Foggy weather visibility identification method and system under complex background

Abstract

The invention discloses a foggy weather visibility identification method and system under a complex background. The method comprises the following steps: acquiring a real foggy weather picture data set; establishing an end-to-end convolutional neural network model, performing training and region calibration on the input foggy day picture data set, and outputting a transmissivity graph; optimizingthe transmissivity graph by using a guided filtering method to obtain a refined transmissivity graph; estimating an atmospheric scattering coefficient according to the transmissivity graph; and solving a regional average atmospheric scattering coefficient, and carrying out grade identification on the foggy day picture. The method is suitable for foggy day visibility identification under a complexbackground in the fields such as meteorological stations, highways, tourist attractions, alpine regions and complex riverways; to a certain extent, no manual judgment is required, the limitation thattraditional manual observation needs expensive hardware equipment such as visibility meters is broken through, and compared with highway foggy weather visibility judgment, the method is wide in application range.

Description

technical field [0001] The invention relates to an intelligent grading recognition method and system for visibility in foggy days under complex backgrounds, and belongs to the technical field of foggy day camera recognition. Background technique [0002] At present, the image dehazing problem has achieved many research results in the scientific research and industrial circles, and has been applied to the video surveillance of highways, aerial photography and other related fields. Under real fog conditions, severe weather conditions will cause various safety hazards in life, which are related to the safety of people's lives and property. If the relevant meteorological department can accurately release the corresponding visibility conditions in foggy days, it will help all walks of life to improve their visibility. Manage quality. At this stage, the observation of fog visibility by the meteorological department is mainly based on manual observation, and the fog conditions are...

AI Technical Summary

Problems solved by technology

In the prior art, there are two main types of methods for judging visibility in foggy weather: the first type of method is a foggy weather detection method based on image space features, this type of method requires a large number of threshold judgments, and threshold adjustment according to different road section information has achieved good results. effect, with poor robustness

The second type of method is a fog detection method based on the dark channel prior. This type of method needs to calculate the image transmittance to further obtain the corresponding visibility value. The image transmittance obtained by this method is not fine enough, and the amount of calculation is large, which will affect the visibility value. accuracy

The convolutional neural network in deep learning has high accuracy and effectiveness for image feature extraction. At present, there are few studies on the judgment of fog levels using deep learning methods at home and abroad, which require a large amount of foggy image data sets, and The existing technologies all adopt the technical scheme of front-end collection and back-end identification

Images