Research And Application Of Picture-based Foot Measurement

Foot measurement is a part of anthropometry,which is mainly used in the field of shoemaking technology or online shoe purchase.Shoes as daily necessities protect and support the feet.If people often wear inappropriate shoes,it will cause foot disease and even lead to foot deformity.Helping consumers to buy suitable shoes can effectively protect consumers' feet.In this e-commerce era,consumers are more willing to shop online,but it is difficult to buy well-fitting shoes online without trying.For this reason,e-commerce has launched a virtual try-on,private customization and other programs to help consumers better choose well-fitting shoes.These programs are inseparable from the foot size of consumers,how to accurately,quickly and easily obtain foot size data is of great value.At present,the main foot measurement methods are manual measurement and measurement based on three-dimensional reconstruction.The tools used in manual measurement are simple and easy to operate,but the efficiency is not high and labor intensity is high;the measurement accuracy based on three-dimensional reconstruction is high,but it is limited by equipment and algorithms.It is difficult to measure the foot conveniently and quickly.This paper focuses on the convenient and fast method of foot measurement,and the main work is as follows:(1)Foot structure analysis and measurement parameters:from the perspective of the shoe industry,this article first analyzes the foot's physiological structure and stress model to understand the important parts of the foot in the process of movement and determine the relevant key points of the foot.Combined with the definition of foot parameters in the shoemaking process,a set of measurement parameters based on the key points of the foot were determined.(2)Research on the method of foot measurement:A picture-based foot measurement method is proposed.This method collects data based on a hand-held camera,uses reference object A4 paper to restore the scale,and combines key points of the foot to complete convenient and quick measurement of foot parameters.The collected data undergoes preprocessing such as data annotation,resampling,standardization,and data enhancement to form a dataset.Then use the MobileNet_V1_SSD network to get the foot area.Inspired by the relevant methods of face key point detection,this paper uses the deep alignment network DAN(Deep Alignment Network)to detect the key points of the foot.At the same time,the DAN model is optimized according to the characteristics of the data set and the shortcomings of the DAN model,and the activation function and loss function of the model are modified to improve the accuracy of the model.Finally,the picture is corrected by perspective transformation,and the reference foot A4 paper is used to restore the foot scale in the image,and the foot size is obtained from the obtained key position and scale of the foot.In this paper,experiments are designed from two aspects:the detection accuracy of key points and the measurement error of foot size,and the performances of cascade regression tree model,the original DAN model and the optimized DAN model are compared.The results show that the method proposed in this paper can conveniently and quickly complete the foot measurement with good accuracy.(3)The realization of foot measurement system:based on the foot measurement method studied in this paper,a foot measurement system is implemented on the Android platform.