Abstract—Tracking-based video surveillance approaches use a pipe line of processes from capture of frames up to video analysis. All these processes consume too much computational cost and generally it is concentrated in the last step of this framework. Particularly for this step, our paper proposes a method for abnormal motion analysis that ensures efficiency in the inferences with less computational effort. For this, we use a region-based model that uses a mobile grid of subregions constructed from scene's ROI (region of interest). In order to avoid the implementation of the complete framework, we have replaced the previous steps with annotated datasets from the real world. From these annotations, we seek a size of subregion that produces the best result in the abnormal motion detection using GMM (Gaussian Mixture Models) and ROC (Receiver Operating Characteristic) curves. The method proved efficient and useful for abnormal motion analysis, especially
in tracking-based approaches.
Index Terms—motion analysis, abnormal motion detection, pattern recognition, video surveillance
Cite: Jorge Henrique Busatto Casagrande and Marcelo Ricardo Stemmer, "Abnormal Motion Analysis for Tracking-Based Approaches Using Region-Based Method with Mobile Grid," Journal of Image and Graphics, Vol. 2, No. 1, pp. 22-27, June 2014. doi: 10.12720/joig.2.1.22-27
Copyright © 2012-2024 Journal of Image and Graphics, All Rights Reserved