Hajime Kanzaki

Video streaming schemes for industrial IoT

Wireless video applications for Industrial Internet of Things (IoT) are expanding into a multitude of new services. In the example of cloud processing for visual object detection, a camera is connected to the cloud via a local server and a data network, allowing the processing load to be handled in a distributed manner. This service model heavy taxes the data network with potentially unneeded traffic, thus degrading the overall quality of service for all users on the network. Edge computing techniques mitigate the degradation of service quality by partially processing the sensor data at the local server before the data is transmitted to the cloud. This is done according to the level of interest of the captured data which is categorized by machine learning algorithms. However, conventional edge computing is not optimally efficient as further recognition attributes of the captured object data are not considered. This paper presents a model that adds control of the camera video rate by considering the attributes of captured object. We then investigate cost trade-offs using dynamic programming, and evaluates the behavior of proposed method under wireless channel condition using NS-3 simulations. Our results show that by adding intelligent adaptive video rate control to the cloud processing of video data capture can reduce overall system power use while improving system efficiency and subsequently network throughput.

Video Streaming Schemes for Industrial IoT

Wireless video applications for Industrial Internet Of Things (IoT) are expanding into a multitude of new services. In the example of cloud processing for visual object detection, a camera is connected to the cloud via a local server and a data network, allowing the processing load to be handled in a distributed manner. This service model heavy taxes the data network with potentially unneeded traffic, thus degrading the overall quality of service for all users on the network. Edge computing techniques mitigate the degradation of service quality by partially processing the sensor data at the local server before the data is transmitted to the cloud. This is done according to the level of interest of the captured data which is categorized by machine learning algorithms. However, conventional edge computing is not optimally efficient as further recognition attributes of the captured object data are not considered. This paper presents a model that adds control of the camera video rate by considering the attributes of captured object. We then investigate cost trade-offs using dynamic programming, and evaluates the behavior of proposed method under wireless channel condition using NS-3 simulations. Our results show that by adding intelligent adaptive video rate control to the cloud processing of video data capture can reduce overall system power use while improving system efficiency and subsequently network throughput.