The reliability of 3D depth sensing is important for automotive perception, particularly in dynamic urban environments. We developed a range-adaptive perception framework for structured light depth sensing that integrates the density-based spatial clustering of applications with noise clustering, Mahalanobis validation gating, Hungarian data association, and Kalman filtering with a distance-dependent measurement-noise model. The framework was evaluated using static distance measurements, dynamic longitudinal tracking, and multi-object lane-change scenarios at vehicle speeds below 40 km/h. For static measurements at 2.0, 3.0, and 5.2 m, the framework reduced the mean absolute error (MAE) from 0.051, 0.079, and 0.169 m to 0.013, 0.023, and 0.061 m, respectively. In dynamic tracking, MAE decreased from 0.84 to 0.43 m and the mean absolute percentage error from 28.3 to 12.9%. Multi-object tracking reduced the average object-count deviation from 1.29 to 0.46. The measured end-to-end processing latency was 275 ms, and the range-adaptive state-estimation stage required only 1.2 ms per frame on the embedded platform. These results indicate that depth estimation and tracking stability were enhanced with limited computational overhead by modeling distance-dependent sensor uncertainty. The evaluation of the framework was limited to outdoor conditions and speeds below 40 km/h. Therefore, it is necessary to evaluate the framework under a wider range of weather, illumination, traffic density, and vehicle speeds and investigate longer-term multi-object tracking.