Feature Dimensions Recognition Of 3D Human Body Based On Digital Image Processing

Three-dimensional(3D) human body feature dimensions recognition is the aspect of 3D digital garment design and garment ergonomics field. In this dissertation, two algorithmic issues of the two-dimensional(2D) view and the digital geometric information gain are focused. The main contributions and innovations are given as follows:In the view of digital image processing, a projection from 3D scanned human body to 2D plane is firstly performed to obtain 2D front view and 2D side view respectively, which is followed by the binarization. On this basis, Canny operator provided with higher accuracy and better anti- noise performance is chosen for the edge detection of the resulted 2D front view and 2D side view. Then a edge tracking operation is employed to result in more smooth body contour. Finally, the Gaussian curvature is computed by chain code to locate the feature points on the human body contour.A new digital geometric information gain based approach to recognize 3D human body feature dimensions is developed in this dissertation. In the proposed algorithm, the whole scanned human body is segmented into such regions as neck, sho ulder, axillary, chest, waist and hip in accordance with the human body proportion relations stipulated in the GB10000-88. Then the definition of digital geometric information gain of the 3D object is given to measure the local geometry of each point on the 3D scanned human body. An appropriate threshold is established for each region, by which the optional feature points are decided. The points with salient values of digital geometric information gain among the optional feature points are finally chosen as the feature points.The recognized feature dimensions are categorized into two types, which are straight-line genre and circumference genre. The straight- line feature dimensions such as height, arm length etc. can be measured by calculating the Euclidean distance between feature points. As far as chest circumference, waist circumference and hip circumference are concerned, the cubic B-spline technique is first applied to fitting the underlying feature points to form the chest line, the waist line and the hip line respectively. The feature dimensions are obtained by measuring the circumference lines individually.Simulation results have proved that the proposed algorithms can reflect the local geometry of the key points on the 3D scanned human body effectively, and are with the capability of low computation cost, low computation complexity, and fast performance.