Smart camera for traffic applications

Smart camera for traffic applications

Článek WoS

In this paper, we present a compact intelligent camera with traffic surveillance and enforcement as a primary application field. The unique features of the camera are 1/ embedded object detection of user defined objects and 2/ excellent image quality achieved by a custom HDR image acquisition and tone mapping. Intelligent cameras are important parts of equipment in production, control, security and especially traffic surveillance as well as enforcement. In many applications, they serve as a versatile source of images processed elsewhere and reduction of required processing and communication achieved by embedded (pre)processing within the intelligent camera is always welcome. The key blocks of the camera are implemented in FPGA to reduce the CPU load and thus also the power requirements. The control software uses an embedded CPU running  Linux. The camera is designed to provide key images interesting from the application point of view and to provide metadata. On the camera prototype, we observed total power consumption of 8 W and object detection accuracy > 99 %. The camera design and functionality can be extensively modified to harness a wide variety of applications, such as traffic surveillance and enforcement, quality control, and others.

In this paper, we present a compact intelligent camera with traffic surveillance and enforcement as a primary application field. The unique features of the camera are 1/ embedded object detection of user defined objects and 2/ excellent image quality achieved by a custom HDR image acquisition and tone mapping. Intelligent cameras are important parts of equipment in production, control, security and especially traffic surveillance as well as enforcement. In many applications, they serve as a versatile source of images processed elsewhere and reduction of required processing and communication achieved by embedded (pre)processing within the intelligent camera is always welcome. The key blocks of the camera are implemented in FPGA to reduce the CPU load and thus also the power requirements. The control software uses an embedded CPU running  Linux. The camera is designed to provide key images interesting from the application point of view and to provide metadata. On the camera prototype, we observed total power consumption of 8 W and object detection accuracy > 99 %. The camera design and functionality can be extensively modified to harness a wide variety of applications, such as traffic surveillance and enforcement, quality control, and others.

smart camera, object detection, WaldBoost, ACF, HDR, tonemapping, traffic surveillance

smart camera, object detection, WaldBoost, ACF, HDR, tonemapping, traffic surveillance

JURÁNEK, R.; MUSIL, P.; MUSIL, M.; NOSKO, S.; ZEMČÍK, P.

2024

18.07.2023

1939-8115

Journal of Signal Processing Systems for Signal Image and Video Technology

95

9

Spojené státy americké

1067

1077

11

https://link.springer.com/article/10.1007/s11265-023-01843-1