Human body trunk posture matching method for measuring 2d image in e-commerce image

The invention discloses a human body trunk posture matching method for measuring a 2d image in an e-commerce image, and relates to the technical field of image matching. The method comprises the following steps: constructing a posture library; for each human body trunk posture image in the posture library, extracting key points of the human body trunk, and calculating the rotation degree; countingthe probability that the upper body trunk and the lower body trunk rotate in three dimensions; extracting key points of a human body trunk of the to-be-matched image, and calculating a rotation degree; calculating rotation degree absolute differences of the sample and the to-be-matched image in three dimensions, and performing weighted summation to obtain a first measurement index; calculating the sum of distances between the sample and corresponding key points in the to-be-matched image as a second measurement index; summing the first measurement index and the second measurement index to obtain a similarity index of