A lane line detection method based on multi-task network

Abstract

A lane line detection method based on multi-task network. The invention discloses a lane line detection method based on deep learning target detection and fitting regression, including: (1) extracting robust features from images through forward propagation in a convolutional neural network (referred to as CNN) expression; (2) The two fully connected layers of the network use the extracted features to return the position of the possible target, and at the same time make a judgment whether the area is a lane line; (3) filter the detected small lane line; ( 4) Fitting the filtered results to n segments of lane lines. The method provided by the present invention uses the convolutional neural network to extract image features, which can determine the position of the lane line more accurately than traditional features, and has better accuracy; Even if there is a small amount of false detection, it can be eliminated by filtering isolated detection results, which improves the robustness of the method of the present invention.

Description

technical field [0001] The invention belongs to the field of image processing and pattern recognition, and relates to a lane line detection method based on multi-task network target detection and regression fitting. Background technique [0002] Traffic safety is always an important area that people are very concerned about. Many traffic accidents cause huge casualties and property losses every year. In recent years, with the development of deep neural networks and the rapid development of computer vision, many manufacturers have begun to try to develop advanced driver assistance systems (Advanced Driver Assistant Systems, referred to as ADAS) through computer vision methods, among which lane line detection is an important The components became the focus of the study. Because computer vision technology has many advantages such as good detection effect, wide application range and low cost, it has become a mainstream technology in the field of lane detection in recent years. ...

AI Technical Summary

Problems solved by technology

This type of method is first of all efficiency, because the huge computational cost of the sliding window will seriously slow down the running time

Images