Research On The Measurement Method Of Human Waist Circumference Based On Image

Image-based anthropometry is a research hotspot in the field of computer vision and has a broad application prospect in the fields of clothing customization and human behavior recognition.Waist circumference plays an important role in the measurement of human body size.However,the existing image-based waist measurement methods have large errors.To solve this problem,this paper proposes a new method for measuring waist circumference.First of all,Harris corner point detection was performed on the body image with a belt,and then,through the Fast corner detection algorithm,color filter algorithm,cluster selection algorithm for corner point selection in order to achieve the purpose of the corner point on the belt,and then,Fitting the conic according to the corner point on the belt,and then further search by quadratic curve and supplement the omissions of angular point,then the discriminant square color by BP neural network,finally,according to the three sections of method to calculate human body waist length.The specific work is as follows: 1.A special soft magnetic belt is designed.The belt is marked with a square mark 1cm apart for easy corner point recognition.In addition,the belt has the advantages of avoiding clothes interference and low cost 2.Since there are many noisy corners in corners detected by Harris method,this paper proposes combining Fast corner detection,color screening and clustering screening algorithms to screen corners.First,the corner point set detected by Harris is screened by Fast method to obtain the corner point set screened by Fast method.Then,the color ratio of four adjacent positions around the corner point is used for color screening.Finally,the coordinate information of the diagonal point is clustered to achieve the purpose of screening noise corner points.3.Aiming at the problem of missing part of corner points,this paper proposes the corner point search algorithm to make up for this defect.First,the edge contour of the image was extracted by Canny operator.Then,fitting the conic by using the known set of corner points in order to obtain the intersection point between the conic curve and the block contour in the belt.Finally,the corner point search algorithm was performed at the intersection point to search along the block contour until the candidate corner points were found.4.In view of the problem of judging the cut-off line on the belt,this paper proposes the algorithm of judging the cut-off line on the belt,and accurately locates the cut-off line by judging the block’s closure.