Lane line detection method based on integrated learning cascade classifier

Abstract

The present invention discloses a lane line detection method based on an integrated learning cascade classifier. By the method, the accurate position and direction information of the lane lines in an image can be real-timely obtained on a single CPU or DSP, so that a lane line equation is obtained, at the same time, a certain robustness about the brightness change of the traffic scenes is realized. The detection method comprises the steps of firstly, erecting an image sensor to obtain color images of which the lane lines need to be extracted; then extracting an interested area according to a detection result of the previous frame; and then calculating an integration graph and the LBP characteristic of a single scale block; then adopting the integrated learning to traverse the interested area to obtain the patches of the lane lines; after the patches of the lane lines are obtained, finally using an optimization-based method to obtain the lane line equation.

Description

technical field [0001] The invention belongs to the technical field of advanced driving assistance systems, and specifically designs a lane line detection method based on an integrated learning cascade classifier. Background technique [0002] Advanced Driver Assistant System (ADAS) for short, uses cameras and other sensors installed on the car to collect environmental data inside and outside the car, and detects, recognizes and tracks static and dynamic objects, so that the driver can drive in the fastest time. The time to detect possible dangers to improve driving safety. Among them, the lane line detection technology uses image sensor data to judge whether there is a lane line and where the lane line appears. [0003] Existing lane detection methods can be divided into methods based on image processing and methods based on machine learning. Image processing-based methods use low-level features such as color, texture, shape, and geometry of lane lines to identify lane li...

AI Technical Summary

Problems solved by technology

These methods based on deep learning work well, but because each pixel has to go through multiple layers and a large number of convolution filtering operations, the amount of calculation and memory usage are very high, and it is not suitable for the application of embedded products in traffic scenarios.

Images