Energy consumption-accuracy controllable sensor dual-parameter distribution system

Abstract

The invention provides an energy-consumption -accuracy controllable sensor dual-parameter distribution system comprising a sensor and a server which are connected via a transmission channel, wherein atraining mode of the system statistically analyzes internal relations between performance indexes such as coding energy consumption, transmission energy consumption and scene classification, and twocontrol parameters, and builds an accuracy parameter optimization model between an accuracy threshold and an (S, Q) control parameter through analyzing how a sampling factor S and a quantization parameter Q influence characteristics such as coding complexity, a bit rate and image quality, of a monitoring image. The working mode of the system can preset a proper (S, Q) control parameter according to an accuracy parameter optimization model, to gradually adjust energy consumption of the sensor under the constraint of classification accuracy, and acquire approximately optimal energy-consumption-accuracy control performance. According to the system provided by the system, a computing method for the control parameter is provided through testing the characteristics such as the coding complexity,bit rate and accuracy, of intelligent wireless image monitoring, and an energy consumption-accuracy controllable mechanism is provided for a monitoring and imaging application.

Description

technical field [0001] The invention relates to a dual-parameter allocation system that minimizes sensor energy consumption under the constraints of classification accuracy, and belongs to the technical field of image communication. Background technique [0002] In intelligent wireless image monitoring, battery-powered sensors can actively collect and encode scene images, and transmit compressed code streams to servers through wireless channels for automatic scene classification. Intelligent wireless image monitoring has many important applications in security monitoring, battlefield intelligence, environmental perception and other fields. For example, the sensor is deployed in a wild animal protection area. When a certain scene change triggers a sensor trigger, the sensor immediately acquires several images of the monitored scene, compresses them and sends them to the server for scene classification, and automatically determines whether the scene image is a person or an ani...

AI Technical Summary

Problems solved by technology

In the process of encoding and transmitting image samples, the sensor may waste too much computing and energy resources, resulting in the quality of the decoded image being much higher than the quality required for scene classification

In recent years, researchers have realized the importance of jointly optimizing image compression and scene classification. However, there is no literature on how to jointly design image compression, wireless transmission, and scene classification.

Images