Unmanned aerial vehicle single image rain removing method based on convolutional neural network

Abstract

The invention discloses an unmanned aerial vehicle single image rain removing method based on a convolutional neural network, and the method comprises the steps: firstly enabling a rain image to be separated into a low-frequency background image and a high-frequency rain line image through a guide filter; secondly, performing sub-channel feature extraction on the low-frequency background image and the high-frequency rain line image by using a multi-scale CNN convolutional neural network; re-allocating weights to the feature maps extracted by the multi-scale CNN convolutional neural network through a channel domain attention mechanism; then, splicing the feature maps of the plurality of channels through a full connection layer, and constructing a mapping relationship between the rain image and the rain-removed image; and finally, respectively outputting high-frequency and low-frequency rain-removed images, and fusing the high-frequency and low-frequency rain-removed images to output a complete rain-removed image. The method has the characteristic of efficiently removing the rain line on the rainy day image shot in the flight process of the unmanned aerial vehicle.

Description

technical field [0001] The present invention relates to the technical field of computer vision, and more specifically, to a method for removing rain from a single image of a UAV based on a convolutional neural network. Background technique [0002] UAVs are widely used in today's society, whether it is military unmanned reconnaissance drones or target drones, or civilian aerial photography, agriculture, plant protection, express transportation, disaster rescue, power inspection, disaster relief, film and television shooting, etc. , have greatly expanded the application field of UAVs, and both reconnaissance and self-navigation require the help of sufficiently clear images. However, UAVs are easily affected by harsh weather environments, such as rain and snow If there is no weather, the visual image with poor effect will be obtained, which not only affects the observation but also easily affects its own visual navigation system. Therefore, an effective UAV rain image removal ...

AI Technical Summary

Problems solved by technology

[0002] UAVs are widely used in today's society, whether it is military unmanned reconnaissance drones or target drones, or civilian aerial photography, agriculture, plant protection, express transportation, disaster rescue, power inspection, disaster relief, film and television shooting, etc. , have greatly expanded the application field of UAVs, and both reconnaissance and self-navigation require the help of sufficiently clear images. However, UAVs are easily affected by harsh weather environments, such as rain and snow If the weather is different, the visual image with poor effect will be obtained, which not only affects the observation but also easily affects its own visual navigation system. Therefore, an effective UAV rain image removal algorithm is very important.

Images