Road work zones pose a great challenge for the environmental perception with sensors and the correct modelling of the scenery. Especially the recognition of the valid lane is of great interest for driver assistance systems. At road work zones temporary lanes invalidate the original lanes while the latter are still present on the road. All lines are detected with a greyscale video camera. In this paper we present an approach which generates different hypotheses for the lanes originated from the two different types of lines. A particle filter is applied for smoothing and filtering while a JIPDA algorithm is used for association and with its existence estimation helps to find the most plausible hypothesis of the currently valid lane.