Unmanned aerial vehicle visual angle video semantic segmentation method based on deep learning

Abstract

The invention belongs to the field of unmanned aerial vehicle vision, and relates to an unmanned aerial vehicle view angle video semantic segmentation method based on deep learning. In order to solve the problem of image semantic segmentation, an encoder-decoder asymmetric network structure is designed, in the encoder stage, channel split and channel recombination are fused to improve a Bottleneck structure so as to carry out down-sampling and feature extraction, in the decoder stage, rich features are extracted and fused based on a spatial pyramid multi-feature fusion module, and in the multi-feature fusion module, the multi-feature fusion is carried out on the basis of the multi-feature fusion module. And finally, up-sampling is performed to obtain a segmentation result. And then, aiming at a video semantic segmentation problem, the image segmentation model designed by the invention is used as a segmentation module of video semantic segmentation, thereby improving a key frame selection strategy and performing feature transfer in combination with an optical flow method, reducing redundancy and accelerating the video segmentation speed.

Description

technical field [0001] The invention belongs to the field of unmanned aerial vehicle vision, and relates to a method for semantic segmentation of unmanned aerial vehicle perspective video based on deep learning. Background technique [0002] In the field of computer vision, the current main applications of neural networks are image recognition, target location and detection, and semantic segmentation. Semantic segmentation is a typical computer vision problem that involves taking as input some raw data (e.g. 2D images) and converting them into masks with highlighted regions of interest. In other words, in semantic segmentation we need to classify the visual input into different semantically interpretable categories, for example, we may need to classify all the pixels in the image that belong to cars and paint these pixels in a certain color . Today, semantic segmentation is one of the key issues in computer vision. More and more application scenarios require accurate and e...

AI Technical Summary

Problems solved by technology

[0005] For the first direction, Fayyaz et al. proposed to use the classic LSTM module to learn the temporal features of the video and assist in the propagation of spatial features, but this will seriously affect the running speed

Images