Lane detection method based on deep learning combined with post-processing
In recent years,deep learning technology has made remarkable progress in image recognition,image processing and other fields,making it possible to perceive the environment in driverless technology.Combined with the existing deep learning technology,the corresponding research on the realization of real-time lane line recogni-tion in complex traffic scenes is carried out,and a lane line detection method based on deep learning semantic seg-mentation algorithm combined with post-processing is proposed.This paper optimizes the convolutional structure of the EfficientNetV2 network,and introduces an information fusion structure with dual branches sharing low-level features to make it suitable for semantic segmentation tasks.At the same time,a new multiscale convolution fusion module is designed,which is used as an enhanced feature extraction structure,and a lane line recognition network Eff-SCNN is further proposed.Finally,the recognition results are post processed to detect lane line instances.Ex-perimental results show that the method can correctly predict lane lines in different traffic scenarios.
万方数据
维普
