Implementation Of Image Dehazing Algorithm Based On Raspberry Pi

Optical imaging is often affected by fog and other weather.It makes images to be foggy,incomplete information,inability to extract key information,and adversely affecting normal work.This dissertation aims at the problem of foggy image degradation,studying with the mechanism of foggy image degradation and the degeneracy model deeply.Based on the degeneracy model,the methods can be divided into two parts:the classical algorithm method and the dark channel prior dehazing method.This dissertation summarizes the classical algorithm and focuses on the dark channel prior defogging method,in combination with deep learning,and the DehazeNet is used to generate transmittance maps.An improved deep learning dehaze algorithm.based on the Caffe framework is attempted to run on the Raspberry Pi.The main research work of the paper is as follows:(1)Introduce the research status of the imaging mechanism in a fog day.(2)Analyzing the advantages and disadvantages of the image dehazing method,we deeply study the theory of deep learning,and improve the dehaze algorithm of convolutional neural network.Based on the establishment of training sets,network pre-training,the network was tested,and the evaluation of the processing effect of the algorithm qualitied.(3)The Raspberry Pi 3B-based image acquisition system was designed.It is consisted of a dedicated platform with a dedicated camera,and VNC is used as a remote connection software.Finally,the image acquisition system was established,and the Python language was used to control the system to achieve image acquisition,making the DehazeNet algorithm run on the embedded platform Raspberry Pi.