Construction method for two-dimensional image human body articulation point positioning model and positioning method

• Main Authors: ZHANG JUNGE, FU LIANRUI, HUANG KAIQI
• Format: Patent
• Language: Chinese; English
• Published: 29.03.2017
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a construction method for a two-dimensional image human body articulation point positioning model and a positioning method on the basis of the construction method. The construction method comprises steps that a colorful image of which human body articulation point position coordinates and shielding states are marked is utilized, and a human body part local characteristic training sample set and a human body part global configuration sample set are constructed; a depth convolutional neural network is constructed, the depth convolutional neural network is trained through utilizing the human body part local characteristic training sample set to acquire a human body part local appearance model; a shielding relationship graph model is acquired through utilizing the human body part local appearance model and the human body part global configuration sample set; the human body part local appearance model and the shielding relationship graph model are determined to be the two-dimensional image