Hi,
I am investigating clustering and tracking approaches in SRR (Short Range Radar), TM (Traffic Monitoring), PC (People Counting) and SDK.
Clustering and tracking approach in TM, PC and SDK are very similar and they are very different from SRR approach.
SRR: Range-Azimuth-Doppler Estimation --> pointCoud --> Clustering (DBSCAN) --> StrongestPointsInClusters --> Multi Target Tracking --> trackedObjects
Multi Target Tracking is based on EKF, gating: some thresholding, and data association: x_diff^2+y_diff^2+d_diff^2.
SDK/PC/TM: Range-Azimuth-Doppler Estimation --> pointCloud --> Multi Target Group Tracking --> trackedObjects
Multi Target Group Tracking is based on EKF, gating: mahalanobis distance threshold derived from constant ellipsoid volume, and data association: max likelihood Gaussian pdf.
Is there any study/analysis done on the performance comparison between both approaches?
or may be pros and cons of both approaches?
or anything close to it? :)
If yes, could you share it?
Thank you.
