Shortened versions in International Conference on Computer Vision, Nice, France, pages 432-439 and Proc. Scale-Space'03, Isle of Skye, Scotland, Springer Lecture Notes in Computer Science, volume 2695, pages 372-387.
To detect spatio-temporal events, we build on the idea of the Harris and Forstner interest point operators and detect local structures in space-time where the image values have significant local variations in both space and time. We estimate the spatio-temporal extents of the detected events by maximizing a normalized spatio-temporal Laplacian operator over spatial and temporal scales. To represent the detected events we then compute local, spatio-temporal, scale-invariant N-jets and classify each event with respect to its jet descriptor. For the problem of human motion analysis, we illustrate how video representation in terms of local space-time features allows for detection of walking people in scenes with occlusions and dynamic cluttered backgrounds.
PDF: (1.6 Mb)
Demonstractions:
handwavedemo.avi (4.6Mb)
 
walk07a_classharris.avi (2.6Mb)
 
walk07a_silhouette.avi (3.0Mb)
 
walk02a_classharris.avi (1.4Mb)
 
walk02a_silhouette.avi (1.5Mb)
Related projects: Recognition of human actions
Related publications: (Recognition of activities from histograms of velocity-adapted spatio-temporal derivatives) (General scale selection principle) (Linear spatio-temporal scale-space) (Separable spatio-temporal scale-space with causal time direction) (Automatic selection of temporal scales in time-causal scale-space) (Monograph on scale-space theory)