Study On Obtaining Anthropometric Data Based On Binocular Vision

With the increase of car ownership, road traffic accidents occur frequently, and have becomemajor global economic and social problems. How to improve the driving safety has become issue thatpeople concern, so all kinds of the active safety equipments were produced, and man-machineengineering application is becoming more and more widely. Building human body model is the basisof research of Automotive Ergonomics, so a fast non-contact body measurement system need to bedeveloped to improve the precision and speed of human body measurement and building human bodymodel which is used in automobile design. This system can measure the human body size to build themodel of human body which lay the foundation of vehicle and its components’ design to makevehicle more comfortable.This paper raises the parameter acquisition method of the human body based on binocular vision.Using two cameras obtain the front and side images. By using various algorithms of imageprocessing and the body contour line, each feature point area of interest are determined, and then findaccurate feature points in the areas; through cameras are calibrated to obtain two cameras’ internalparameters and position relationship, Body height is obtained by3d reconstruction, and then other2ddimension values are calculated using the anchoring approach. Based on those2d data, surroundeddegree size values are calculated by using linear regression models. The main research contents are asfollows:1. Build binocular stereo vision sysytem platform and calibrate two cameras. Image acquisitionis the basis of follow-up processing. Firstly, images were caputured using the image acquisitionplatform. Then using obtained the different angles images of planar checkerboard, two cameras werecalibrated based on Zhang Zhengyou calibration method.2. Image preprocessing and human body contour extraction. The human body contour is thefoundation of the body feature points. First of all, chose Sobel operator to extract the body outline,then used the gaussian smoothing method to smooth image, and segment images by using iterativethreshold method. Finally, adopted freeman chain code gets the front and side images’ contours of thehuman body.3. Feature points of the human body extraction. Access to get feature points effectively is thebasis of size measurement. First based on the clothing design standard described the height of10 curve of human body, and then got the initial feature points on the human body contour line. Then theinitial feature points were used as the center to determine the interest area. Finally in the interestedareas using the Harris corner detection mothed found effective human feature points.4. Two-dimensional size and surrounded degree calculation. First of all, according to3dreconstruction technology height of human body was calculated. Then based on the height data，using the reference method calculated the other two dimensions data, such as head width, head high,thick waist width and soon. Finally use calculated two dimensions data, and curve fitting and dualdegree of a regression analysis calculated the circumference size, and analyzed the error for eachsize.Based on Visual Studio2008software, and combined with the binocular stereo vision system,software system for measuring the parameters of the human body and related algorithm wasdeveloped. Through collecting the human body images, drivers of human body size data wereobtained to provide support for the man-machine engineering research. The experimental resultsshowed that the method can quickly and accurately obtaine the body size and has good real-timeperformance and accuracy.