Hash-based Gaussian-Mixture-Model-for-Roadside-LiDAR-Smart-Infrastructure-Applicataion

T. T. Zhang, Y. Ge, A. Chen, M. Sartipi and P. J. Jin, "Hash-Based Gaussian Mixture Model (HGMM) for Roadside LiDAR Smart Infrastructure Applications," in IEEE Transactions on Intelligent Transportation Systems, doi: 10.1109/TITS.2024.3434749. Paper_Link

Fig.1 LiDAR Detected Trajectory

Fig.2 Pre-Trained Bakcgrounds are overlayed with moving foreground points. In this figure, only the foreground objects are updated. The background infrastructure points are considered as stationary components.

Fig.3 Snowy Weather. The snowfalls cause phatom reflections with irregular and random points as displayed in the figure.

Fig.4 Signalized Intersection. Redlight phases usually have more severe occulusions than greenlight phases.

Fig.5 Highway Segment. High Volume Traffic at Afternoon Peak Hour.

Fig.6 Adaptive GMM Model Running in RealTime. The bakcground points are getting fewer and fewer when the GMM fully captured the background modes.