Research On Non-contact Anthropometric Method Based On Image Processing

In the field of human clothing,human body measurement technology is constantly developing in the direction of digitization and informatization.Non-contact anthropometric measurement can achieve rapid acquisition of human body size information,which greatly improves the work efficiency of traditional manual anthropometrics.At present,the equipment required for non-contact three-dimensional measurement is generally expensive and large,and there are also certain inconveniences in the processing and data storage of three-dimensional data.The noncontact two-dimensional measurement is based on the human body image.Through the process of human body image processing,segmentation and extraction,the human body size data can be directly output.Non-contact human body measurement based on image processing has a very broad research and application prospects for human body size acquisition and clothing network customization.In this project,the current process of image-based anthropometric measurements has been optimized and improved in order to solve the problems of low extraction accuracy and difficulty in obtaining body girth dimensions.The main research work is as follows:(1)Aiming at the problems of large noise and discontinuity in the extraction of human contours in the image,a refined and systematic image processing method was proposed,and a human body measurement system based on image processing was established and developed independently.An image preprocessing scheme suitable for the research system of this subject is proposed.Realize image grayscale through color space conversion.The threshold adjustment function is implemented in the self-built system,which improves the accuracy of image segmentation.Through experimental comparison,the canny operator is selected for image edge detection,which strengthens the uniformity of human contour extraction.Use the knowledge of morphology to remove small noises,fill in the voids of the human body,and strengthen the outline of the human body.(2)The human body size extraction scheme for establishing a preliminary search area of human body parts is proposed.In the search area of each part of the human body,the coordinates of the breast point,hip peak point,acromion point and side waist point are determined based on the characteristics of human body shape.Use the height manually entered to determine the relationship between the real size of the human body and the coordinate size of the image.(3)Aiming at the problem that it is difficult to predict the circumference of the human body width and thickness,based on the manual measurement data of 210 people,4 types of circumference prediction models have been established using the method of machine learning.According to the experimental verification and comparison of the three girths of bust,waist and hip,the optimal model for each girth was selected,and each girth model was integrated into the self-built body measurement system.