Within the framework of MOCoPo research project, an in-situ experiment was conducted during one- month in Grenoble, France (400,000 inhabitants) on a high-capacity road (ring, 2x2 lanes) close to an area of exchanges and frequently under congested conditions. Traffic counting was performed at 6- minutes intervals. Four video-cameras were installed to capture in real-time vehicle license plates. Their identification through the national registrations enabled the characterization of the fleet composition. These data were used to simulate pollutant emissions from the traffic, by the COPCETE calculation plate-form. Based on the COPERT4 (V9) methodology, it enables the computation of around 7000 observations (i.e. two full weeks of valid 6-minutes data), integrated on a distance of 1 km, and described by the real-world traffic data (flow, speed) and fleet composition as observed by the video. These works have first enabled the implementation of the video technique to capture the in- situ fleet composition. 1.7 million of observations and 350,000 French registrations allowed a good and detailed characterization of the local fleet, its specificities and variability with the time-periods. Analyses of pollutant emissions demonstrate that they are mainly linked to the traffic flows, thus to the fleet composition (heavy vehicle share) and then to the traffic conditions. Complementary analyses demonstrate the importance of cold start over-emission.