Topological features for monitoring human activities at distance

摘要: In this paper, a topological approach for monitoring human activities is presented. This makes possible to protect the person’s privacy hiding details that are not essential processing security alarm. First, stack of silhouettes, extracted by background subtraction and thresholding, glued through their gravity centers, forming 3D digital binary image \(I\). Secondly, different orders simplices applied on simplicial complex obtained from \(I\), which capture relations among parts body when walking. Finally, signature persistence diagrams according each order. The measure cosine used give similarity value between signatures. way, powerful tool known as persistent homology novelty adapted deal with gender classification, person identification, carrying bag detection simple action recognition. Four experiments show strength feature used; three they use CASIA-B database, fourth KTH database present results in case actions first experiment named evaluated, obtaining \(98.8\,\%\) (lateral view) correct classification rates identification. second one shown an average \(98.5\,\%\). third result \(93.8\,\%\) detection. And last were \(97.7\,\%\) walking \(97.5\,\%\) running, took database.