Lane line detection system and method combining double-branch network and user-defined function network

Abstract

The invention discloses a lane line detection system and method combining a double-branch network and a user-defined function network. The method comprises the steps of improving a classic convolutional neural network structure, reconfiguring a coder-decoder composition structure, and forming a double-branch network detection lane line; and training a convolutional network with a self-defined lossfunction, allowing the transformation matrix parameters to be adjusted when the road gradient changes, and constructing a self-defined function network to fit the lane pixels. According to the method, the double-branch network comprises a lane segmentation branch and a lane embedding branch, wherein the segmentation branch divides an image into a background part and a lane part, the lane embedding branch divides the lane pixels into different examples, and the method for detecting fixed lane lines is improved. According to the method, the user-defined function network generates the convertible matrix fitting lane pixels, so that the problem of poor fitting of distant lane pixels is solved. According to the present invention, the lane line is detected by adopting the double-branch networkand the user-defined function network, so that the technical problem of low detection accuracy under complex road conditions is solved.

Description

technical field [0001] The invention belongs to the technical field of unmanned driving, and specifically relates to a lane line detection system and method combined with a double-branch network and a self-defined function network, which is suitable for detection of vehicle lane changes and changes in road plane slopes. Background technique [0002] As an important link in the field of unmanned driving and an important part of intelligent assisted driving, lane line detection plays an important role in reducing road traffic accidents, and has been highly valued by researchers and engineering practitioners. [0003] Lane line detection is defined as follows: by detecting the input image, then using a series of feature extraction algorithms to extract the lane line pixel features, using high-order polynomial fitting to extract pixels from the lane, and finally displaying in the original input image and outputting the image. The purpose of lane line detection is to locate the e...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the shortcomings of the existing lane line detection technology, the present invention provides a lane line detection method combined with a double-branch network and a self-defined function network, which can effectively solve the problem of lane changes, lane line occlusion, and road plane gradient changes. In order to adapt to the detection of complex road conditions and improve the robustness and real-time performance of lane line detection

Images