This paper examines a short-time estimation problem of an origin-destination (OD) matrix, where each element is a volume of vehicle flow between one of the OD pair of zones of a signalised junction. The estimation is based on the use of traffic measurements provided by video sensors and on the knowledge of the traffic lights. This data is subject to redundancy, imprecision and uncertainty. The main purpose of this paper is to obtain the best estimates of the OD matrix by modelling the data imperfection, using a two-step method. First, relationships between the observed data are built in real-time using High-Level Petri Nets. Due to the imperfection of data the system obtained is underdetermined and inconsistent. Second, the fuzzy sets theory is used to model this imperfection and to overcome the inconsistency of the system.