In English

A model based approach to lane detection and lane positioning using OpenCV

Daniel Posch ; Jesper Rask
Göteborg : Chalmers tekniska högskola, 2017. 41 s.
[Examensarbete på grundnivå]

The aim of this thesis was to implement and develop a computer vision based method that would play a important part in the implementation of an autonomous RC car. In particular this thesis provides the initial steps of image pre-processing, an algorithm for lane detection, position identification and a communication model.

The study was preformed with two main goals. The first goal consists of an investigation of a suitable algorithm to efficiently detect specific information from an image, to be able to act in a way based on the extracted data. The second goal was to determine what camera specifications are needed for the chosen algorithm.

Using different approaches of algorithms for detecting a path between lanes, this thesis present an implementation of the B-snake model. The program is evaluated in the Udacity game simulator, as well as on real hardware with challenging benchmarks such as lanes with hard curvature, high speed and noisy environments. The program performed well in different environments together with a limited speed. However lanes with curvature which exceeds 25 degrees has to be further developed in the future.

Nyckelord: Autonomous vehicles, OpenCV, lane detection, lane keeping, CHEVP, B-snake

Publikationen registrerades 2017-06-21. Den ändrades senast 2017-06-21

CPL ID: 250038

Detta är en tjänst från Chalmers bibliotek