To increase driving safety, many researcher works on Advanced Driving Assistance Systems (ADAS) have been developed and embedded in real prototypes during the last decades. For some of these applications like Lane Keeping System, lane perception is an essential task. For others applications like Emergency Brake Assist, lane perception modules provide useful information helping the system to select only the most dangerous obstacles. Proposed solutions to perform lane detection become more and more elaborated, however no generic solution has been proposed to calculate performances of these algorithms. Lots of solutions have been proposed to perform this lane detection. However, no generic solution has actually been proposed to quantify the quality of such applications. It is appearing that this evaluation task is now very important and critical. Most of the existing evaluation stages can be classified in two main parts. In the first case, evaluation is based on natural images databases with ground truth of road marking and/or geometrical truth of lanes. In the second case, evaluation uses virtual data and simulated images. The first one is relatively hard to perform because it is based on manual labeling of natural images. The second one has automatic labeling clustering but a realistic virtual environment is required and more precisely both realistic road bitumen and road marking textures. This paper presents an efficient solution in order to simulate roads environment for the evaluation stage of road marking detection algorithms. Moreover, a powerful tool dedicated to the road marking texture generation is proposed. It takes into account both imperfection and wear of the road marking. A virtual database using this tool will be applied on a set of road marking extractor to validate the evaluation process with our virtual approach.