Abnormal Action Recognition in Power Production
Safety is of importance for power production. It is known that kinds of abnormal action in power production are concerned with objects. Different from many abnormal action recognitions researches, which ignore the interactive relationship between human and objects, this paper proposes an approach for the specific abnormal action based objects recognition in power production. The major contributions of the paper are to employ an object-based framework for description of hand-trajectory information, reference particle filter for primitive action locating and realize time cost decrease through human silhouette block analysis. Different from simplex trajectory based approach, the presented approach consists of types of trajectory features which belong to specific object and each primitive action is relevant to specific object model. It is verified by experiment that the approach performs on certain abnormal action recognition at an effective level.
Helen Zhang, Gang Shen and David Jin
Z. H. Wei et al., "Abnormal Action Recognition in Power Production", Advanced Materials Research, Vols. 225-226, pp. 311-314, 2011