Research On 3D Facial Morphology Prediction Method Based On Gene Data

As the basic unit of heredity,genes control the basic characters of organisms and play a decisive role in individual identification and consanguinity identification.It has been a research hotspot in the field of criminal investigation and forensic medicine to predict the human facial morphology by gene.At present,the relatively popular DNA-based facial portrait technology mainly uses genome wide association study(GWAS)combined with single variable difference analysis technology or based on three-dimensional dense data points to study the correlation between gene and facial morphology,and then through machine learning algorithm to establish the corresponding prediction model,these methods often have complex process and need a lot of manual interaction process.In this paper,an end-to-end method is proposed by using deep learning method.It improves the prediction efficiency and ensures considerable prediction accuracy.It makes a pioneering attempt for the development of this technology in the field of deep learning in the future.The work of this paper mainly includes the following two parts:By referring to the published genetic locus related to facial features,combined with a part of randomly selected genetic locus as samples,quantitative analysis and visualization experiments are carried out in this paper.The experiment first requires some data preprocessing to segment,align and standardize the facial data,then groups the different performance traits of each genetic locus,and finally the difference between the average facial morphology of these groupings and the overall average facial morphology will be visualized to get the correlation between different genetic locus and different facial features.Compared with the traditional methods,this method can make quantitative analysis and visualization of the correlation between genes and facial features,which is more intuitive and reliable.According to the characteristics of gene data,a general adversarial network(GAN)is designed to fit the correlation between genetic locus and facial features.The network is divided into two subnetworks: generator and discriminator.The generator is responsible for generating the corresponding facial morphology according to the gene data,and the discriminator is responsible for judging whether the facial morphology is the real data corresponding to the gene data.Through analysis and validation of the experimental results,the facial morphology predicted by the network is very close to the real facial morphology,but also shows a high correlation with the corresponding gene data,which proves the feasibility and reliability of the method in this paper.