A Novel Lane Detection System Using Dynamic V...

A Novel Lane Detection System Using Dynamic Vision Approach

Author Name : Mrs. S. L.Soniya, P. Indira

ABSTRACT Lane detection is a challenging problem. It has attracted the attention of the computer vision community for several decades. Essentially, lane detection is a multi-feature detection problem that has become a real challenge for computer vision and machine learning techniques. Although many machine learning methods are used for lane detection, they are mainly used for classification rather than feature design. But modern machine learning methods can be used to identify the features that are rich in recognition and have achieved success in feature detection tests. With increase in the number of road accidents, it has led to concern over the nature of accidents. Most of the time, it is due to human error. So the lane detection systems are being developed for assisting the driver. The main purpose of it is to detect the lanes and warn the driver of lane departure. The proposed system is a novel lane detection system, called Scene Understanding Physics-Enhanced Real-time (SUPER) algorithm. The proposed method consists of two main modules: Hierarchical Semantic Segmentation network as the scene feature extractor and a physics enhanced multi-lane parameter optimization module for lane inference. It includes an optimization framework to estimate lane parameters. The hierarchical structure extracts the objects in each level and this structure tries to label an image of a street scene into coarse geometric classes that are useful for tasks such as navigation, object recognition, and general scene understanding. Assuming that lane markings are largely parallel polynomials, it separate the lane parameters into shared parts (heading angle and curvature) and unique parts (offsets and lane marking attributes). To cope with non-flat ground, a polynomial road model is adopted. Then the lane parameter estimation problem is formulated as an optimization problem based on the pixel-wise scene labels, solving which can efficiently estimate all parameters.