Occluded Human Body Segmentation and its Application to Behavior Analysis

Abstract

This paper addresses the problem of occluded human segmentation and then uses its results for human behavior recognition. To make this ill-posed problem become solvable, a novel clustering scheme is proposed for constructing a model space for posture classification. To construct the model space, we use a triangulation-based method to divide a posture into different triangular meshes from which a posture descriptor “centroid context” is then extracted for posture recognition and model selection. Then, a model-driven approach can be proposed for separating an occluded region to individual objects from the model space. Due to partial occlusions, the task of model selection is very challenging. For reducing the model space, a particle filtering technique is then used for locating possible positions of each occluded object. Then, from these positions, the best model of each occluded object can be then selected using its distance maps. Then, a novel template re-projection technique is proposed for repairing an occluded object to a complete one. Then, each action sequence can be converted to a series of symbols through posture analysis. Since occluded objects are handled, there will be many posture symbol converting errors in this representation. Instead of using a specific symbol, we code a posture using not only its best matched key posture but also its similarities among other key postures. Then, recognition of an action taken from occluded objects can be modeled as a matrix matching problem. With the matrix representation, different actions (even caused by occluded persons) can be more robustly and effectively matched by comparing their Kullback–Leibler distance. Experimental results show the effectiveness and superiority of the proposed method in classifying human behaviors from occlude objects.