Deep learning image compressed sensing algorithm and system based on Internet of Vehicles

Abstract

The invention relates to a deep learning image compressed sensing algorithm based on the Internet of Vehicles, and the algorithm comprises the steps: partitioning compressing and sampling a picture obtained by an Internet of Vehicles intelligent terminal through a measurement matrix, and obtaining a measurement data set Y; transmitting the measurement data set Y to a cloud server in a wireless manner, and importing the measurement data set Y into a trained deep neural network model, i.e., recovering the measurement data set Y into intermediate reconstruction through a full connection layer toobtain a preliminary reconstructed image X '; reconstructing the image through a residual denoising network to obtain a final reconstructed image X ''; and determining a loss function of the originalimage, and reversely transmitting the updated network weight by the loss function to realize optimization. The invention further discloses a deep learning image compressed sensing system based on theInternet of Vehicles. The data transmission quantity is greatly reduced, the bandwidth pressure is reduced, the flow charge and the limitation of the storage space are greatly saved, and meanwhile, the real-time response recovery, the accuracy of the reconstructed image and the suppression of the image noise can achieve better effects.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a deep learning image compression sensing algorithm and system based on Internet of Vehicles. Background technique [0002] With the construction of the Internet of Vehicles cloud data center and comprehensive service platform and the application of the industry, a large amount of vehicle visual data, location data, vehicle status, faults, environmental data, speed, acceleration and other driver behavior data, road network data and Visual data provides a basic data source for analyzing the behavior of moving objects. In the development process of building end-cloud integration of IoV big data, secure data transmission, traffic charges, efficiency, etc. are factors that need to be considered urgently. [0003] Images and videos, as important carriers for the material reproduction and recording of human visual information, are generally represented by discrete values ​​after...

AI Technical Summary

Problems solved by technology

Reconstruction of compressed sensing based on neural network, when a compressed signal with a large dimension is input into the network, a fully connected layer and a highly deep convolutional neural network require a large number of parameters, and a large number of parameters are prone to gradient explosion or gradient disappearance. , unable to complete a high-quality reconstruction

Images