Road lane detection is a fundamental component of the vision-based selfdriving automobile system.It serves as a basis for more advanced course planning and control operations such as braking and steering.In feature-based lane identification,the results of lane edge detection are critical.The edges left by tough conditions（illumination and shadows）in the form of noise on roadways interfere with accurate lane line detection,providing a severe impediment.This work proposes an improved algorithm to take advantage of edge orientations for robust lane detection based on edge detection with symmetric molecules,which exploits the symmetric properties of directionally sensitive analytical functions in multi-scale systems,adaptive HSI colour model,perspective transformations,and histogram analysis.This final approach may be used to identify straight and curved lane lines.The simulation results from various scenarios are shown to show how the proposed lane identifying approach outperforms previous methods.