Research On Rapid Measurement Of Anthropometry Based On Monocular Vision Photography

The goal of this paper is to get anthropometric data by combining monocular camerameasurement with multivariate linear regression analysis and factors analysis.Anthropometric data can be defined into two different kinds: static data and dynamic data.And this focus on studying of static anthropometric data’s measurement and prediction.Using the28different kinds of anthropometric data of this paper, can create the ergonomicsdatabase of the software CATIA-V5. At the same time, these data can also be used in thestudying fields as drivers’ posture prediction, occupant comfort evaluation, auto-bodyinteriors, digital ergonomic modeling, and seat designing.The feasibility of photogrammetric anthropometric measurement has been verified byprevious experiments. And it has lots of advantages such as low cost, simple equipments,environment adaptability, efficient and so on. But, its measurement error is difficultlycontrolled. The main method of this paper is photogrammetric measurement by a definedreference object. By the way of error experiments, the paper finds that there are four factorswhich can influent the error accuracy:1.the object measured is not strictly coplanar with thereference object;2. The reference is too bigger or smaller than the object measured;3. Theobject distance of photographic is unreasonable;4. The subjectivity errors such as site erroror recording error. The paper chooses the gradual elimination methods to analyze thevariation of measurement errors, and to design the program of maximally eliminatingmeasurement errors by the results of experiments.This paper designs a photogrammetric auxiliary bench, which can help the personmeasured to adjust his/her postures adapting to the national standard. And it can also helpto eliminate the errors caused by no-standard posture. The bench and its ancillaryequipments size are designed according to the large size of anthropometric data of west, which can also adapt to the use of different countries. The bench combined with thereference object can achieve the gale of optimal design intent and get the bestanthropometric data. At the beginning of the photogrammetric measurement, the photosmust be strengthened contrast and shaped edges, to get clearer body size markeridentification and to measure the target parameters by software.According to the disadvantages of anthropometry based on constant of proportionality,this paper chooses the method of multiple linear regression analysis to predict the humanbody size. Because of the one-dimensional normal distribution of anthropometric data, andliner relationship between different human body sizes, use the methods of multiple linearregression analysis, factors analysis and variables contract to eliminate multi-collinearityeffects between the predictor variables. By the way of stepwise regression analysis toestablish linear regression predictor models of human limb segments. And the models aredetected by many aspects such as goodness of fit detection, collinearity detection, forecaststandard error detection, model F detection, and coefficients significant test and so on. Thefinal prediction models are best optimal.At the designing of pro-measurement plan and later anthropometric test, obtains ofhuman size data can be defined into three stages. First stage, the anthropometric data isobtained by directly measured, such as weight and high, which are two macro-scales.Second stage, there are nine different scales as distance between shoulder and elbow, whichcan be obtained by photogrammetric measurement. Third stage,17scales in which there areforearm length and etc. which can be forecasted by regression models. And the dates in thefirst and second stages are used as predictor factors in the third stage. The photogrammetricmeasurement data contain the scales of the second and the third stages, the third stage dataare use on regression models verification.According to the linear correlation between anthropometric data, this paper divides theanthropometric programs into four sections: head, arms, torso and feet. The anthropometricprograms in each section have significant correlation, which can help to improve the efficiency of independent variables selection. This paper eventually establishes male andfemale linear regression models of anthropometric data and verifies the models’ accuracyby experience test.By the way of photogrammetric measurement error test, the paper finds severalsignificant laws, which could be used widely if there is a lager sample measurement. At thesame time, this paper finds an efficiency way to establish multiple linear regression modelsof anthropometric data, which could reduce the time of model establishment. The modelsare best optimal and could achieve the goal of rapid measurement. The results of the work,combined with monocular vision photogrammetric and image processing, could achieve thegoal of rapid measurement and estimate of anthropometric data. And it has advantages oflow cost, convenient, accurate and fast.