Image segmentation for human motion analysis: methods and applications

8th. World Congress on Computational Mechanics (WCCM8) 5th. European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS 2008) June 30 – July 4, 2008 Venice, Italy IMAGE SEGMENTATION FOR HUMAN MOTION ANALYSIS: METHODS AND APPLICATIONS *Maria João M. Vasconcelos1 and João Manuel R. S. Tavares2 1,2 Faculdade de Engenharia da Universidade do Porto Instituto de Engenharia Mecânica e Gestão Industrial Rua Dr. Roberto Frias, s/n 4200-465 Porto PORTUGAL 1 [email protected] 2 [email protected] http://www.fe.up.pt/~tavares Keywords: Human Motion, Image Segmentation, Biomechanics, Applications. ABSTRACT Human motion analysis is closely connected with the development of computational techniques capable of automatically identify objects represented in image sequences, track and analyse its movement. Feature extraction is generally the first step in the study of human motion in image sequences which is strictly related to human motion modelling [1]. Next step is feature correspondence, where the problem of matching features between two consecutives image frames is addressed. Finally high level processing can be used in several applications of Computer Vision like, for instance, in the recognition of human movements, activities or poses. This work will focus in the study of image segmentation methods and applications for human motion analysis. Image segmentation methods related to human motion need to deal with several challenges such as: dynamic backgrounds, for instance when the camera is in motion; lighting conditions that can change along the image sequences; occlusion problems, when the subject does not remain inside the workspace; or image sequences with more than one subject in the workspace at the same time. It is not easy to develop methods which can deal with all these problems at once, so it is common to make some assumptions, however each day more robust and accurate methods are being developed. A typical method of image segmentation is background subtraction, which involves the calculi of a reference image followed by the subtraction of each frame of the image sequence from the reference and further threshold of the result [2]. The simplest form is using a timeaveraged background image as reference but it requires a training period absent of foreground objects. Other possibility is describing each pixel in the scene by a mixture of Gaussian distributions, where the weight parameters of the mixture represent the time proportions that those colours stay in the scene, so background components will be the ones with the highest probable colours. However, this last method usually fails in busy environments where a clean background is rare. In [3] is presented an improvement for the background mixture models that also describes a method to detect moving shadows. In [4] is proposed a method based on Bayes decision theory to detect foreground objects from complex image sequences which contain both stationary and moving backgrounds. It starts to establish a Bayes decision rule for background and foreground classification from a general feature vector and then applies to both stationary and moving background with suitable feature vectors. This method showed to work well in complex backgrounds including sequences with variable light conditions and shadows of moving objects however has a problem of absorbing foreground objects when they are motionless for a long time. In this work we will present some experimental results using these image segmentation methods for human motion analysis, discuss their advantages and disadvantages and address possible practical applications related with human motion. The analysis of human motion is motivated by the advantage to improve men/machine interaction in several applications, such as: surveillance systems, virtual reality animations, clinical study and diagnosis and analysis of athletic performances. REFERENCES 1. Aggarwal, J. and Q. Cai, Human Motion Analysis: A Review. Computer Vision and Image Understanding, 1999. 73(3): p. 428-440. 2. Chen, X., et al. Adaptive Silhouette Extraction and Human Tracking in Complex and Dynamic Environments. in International Conference on Image Processing. 2006. Atlanta, Georgia. 3. KaewTraKulPong, P. and R. Bowden. An Improved Adaptive Background Mixture Model for Real-time Tracking with Shadow Detection. in 2nd European Workshop on Advanced Video-based Surveillance Systems. 2001. Kingston upon Thames. 4. Li, L., et al. Foreground object detection from videos containing complex background. in Proceedings of the eleventh ACM international conference on Multimedia. 2003. Berkeley, CA, USA.