Event camera lane line extraction method based on deep learning

Abstract

The invention provides an event camera lane line extraction method based on deep learning. The invention provides a network based on structure prior, and the network can well capture the spatial relation between pixels through employing an omnidirectional slice convolution module, and especially can capture the spatial relation of a target in a slender shape. In order to further improve the lane line extraction precision, a polynomial fitting post-processing method based on model Carlo sampling and least squares is introduced to fit the lane line, and finally the lane line extraction task is completed.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a deep learning-based event camera lane line extraction method. Background technique [0002] Lane line extraction is a basic and important task in the field of automatic driving. In recent years, advanced lane line extraction methods have used deep learning models. These deep learning methods usually use RGB images, that is, ordinary optical camera images. However, ordinary optical images naturally have the disadvantages of motion blur and small dynamic range due to the imaging mechanism. In order to solve these problems, the invention introduces a dynamic vision sensor (DVS, an event camera) with the advantages of low latency and high dynamic range, and constructs a data set for lane line extraction tasks. In order to extract lane lines well on DVS images, the present invention proposes a network based on structure prior. The network can well capture the spatial rel...

AI Technical Summary

Problems solved by technology

[0003] In view of the defects in the prior art, the purpose of the present invention is to provide a method for extracting lane lines of event cameras based on deep learning, which can better solve the problem of poor imaging quality caused by ordinary optical cameras in some harsh environments (such as tunnel entrances and exits). The problem that leads to the difficulty of lane line extraction, the proposed network based on the structure prior can extract the lane lines on the DVS image with high precision

Images