Driver Anomaly Detection Using Skeleton Images

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F23%3A10254632" target="_blank" >RIV/61989100:27240/23:10254632 - isvavai.cz</a>

Výsledek na webu

<a href="https://link.springer.com/chapter/10.1007/978-3-031-47969-4_36" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-031-47969-4_36</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-031-47969-4_36" target="_blank" >10.1007/978-3-031-47969-4_36</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Driver Anomaly Detection Using Skeleton Images

Popis výsledku v původním jazyce

Many unexpected situations can occur while driving that may lead to dangerous accidents. Some of them may be caused by sudden health problems (e.g. heart attack, stroke, total collapse) or by driver inattention (e.g. microsleep, visual distraction). This has motivated the need for developing the methods that are able to monitor the driver&apos;s state in the first step and to prevent the accidents in the second step (e.g. by activating an acoustic signal, or even by taking over driving). In this paper, we propose a method that can be used for detecting the abnormal driving situations. Our approach is based on two main steps. In the first step, the MNIST-like skeleton images are created with the use of human pose detector. In the second step, an appropriate neural network is used for the final classification. Since we also include the anomalies consisting in an unusual trajectory of a certain body part (not only an unusual shape of body, which can be detected from the isolated images), short sequences of images are examined. The LSTM (long short-term memory) autoencoder is used as a main network architecture. The experiments that are presented show that the proposed method achieves better results than other compared methods.

Název v anglickém jazyce

Driver Anomaly Detection Using Skeleton Images

Popis výsledku anglicky

Many unexpected situations can occur while driving that may lead to dangerous accidents. Some of them may be caused by sudden health problems (e.g. heart attack, stroke, total collapse) or by driver inattention (e.g. microsleep, visual distraction). This has motivated the need for developing the methods that are able to monitor the driver&apos;s state in the first step and to prevent the accidents in the second step (e.g. by activating an acoustic signal, or even by taking over driving). In this paper, we propose a method that can be used for detecting the abnormal driving situations. Our approach is based on two main steps. In the first step, the MNIST-like skeleton images are created with the use of human pose detector. In the second step, an appropriate neural network is used for the final classification. Since we also include the anomalies consisting in an unusual trajectory of a certain body part (not only an unusual shape of body, which can be detected from the isolated images), short sequences of images are examined. The LSTM (long short-term memory) autoencoder is used as a main network architecture. The experiments that are presented show that the proposed method achieves better results than other compared methods.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

ADVANCES IN VISUAL COMPUTING, ISVC 2023, PT I

ISBN

978-3-031-47968-7

ISSN

0302-9743

e-ISSN

1611-3349

Počet stran výsledku

13

Strana od-do

459-471

Název nakladatele

SPRINGER INTERNATIONAL PUBLISHING AG

Místo vydání

CHAM

Místo konání akce

Lake Tahoe, NV, USA

Datum konání akce

16. 10. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

001159734600036