Lane line detection and segmentation method based on attention space convolutional neural network

Abstract

The invention discloses a lane line detection and segmentation method based on an attention space convolutional neural network, and the method comprises the steps: segmenting each lane line in a lane image into one class through the space convolutional neural network, outputting a probability graph of the lane lines, wherein the numerical value of each pixel point in the graph is the probability that the point belongs to each class; and finally, connecting each type of pixel points through cubic spline interpolation to obtain a lane line. According to the method, a spatial convolution layer is embedded into a traditional convolutional neural network, and an attention gate structure is introduced, so that spatial information in an image can be propagated in neurons on the same layer, the structured information can be better extracted, and under the condition that the calculation amount of a network model is not increased, the calculation complexity of the network model is reduced. And the detection capability of long-distance continuous targets such as lane lines is enhanced. The method mainly aims at a structured road with clear road marker lines, can detect and segment complete lane lines in complex driving conditions such as crowding, dim and narrow, and the performance of the method is superior to that of an existing method.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to an attention mechanism-based spatial convolutional neural network lane line detection and segmentation method. Background technique [0002] Lane line detection is the basic link of many advanced driving assistance functions such as deviation warning, lane keeping and automatic lane change, and plays a vital role in the automatic driving system. The stability and accuracy of the lane line detection method affect the automatic performance of the driving system. Although existing lane line detection methods can meet the usage requirements, it is still a challenging task to achieve accurate detection in real driving scenarios. The main difficulty is that there are various interference factors in the actual environment, such as bad weather, light changes, traffic congestion, etc., resulting in incomplete target information detected by the sensor, which greatly increases the ...

AI Technical Summary

Problems solved by technology

[0005] Traditional computer vision technology is not ideal for the detection of lane lines, because most of the model-based and feature-based methods have strict assumptions, are not universal, and can only target certain characteristics in specific scenarios. The lane line completes the detection task

Images