Lane Detection Using Orientation of Gradient and Vehicle Network Signals

The continuous growth of market for Advanced Driver Assistance Systems based on image processing features leads to the advance of the applied techniques, increasing thus the driving safety. Mostly of the edge detection algorithms are traditional approaches, and to achieve improvements it is necessary to combine different methods. The purpose of this work is to implement a strategy for road lanes detection using the traditional Canny operator. Oriented filters are used to remove unnecessary information and vehicle’s yaw rate signal is used to adaptively correct the filter orientation according to the lane boundaries directions. In sequence, morphological filters using dilation and analysis of connected components are applied in order to remove the noise components of the edge detection stage. A linear approach for lane tracking based on Hough transform was applied and proved to be an effective technique considering a near field as region of interest, enriching the evaluation of the results. The algorithm has shown to be robust in a variety of test scenarios including reflective roads, shadows and weakened lane markings.