A Semi-Automatic 2D Solution for Vehicle Speed Estimation from Monocular Videos

Kumar, Amit; Khorramshahi, Pirazh; Lin, Wei-An; Dhar, Prithviraj; Chen, Jun-Cheng; Chellappa, Rama

doi:10.1109/cvprw.2018.00026

Cited by 35 publications

(38 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…These methods involve different image processing techniques: background extraction [3,20,23,14], image rectification [3,20,23,12], detecting and tracking reference points [3,20,14] or centroids [23] over successive frames, converting the displacement vectors from the image to the real-world coordinate system. The state-of-the-art results of deep learning in vision tasks makes object detection and tracking [12], locating license plates on vehicles [14], 3D convolutional networks [4] other promising directions in the task of speed estimation. Disadvantages of these approaches and of the traffic enforcement solutions already in use, such as speed or point-to-point cameras, include the need for calibration processes, meticulous positioning of the devices at predefined locations, investment in infrastructure and maintenance.…”

Section: Related Workmentioning

confidence: 99%

Intuitive Estimation of Speed using Motion and Monocular Depth Information

Rill

2020

Studia UBB Informatica

View full text Add to dashboard Cite

Advances in deep learning make monocular vision approaches attractive for the autonomous driving domain. This work investigates a method for estimating the speed of the ego-vehicle using state-of-the-art deep neural network based optical flow and single-view depth prediction models. Adopting a straightforward intuitive approach and approximating a single scale factor, several application schemes of the deep networks are evaluated and meaningful conclusions are formulated, such as: combining depth information with optical flow improves speed estimation accuracy as opposed to using optical flow alone; the quality of the deep neural network results influences speed estimation performance; using the depth and optical flow data from smaller crops of wide images degrades performance. With these observations in mind, a RMSE of less than 1 m/s for ego-speed estimation was achieved on the KITTI benchmark using monocular images as input. Limitations and possible future directions are discussed as well.

show abstract

Section: Related Workmentioning

confidence: 99%

Intuitive Estimation of Speed using Motion and Monocular Depth Information

Rill

2020

Studia UBB Informatica

View full text Add to dashboard Cite

show abstract

“…Также алгоритмы могут отличаться связями между этапами трекинга. В некоторых случаях все этапы реализованы отдельными модулями, которые оптимизируются отдельно ( [7,8,9,10]). А есть подходы, в которых 3 последних этапа объединяются в единую end-to-end сеть и оптимизируются совместно ( [4], [5], [6]).…”

Section: обзор существующих методовunclassified

“…Еще одно отличиеинформация, на основе которой принимается решение о сопоставлении новых обнаружений с траекториями. Могут быть рассмотрены связи между соседними обнаружениями [9], могут учитываться несколько последних обнаружений [5] или же на вход подается вся видеопоследовательность целиком, и есть возможность учитывать обнаружения на всех кадрах: предыдущих и последующих. А в некоторых работах предлагают строить траектории из обнаружений с помощью MCMC (Markov chain monte carlo) [12], [14]…”

Section: обзор существующих методовunclassified

“…Высокая частота детекции приводит к выполнению этого условия. В строке 1 Таблицы 2 в качестве примера было рассмотрено одно из призовых решений [9] NVIDIA AI City Challenge, использовавшее для отслеживания автомобилей метод SORT [15].…”

Section: обзор существующих методовunclassified

See 1 more Smart Citation

Algorithm for Counting Cars in Large-scale Video Surveillance Systems

Shirokov

Kuplyakov

Konushin

2020

Proceedings of the 30th International Conference on Computer Graphics and Machine Vision (GraphiCon 2020). Part 1

View full text Add to dashboard Cite

The article deals with the problem of counting cars in large-scale video surveillance systems. The proposed method is based on car tracking and counting the number of tracks intersecting the given signal line. We use a distributed tracking algorithm. It reduces the amount of necessary computational resources and increases performance up to realtime by detecting vehicles in a sparse set of frames. We adapted and modified the approach previously proposed for people tracking. Proposed improvement of the speed estimation module and refinement of the motion model reduced the detection frequency by 3 times. The experimental evaluation shows that the proposed algorithm allows reaching an acceptable counting quality with a detection frequency of 3 Hz.

show abstract

“…Some methods make use of line markings on the road to deal with perspective projection [15][16][17]. In [18], the width of lane and road white strips length are essential information for camera calibration. In this study, Mask-RCNN and deep-SORT [19] are vehicle detector and tracker, respectively.…”

mentioning

confidence: 99%

Vehicle Speed Estimation Based on 3D ConvNets and Non-Local Blocks

2019

View full text Add to dashboard Cite

Vehicle speed estimation is an important problem in traffic surveillance. Many existing approaches to this problem are based on camera calibration. Two shortcomings exist for camera calibration-based methods. First, camera calibration methods are sensitive to the environment, which means the accuracy of the results are compromised in some situations where the environmental condition is not satisfied. Furthermore, camera calibration-based methods rely on vehicle trajectories acquired by a two-stage tracking and detection process. In an effort to overcome these shortcomings, we propose an alternate end-to-end method based on 3-dimensional convolutional networks (3D ConvNets). The proposed method bases average vehicle speed estimation on information from video footage. Our methods are characterized by the following three features. First, we use non-local blocks in our model to better capture spatial–temporal long-range dependency. Second, we use optical flow as an input in the model. Optical flow includes the information on the speed and direction of pixel motion in an image. Third, we construct a multi-scale convolutional network. This network extracts information on various characteristics of vehicles in motion. The proposed method showcases promising experimental results on commonly used dataset with mean absolute error (MAE) as 2.71 km/h and mean square error (MSE) as 14.62 .

show abstract

A Semi-Automatic 2D Solution for Vehicle Speed Estimation from Monocular Videos

Cited by 35 publications

References 12 publications

Intuitive Estimation of Speed using Motion and Monocular Depth Information

Intuitive Estimation of Speed using Motion and Monocular Depth Information

Algorithm for Counting Cars in Large-scale Video Surveillance Systems

Vehicle Speed Estimation Based on 3D ConvNets and Non-Local Blocks

Contact Info

Product

Resources

About