Skull appearance restoration model construction method and skull appearance restoration method and system

A construction method and skull technology, applied in the field of 3D reconstruction, can solve the problems of lack of explicit knowledge, insufficient representation of craniofacial nonlinear deformation, and insufficient model representation ability in knowledge analysis model methods, achieving convenient operation, high efficiency, The effect of high restoration accuracy

A construction method and skull technology, applied in the field of 3D reconstruction, can solve the problems of lack of explicit knowledge, insufficient representation of craniofacial nonlinear deformation, and insufficient model representation ability in knowledge analysis model methods, achieving convenient operation, high efficiency, The effect of high restoration accuracy

CN112288645APending Publication Date: 2021-01-29NORTHWEST UNIV

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Skull appearance restoration model construction method and skull appearance restoration method and system
  • Skull appearance restoration model construction method and skull appearance restoration method and system
  • Skull appearance restoration model construction method and skull appearance restoration method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0058] The skull data and skin data in this embodiment are three-dimensional grid data, and the three-dimensional grid data adopted is obtained by scanning the head CT to obtain head section image data conforming to the DICOM standard, and performing denoising and redundancy removal on the CT data , using the Marching Cubes method to reconstruct the 3D mesh data of the skull and skin. Each 3D mesh data contains about 100,000 vertices and 200,000 triangular patches. The data volume is 400 skull data samples and 400 skin data samples. And one-to-one correspondence, from the same person. The following are the specific implementation steps:

[0059] Step 1: Unify the three-dimensional skull data and three-dimensional skin grid data in the training sample set into the Frankfurt coordinate system, and use the distance between the center points of the left and right ear holes of each sample as the distance measurement unit of the sample, that is, for all points on the sample Normali...

Embodiment 2

[0076] This embodiment carries out craniofacial restoration with the model constructed in embodiment 1:

[0077] Step 1, the skull to be restored ( figure 2 ) and randomly select the three-dimensional grid data of the skin image in the data used in Example 1 to convert to a unified Frankfurt coordinate system and carry out scale normalization to normalize its attitude and size; use 3D MAX to process the step one Take two-dimensional screenshots of the skull data and skin data to obtain the two-dimensional data of the skull to be restored and random skin data, and adjust the pixels to 256×256;

[0078] Step 2: Using the model trained in the above-mentioned embodiment 1, input the random skin data and the two-dimensional data of the skull to be restored into the trained network at the same time to obtain the restored face image of the skull to be restored, as shown in image 3 .

[0079] Using the improved Inception convolutional neural network ("Research on craniofacial retr...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a skull appearance restoration model construction method and a skull appearance restoration method. The model construction method comprises the steps of creating a sample dataset, wherein the sample training set comprises a plurality of skull samples and skin samples of the skull samples, and the skull samples and the skin of the skull samples are subjected to coordinate normalization and scale normalization processing; wherein the skull samples and the skin samples are two-dimensional data; and training an initial generative adversarial network by using the sample training set, with the initial generative adversarial network comprising a generator and a discriminator, the generator being an improved U-Net network, and the discriminator being a PatchGAN network. According to the restoration method, the constructed model is used for carrying out facial restoration on the unknown skull. The invention is convenient to operate, and the defect that a traditional method is insufficient in craniofacial nonlinear deformation representation is overcome; the defect that a traditional method depends on template selection is overcome; and the defect of insufficient representation of the texture and shape of a wrapper in the existing method is overcome.

Description

technical field [0001] The invention belongs to the field of three-dimensional reconstruction, and relates to a method for restoring the appearance of a skull by using a generative confrontation network. It is mainly used in criminal investigation, archaeology, forensic anthropology and other fields. Background technique [0002] In the field of biometric identification, face restoration through skull is an important research content in this field. Facial restoration has certain applications in the fields of forensic anthropology, criminal investigation, and identification of unknown skulls, providing an important basis for future related research. [0003] The early expression and description of the essential relationship between craniofacial morphology relied on the subjective understanding of craniofacial morphology by anthropological experts, and there was no exact definition. Therefore, early craniofacial restoration relied on experts' manual restoration, repeatability...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
29 Jan 2021
Publication
CN112288645A
IPC
G06T5/00; G06T7/13; G06T17/00; G06N3/04; G06N3/08
CPC
G06T7/13; G06T17/00; G06N3/08; G06T2207/20104; G06T2207/20081; G06T2207/30201; G06N3/045; G06T5/77
Inventors
刘晓宁; 林芃樾