- Tytuł:
- Lane queues length estimation using single 2D camera and image depth estimation algorithm
- Autorzy:
-
Bolek, K.
Liver, K. - Powiązania:
- https://bibliotekanauki.pl/articles/1833868.pdf
- Data publikacji:
- 2020
- Wydawca:
- Polskie Stowarzyszenie Telematyki Transportu
- Tematy:
-
queues length estimation
RetinaNet
monodepth2
szacowanie długości kolejek - Opis:
- During last 10 years the technology connected to video streaming analytics and deep learning algorithms connected to Intelligent Transport System and later, initiative of worldwide concept called “Smart City” pushed in reality a telematics into direction of IoT and possible use of neural network to support and create an added value in the way the Traffic Management Systems are collecting raw date in a massive data streaming mode, to be able to manage in an optimized and wright way big cities transportation networks. Recently HD cameras started to be one of the more widely used detectors replacing inductive loops and radars. The observations during productive life time of this kind of sensors on the field, created other challenges, efficiency aspects and unfavourable cost structure as consequence of usage of this product (HD camera), on such a big scale. Cameras designed to detect the length of queues at intersections require configuration by the operator. This process is extremely arduous, tedious and time consuming , considering how many cameras are located in cities. In addition, the configuration has to be done by a trained person, and what’s more, moving or replacing the camera involves its manual reconfiguration. The aim of this article is to present a prototype of an algorithm that uses RetinaNet neural network to detect bikes/motorbikes on the street, and using monodepth2 determines the length of the queue and autonomously determines the direction of cars on the road. All the work undertaken confirmed that the approach used is effective and additionally allows to limit the operator’s work only to defining the focal length and size of the camera sensor.
- Źródło:
-
Archives of Transport System Telematics; 2020, 13, 1; 6--9
1899-8208 - Pojawia się w:
- Archives of Transport System Telematics
- Dostawca treści:
- Biblioteka Nauki
Artykuł