A Resource Adaptive Adjustment Method in Edge Computing Environment

Abstract

The invention discloses a resource self-adaptive adjustment method in an edge computing environment, comprising: dividing video stream data into windows and segments; selecting a reference configuration combination and an optional configuration set; and calculating the precision score of each configuration option value; Select the score threshold; comprehensively consider the configuration combinations not less than the threshold and the model calculation consumption, select the final configuration to apply to the window, and apply the window configuration to the corresponding windows of all acquisition devices in the group. This method takes into account various factors that affect the resource consumption of target detection applications in the edge computing environment. In addition, it also considers the adaptive adjustment of network bandwidth, and can flexibly change various parameters according to changes in the network environment to achieve bandwidth consumption and calculation. A balance of speed and precision.

Description

technical field [0001] The invention relates to the technical field of resource allocation in an edge computing environment, in particular to a resource adaptive adjustment method in an edge computing environment. Background technique [0002] With the rise and development of the industrial Internet of Things, the capabilities of information sensing devices such as sensors and cameras have been greatly improved, so that a large number of power videos with different needs and formats can be obtained. At present, the analysis and application of power video data mainly adopt the centralized cloud computing mode. Upload all visual data to the cloud computing center through network communication, and use the powerful computing power of the cloud computing center to store and analyze visual data. It has been practiced in many power scenarios such as transmission line defect detection and substation equipment identification. and apply. However, the power depth vision method based...

AI Technical Summary

Problems solved by technology

[0005] The most typical application of video stream data analysis is object detection based on deep neural network. However, due to the requirements of reliability and cost, the computing power of the video acquisition device itself is very low. Therefore, most or even all neural network operations need to be handed over to the edge. devices to process, but this also produces a large amount of video transmission bandwidth consumption, and with more and more acquisition devices, the computing power of edge servers is also limited, so it is necessary to find a suitable method to balance network bandwidth consumption and edge device computing response Speed ​​and Video Analysis Quality These Metrics

Images