Online Object tracking:a Benchmark paper notes

Transferred from: http://blog.csdn.net/lanbing510/article/details/40411877

factors that affect the performance of a tracing algorithm 1 illumination variation2 Occlusion3 Background clutters Main modules for object tracking 1 Target representation scheme2 Search MECHANISM3 Model Update Evaluation Methodology 1 Precison plot:the Percentage of frames whose estimated location is within the given threshold distance of the ground Tru th.x Coordinate:threshold 2 Success plot:the ratios of successful frames at the thresholds varied from 0 to 1x Coordinat E:threshold 3 robustness Evaluationa ope:one-pass evaluationb TRE temporal robustness evaluationc SRE spatial Robustness Evaluation Overall Performance See thesis 1 TLD performs well in long sequences with a redetection Module 2 struck only estimates the location of target and do Es not handle scale Variation3 Sparse representations is effectivemodels to account for appearance change (e.g., Occlusio N). 4 Local Sparse representations is more effective than the ones with holistic sparse
Templates.5 It indicates the alignmentpooling technique adopted by ASLA are more robust to misalignments and background clu Tters.6 when an object moves fast, dense sampling based trackers (e.g., struck, TLD and CXT) perform much better than othe RS7 on the OCC subset, the struck, SCM, TLD, LSK and ASLA methods outperform others. The results suggest that structured learning and local sparse representations is effective in dealing with occlusions.8 O n the SV Subset,asla, SCM and struck perform best. The results show that
Trackers with affine motion models (e.g., ASLA and SCM) often handle scale variation better than others that is designed To translational motion with a few exceptions such as Struck9 the performance of TLDs, CXT, DFT and Lot de Creases with the increase of
Initialization scale. This indicates these trackers is more sensitive to background clutters. Hand, some trackers perform well or even better when the initial bounding box is enlarged, such as struck, OAB, Semit, and BSBT.  This indicates the haar-like features is somewhat robust to background clutters due to the summation operations when Computing features. Overall, struck is less sensitive to scale variation than other well-performing methods.11 Some trackers perform better wh En the scale factor are smaller, such as L1APG, MTT, Lot and CPF//supplement concluding Remarks1.background information is critical For effective tracking. 2.local Models is important for tracking 3.motion model or dynamic model is crucial to object tracking, especially when the motion of target
is large or abrupt

Good location prediction based on the dynamic model could reduce the search range and thus improve the TRA cking efficiency and robustness.


  Dataset 1 caviar:http://homepages.inf.ed.ac.uk/rbf/caviardata1/

corresponding website http://visual-tracking.net/#

Online Object tracking:a Benchmark paper notes (RPM)