A Study On Fall Human Object Recognition Based On Videos

With the rapid growth of the population of the elderly,accidental falls of elderly people are a major cause of fatal injuries,especially for those living alone.Hence,reliable fall prevention and detection methods are essential for reducing falls and fall-related injuries in a house care environment.The research for such methods has become one of the major scientific trends all over the world.Great progress has been made in computer vision and image processing techniques in recent decades,which has opened up new opportunities to improve fall detection systems.Automatic detection of falls from video sequences is an assistant technology for low cost health care systems.This paper is devoted to the study of the basic problems of human fall detection and prevention and analyzes abnormal behaviors by using image sequences,so as to improve the detection accuracy and the time of emergency treatment.Our main contributions are summarized as follows:(1)The falling human body is extracted from the frames by building the chromaticity and brightness difference indices between the foreground and background pixels,which are used to classify the pixels into background mask,foreground mask and shadow mask.Meanwhile,the moving shadows are also separated from the human body.(2)A novel approach called minimum area-enclosing ellipse is proposed to locate the fragmented human body in the foreground image.Then,a normalized directional histogram(NDH)is developed around the center of the ellipse to represent a human posture by multi-directional statistical analysis.Moreover,shape features and statistical features are extracted from the region covered by the fitted ellipse.(3)To take full advantages of the binary classifiers support vector machine(SVM),we propose DAGSVM(directed acyclic graph support vector machine),which utilizes the directed acyclic graph strategy to combine several binary classification SVMs.It classifies the postures first into two categories(standing-lying,crouching-sitting)and then into four categories,namely standing,crouching,sitting and lying.A fall-like accident is detected by counting the occurrences of lying postures.After conducting the majority voting,the fall event is determined by immobility verification.(4)We present a novel slow feature analysis based framework for fall detection in a house care environment.First,a foreground human body is extracted by a background subtraction technique.After morphological operations,the human silhouette is refined and covered by a fitted ellipse.Second,six shape features are quantified from the covered silhouette to represent different human postures.With the help of the learned slow feature functions,the shape feature sequences are transformed into slow feature sequences with discriminative information about human actions.To represent the fall incidents,the squared first order temporal derivatives of the slow features are accumulated into a classification vector.Finally,falls are distinguished from other daily actions,such as walking,crouching,and sitting,by the trained directed acyclic graph support vector machine.Having been compared with two state-of-the-art methods on two public datasets,our approach gives much better results.