Point Calibration of Face Feature in 3D Image Based on Genetic Algorithm
Faces have diversity and complexity,and the same face may exhibit different shapes and structures under different conditions,which makes it difficult to calibrate facial feature points.In order to accurately extract facial feature points,a 3D image facial feature point calibration based on genetic algorithm is proposed.A binocular stereo vision system is used to achieve multi angle collection of facial information,extracting main feature information from the collected information,and combining the commonality of the"three courtyards and five eyes"position relationship,compensating for facial contours and local transformation coordinate point information of facial features on a general facial model,obtaining a set of 3D coordinate point position information that is closer to the real face.A multi-population genetic algorithm is introduced to represent the shape and pose parameters of a 3D facial model based on the basic chromosomes within the subpopulation,and select the optimal chromosomes through different crossover and mutation operations to achieve accurate calibration of 3D facial feature points.The experimental results demonstrate that the proposed method establishes a three-dimensional facial model with high realism,accurate feature point calibration,and a highest MSE(Mean Squared Error)of 7.9％.
binocular stereo vision system3D facial modelmulti population genetic algorithmscreening of optimal chromosomes
国家科技期刊平台
NSTL
万方数据
