Research on Lane Line Segmentation Algorithm Based on Deeplabv3

In order to solve the problems of low accuracy of lane line detection caused by variation of light intensity, the blocking of pedestrians, vehicles and green plants, a lane line segmentation algorithm combining the enhanced network Retinexnet and the instance segmentation network Deeplabv3 was proposed. Firstly, Retinexnet network was used to enhance the original images to improve the image contrast and clarity. Then, lane line detection and segmentation were carried out based on the segmentation network Deeplabv3, at the same time, the shape analysis algorithm was introduced. Experimental results showed that the proposed method could segment the lane lines well in the changeable road environment and at night, and the forward detection rate of the proposed method could reach up to 94.6%, which improved the detection accuracy by about 2 percentage points compared with direct Deeplabv3 network.