Non-fatal road casualties are under-reported, with official figures coming from the police. In the French Rhône county, a road trauma registry constitutes a second source of reporting but its completeness needs to be assessed. We also wish to estimate an incidence rate of non-fatal road casualties that is corrected for under-count.Having two sources of reporting available, we can apply the capture-recapture method under certain conditions. To take into account different reporting probabilities among casualties, the capture-recapture analysis is stratified according to injury severity (New Injury Severity Score=NISS), road user type and human third party. To evaluate the sensitivity of the capture-recapture estimate on the number of matched casualties between the police file and the registry, three scenarios of record-linkage are considered.For serious (NISS 9+) non-fatal road casualties, the police ascertainment rate is at most 57%, the registry ascertainment rate is at most 87%, and the aggregate ascertainment rate is at most 95%. The ascertainment-corrected incidence rate for serious (NISS 9+) non-fatal road casualties is at least 65 per 100,000.The degree of completeness of the registry for serious casualties is rather high, though not satisfactory.