Severe weather image classification enhancement method based on convolution model and feature fusion

Abstract

The invention relates to a severe weather image classification enhancement method based on a convolution model and feature fusion. The method comprises the following steps: S1, establishing a weatherimage set; s2, constructing an image classification fusion model; s3, training an image classification fusion model through the weather image set; s4, inputting weather images to be classified; s5, the image classification fusion model classifies the weather images to be classified to obtain image types of the weather images to be classified, if the image types are clear, the step S7 is executed,and otherwise, the step S6 is executed; s6, selecting an image enhancement algorithm corresponding to the image type to enhance the weather image to be classified, obtaining an enhanced image, and inputting the obtained enhanced image into the step S5 as the weather image to be classified; and S7, outputting a weather image. Compared with the prior art, the improved VGG16 model is adopted, traditional features and depth features are fused, the recognition accuracy is high, the corresponding algorithm is adopted to enhance the image, and the enhancement effect is good.

Description

technical field [0001] The invention relates to the field of severe weather image enhancement, in particular to a method for classifying and enhancing severe weather images based on convolution models and feature fusion. Background technique [0002] Low-quality images captured under severe weather conditions such as smog and heavy rain are often difficult to identify and analyze due to heavy fog, rain streaks, and raindrops on the screen. This low-quality image will bring great difficulty to surveillance and other systems that require image recognition. The key to dealing with this situation is how to classify and enhance such images. [0003] Chinese patent CN201610079472.X discloses a low-quality image enhancement method under extreme weather conditions. For a single input image, the method first converts the image to the CIE-Lab color space, and sets a color cast factor D. According to experience , if D ≤ 1.4, the image is a clear image and no processing is required; if...

AI Technical Summary

Problems solved by technology

This method is a classification enhancement method for distinguishing fog and haze images from rain and snow images according to the chrominance component value. This method uses a purely physical algorithm, which can distinguish few image categories and has poor robustness, and it is extremely prone to misjudgment.

Images