ABSTRACT
Lanes can be described as the vertically printed markings on the road, whereas lane detection and tracking is the identification and tracking of these printed lane markings within and across the frames. Assuming that there is a camera mounted on the dashboard of the vehicle, lane detection is the capability of the system to detect lanes on the road. Lane detection is an integral part of an advanced driver-assistance system (ADAS) and intelligent vehicles. Many accidents occur due to driver error in cases of heavy traffic on highways. Therefore, there should be a guidance system that alerts the driver to and guides them through the surrounding traffic situation and that helps to increase the safety of the driver. Also, there should be a lane departure warning for avoiding the accidents that occur due to sudden lane changes. We have proposed a spatiotemporal incremental clustering algorithm that detects and tracks the lanes accurately in real time.
Many times the lanes are not properly visible on the road because of occlusion due to a high volume of traffic or wear and tear on the lane markings. It is possible that the lane does not get detected because of these mentioned reasons. Therefore, it is necessary to do lane prediction to be able to make the lane departure warning system and ADAS more robust and to enable these systems to continue working smoothly. Lane prediction can be described as accurately predicting the lanes where the lanes are not present on the road. We propose to predict the lane using the information on the vehicle and its track and on the last detected lanes. For this, we do vehicle detection and tracking in real time. We further describe how these terms are helpful in path planning and make an ADAS complete and robust.
As the driver is driving, he needs to decide whether he wants to remain in same lane or to change lanes. Our lane detection and prediction framework may provide static information on the real-time surrounding view that can be helpful for path planning. In this chapter, we discuss our lane detection, prediction, and path planning framework and present experimental results to show the robustness and accuracy of our framework.
