Detecting Anomalous Trajectories via Recurrent Neural Networks

Cong Ma, Zhenjiang Miao, Min Li, Shaoyue Song, Ming Hsuan Yang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Detecting anomalies from trajectory data is an important task in video surveillance. However, it is difficult to give a precise definition of this term since trajectory data obtained from different camera views may vary in shape, direction, and spatial distribution. In this paper, we propose trajectory distance metrics based on a recurrent neural network to measure similarities and detect anomalies from trajectory data. First, we use an autoencoder to capture the dynamic features of a trajectory. The distance between two trajectories is defined by the reconstruction errors based on the learned models. We then detect anomalies based on the nearest neighbors using the proposed metric. As such, we can deal with various kinds of anomalies in different scenes and detect anomalous trajectories in either a supervised or unsupervised manner. Experiments show that the proposed algorithm performs favorably against the state-of-the-art anomaly detections on the benchmark datasets.

Original languageEnglish
Title of host publicationComputer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers
EditorsGreg Mori, Hongdong Li, C.V. Jawahar, Konrad Schindler
PublisherSpringer Verlag
Pages370-382
Number of pages13
ISBN (Print)9783030208691
DOIs
Publication statusPublished - 2019 Jan 1
Event14th Asian Conference on Computer Vision, ACCV 2018 - Perth, Australia
Duration: 2018 Dec 22018 Dec 6

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11364 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th Asian Conference on Computer Vision, ACCV 2018
CountryAustralia
CityPerth
Period18/12/218/12/6

Fingerprint

Recurrent neural networks
Recurrent Neural Networks
Anomalous
Trajectories
Trajectory
Anomaly
Video Surveillance
Distance Metric
Anomaly Detection
Similarity Measure
Spatial Distribution
Spatial distribution
Nearest Neighbor
Camera
Cameras
Vary
Benchmark
Metric
Term
Experiment

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ma, C., Miao, Z., Li, M., Song, S., & Yang, M. H. (2019). Detecting Anomalous Trajectories via Recurrent Neural Networks. In G. Mori, H. Li, C. V. Jawahar, & K. Schindler (Eds.), Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers (pp. 370-382). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11364 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-20870-7_23
Ma, Cong ; Miao, Zhenjiang ; Li, Min ; Song, Shaoyue ; Yang, Ming Hsuan. / Detecting Anomalous Trajectories via Recurrent Neural Networks. Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. editor / Greg Mori ; Hongdong Li ; C.V. Jawahar ; Konrad Schindler. Springer Verlag, 2019. pp. 370-382 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{790c1e68ad0049b0b1b96e2dc9bb992f,
title = "Detecting Anomalous Trajectories via Recurrent Neural Networks",
abstract = "Detecting anomalies from trajectory data is an important task in video surveillance. However, it is difficult to give a precise definition of this term since trajectory data obtained from different camera views may vary in shape, direction, and spatial distribution. In this paper, we propose trajectory distance metrics based on a recurrent neural network to measure similarities and detect anomalies from trajectory data. First, we use an autoencoder to capture the dynamic features of a trajectory. The distance between two trajectories is defined by the reconstruction errors based on the learned models. We then detect anomalies based on the nearest neighbors using the proposed metric. As such, we can deal with various kinds of anomalies in different scenes and detect anomalous trajectories in either a supervised or unsupervised manner. Experiments show that the proposed algorithm performs favorably against the state-of-the-art anomaly detections on the benchmark datasets.",
author = "Cong Ma and Zhenjiang Miao and Min Li and Shaoyue Song and Yang, {Ming Hsuan}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-20870-7_23",
language = "English",
isbn = "9783030208691",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "370--382",
editor = "Greg Mori and Hongdong Li and C.V. Jawahar and Konrad Schindler",
booktitle = "Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers",
address = "Germany",

}

Ma, C, Miao, Z, Li, M, Song, S & Yang, MH 2019, Detecting Anomalous Trajectories via Recurrent Neural Networks. in G Mori, H Li, CV Jawahar & K Schindler (eds), Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11364 LNCS, Springer Verlag, pp. 370-382, 14th Asian Conference on Computer Vision, ACCV 2018, Perth, Australia, 18/12/2. https://doi.org/10.1007/978-3-030-20870-7_23

Detecting Anomalous Trajectories via Recurrent Neural Networks. / Ma, Cong; Miao, Zhenjiang; Li, Min; Song, Shaoyue; Yang, Ming Hsuan.

Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. ed. / Greg Mori; Hongdong Li; C.V. Jawahar; Konrad Schindler. Springer Verlag, 2019. p. 370-382 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11364 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Detecting Anomalous Trajectories via Recurrent Neural Networks

AU - Ma, Cong

AU - Miao, Zhenjiang

AU - Li, Min

AU - Song, Shaoyue

AU - Yang, Ming Hsuan

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Detecting anomalies from trajectory data is an important task in video surveillance. However, it is difficult to give a precise definition of this term since trajectory data obtained from different camera views may vary in shape, direction, and spatial distribution. In this paper, we propose trajectory distance metrics based on a recurrent neural network to measure similarities and detect anomalies from trajectory data. First, we use an autoencoder to capture the dynamic features of a trajectory. The distance between two trajectories is defined by the reconstruction errors based on the learned models. We then detect anomalies based on the nearest neighbors using the proposed metric. As such, we can deal with various kinds of anomalies in different scenes and detect anomalous trajectories in either a supervised or unsupervised manner. Experiments show that the proposed algorithm performs favorably against the state-of-the-art anomaly detections on the benchmark datasets.

AB - Detecting anomalies from trajectory data is an important task in video surveillance. However, it is difficult to give a precise definition of this term since trajectory data obtained from different camera views may vary in shape, direction, and spatial distribution. In this paper, we propose trajectory distance metrics based on a recurrent neural network to measure similarities and detect anomalies from trajectory data. First, we use an autoencoder to capture the dynamic features of a trajectory. The distance between two trajectories is defined by the reconstruction errors based on the learned models. We then detect anomalies based on the nearest neighbors using the proposed metric. As such, we can deal with various kinds of anomalies in different scenes and detect anomalous trajectories in either a supervised or unsupervised manner. Experiments show that the proposed algorithm performs favorably against the state-of-the-art anomaly detections on the benchmark datasets.

UR - http://www.scopus.com/inward/record.url?scp=85066870479&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85066870479&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-20870-7_23

DO - 10.1007/978-3-030-20870-7_23

M3 - Conference contribution

AN - SCOPUS:85066870479

SN - 9783030208691

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 370

EP - 382

BT - Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers

A2 - Mori, Greg

A2 - Li, Hongdong

A2 - Jawahar, C.V.

A2 - Schindler, Konrad

PB - Springer Verlag

ER -

Ma C, Miao Z, Li M, Song S, Yang MH. Detecting Anomalous Trajectories via Recurrent Neural Networks. In Mori G, Li H, Jawahar CV, Schindler K, editors, Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. Springer Verlag. 2019. p. 370-382. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-20870-7_23