978-1-4244-6516-3/10/$26.00 ©2010 IEEE 2765 2010 3rd International Congress on Image and Signal Processing (CISP2010) Extraction Of Human Body Contours And Position Analysis Engin Mendi Computer Science Department University of Arkansas at Little Rock Little Rock, AR, USA Mariofanna Milanova Computer Science Department University of Arkansas at Little Rock Little Rock, AR, USA Roumen Kountchev Department of Radio Communications Technical University of Sofia Sofia, Bulgaria Roumiana Kountcheva T&K Engineering Co. Sofia, Bulgaria Vladimir Todorov T&K Engineering Co. Sofia, Bulgaria Abstract— The paper presents a new method for extracting and positioning contours of moving objects. The method is applicable in the surveillance of elderly individuals and facilitates the detection of critical situations when the elderly individuals find themselves in need of immediate help. For this, single frames from the video sequence are extracted in regular time intervals and the position of the human body is analyzed. The position is traced using the body contours. For the surveillance, a static video camera is used. The body contours extraction is performed using two consecutive filters: the first - for the adaptive texture suppression, and the second – for the contours extraction. The main presumption is that the environment is known and this makes the influence of multiple variable parameters (illumination, shadows, etc.) lower. A Fourier descriptor is used for the body position analysis. Keywords-component; Body contours extraction, Active contours extraction, Fourier descriptor I. INTRODUCTION Much work has been done to create various systems for visual based human pose estimation using monocular images [1-3]. One of the main techniques is to define human pose recovery as model-based and learning–based approaches. In a model-based approach, a known parametric body model is assumed. The pose recovery is solved by matching the pose variables to a forward rendered human model based on labeled extracted features. In a learning based approach, a relation from extracted features to pose variables is learned from training data. For the given input image, a similarity search is performed and the poses are defined. This work is focused on learning the shape pose evaluation. The presented approach aims to recover body geometric information like silhouettes, contours and deformed shapes to represent human activity. The key points of these applications are: how silhouettes are extracted and what kind of shape descriptor is used. Silhouettes and edges are used the most because they can be easily extracted and are to some extent lighting invariant. Active contours [4] are often used in computer vision and image analysis to detect and locate objects and to describe their shapes. The general disadvantages of these models are that the active contour may leak out of the ideal contour when the edges are weak and in many cases its extraction is very difficult. Since the active contour model is edge-based, the dynamic active contour captures the sharp edge of the shadow. Other algorithms for contours extraction are based on various techniques [6-8]. For the contours presentation different shape descriptors have been developed. Many authors proposed to use Fourier descriptor (FD) for shape analysis, and shape coding. The main advantages of FD are that it not only overcomes the weak discrimination ability of moment descriptors and global descriptors but also overcomes noise sensitivity [5]. In this work, the FD was applied on the human body silhouettes represented by their main contours. The paper is arranged as follows: Section 2 presents an explanation of the method used for the contour extraction. Section 3 presents the Fourier descriptor application. Section 4 presents some experimental results and Section 5 presents the conclusions. II. BODY CONTOURS EXTRACTION The main presumption here is that the images are single TV frames extracted in pre-defined intervals from video sequences. These frames are converted into grayscale and processed as individual still images. The first operation is to remove the known background thus retaining the moving figure only. The processing continues with the following steps: • Adaptive texture suppression; • Contours extraction. The first step is used to smooth the relatively small textures in the image, retaining the sharp transitions. For this, a two-dimensional fuzzy adaptive filter (2DAFF) is used with a sliding window of size M×N pixels. The filter performance is described by the relation: