length 2.0 seconds. up 50% lower compared baseline method.
منابع مشابه
Deep Learning of Player Trajectory Representations for Team Activity Analysis
Team sports such as ice hockey and basketball involve complex player interactions. Modeling how players interact with each other presents a great challenge to researchers in the ield of sports analysis. The most common source of data available for this type of analysis is player trajectory tracking data, which encode vital information about themotion, action, and intention of players. At an ind...
متن کاملTraffic Accident Prediction Using Vehicle Tracking and Trajectory Analysis
Alsau?--lntelligent visual surveillance for road vehicles is a key camponmt for developing autonomous intelligent transportation sys tem. I n this paper, a probabilistic model for prediction of traliic accidents using 3D model based vehicle tracking is proposed. Sample data including motion trajectories are first obtained by 3D model based vehicle tracking. A fuzzy self-organizing neural networ...
متن کاملCreating a Bird-Eye View Map
An omnidirectional camera has properties which are highly beneficial for navigation of a mobile robot. In this thesis this is demonstrated with a new application; a visual map of the environment created from bird-eye view perspective. This visual map can be created on-line, which allows an operator to navigate the robot through an unknown environment based on this visual map.
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Applied Intelligence
سال: 2022
ISSN: ['0924-669X', '1573-7497']
DOI: https://doi.org/10.1007/s10489-022-03961-y