A fusion registration method based on image data and point cloud data

A technology of point cloud data and image data, applied in image data processing, image analysis, image enhancement, etc., can solve the problems of different non-rigid models of point cloud data and image data, difficulty, and variable scale of point cloud data

Active Publication Date: 2021-07-06
ZHEJIANG LAB +1
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The non-rigid registration of fusion image information and point cloud information mainly faces the following problems: 1) The non-rigid models of point cloud data and image data are different, and there is no unified non-rigid transformation model; 2) The data of point cloud data The scale is variable and the index is irrelevant, so it is difficult to be directly used in UNet used by mainstream image registration algorithms; 3) There is still a lack of registration methods based on deep learning to fuse image information and point cloud information

Method used

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
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A fusion registration method based on image data and point cloud data
  • A fusion registration method based on image data and point cloud data
  • A fusion registration method based on image data and point cloud data

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0027] Such as figure 1 It is a schematic diagram of fusion registration based on image data and point cloud data in the present invention, by constructing a fusion registration model, and simultaneously inputting images and point cloud data of images for registration, the method specifically includes the following steps:

[0028] Step 1: Acquire CT images of a large number of patients (such as image 3 shown) data, and obtain the corresponding tracheal tree point cloud through image segmentation, such as Figure 4 shown;

[0029] Step 2: Perform data processing such as denoising, smoothing, and simplification on the tracheal tree point cloud, and make the scale of the point cloud data the same; the point cloud simplification refers to a method of controlling the data scale by deleting data that contributes less to the shape The processing method can be realized by curvature reduction or cluster reduction; Figure 5 Shown is the simplified point cloud data;

[0030] Step 3...

Embodiment 2

[0039] Step 1: Use digital human technology to obtain a large amount of CT image data, and obtain feature points in the image through feature extraction, where feature extraction can use SIFT or MIND methods;

[0040] Step 2: Design as figure 2 The fusion registration model shown, the objective function of the fusion registration model is composed of image distance item, point cloud distance item, image deformation field regularization item, point cloud deformation field regularization item and deformation field consistency constraint item;

[0041] Step 3: Input the CT image obtained in step 1 to form a reference image, a floating image pair, and the corresponding simplified point cloud into the fusion registration model obtained in step 3 for training, so that the loss function converges and stabilizes, and the fusion registration is completed model training;

[0042] Step 4: Use the trained fusion registration model to register the floating image to be registered:

[004...

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 fusion registration method based on image data and point cloud data. The method first obtains the boundary point cloud of tissues and organs through image segmentation or obtains the feature point cloud in the image through feature extraction, and combines the image data and point cloud The data is input into the designed fusion registration model, and the registered image, point cloud data and deformation field are obtained. The fusion registration model includes an image registration network and a point cloud registration network. During training, the loss function consists of an image distance item, a point cloud distance item, a regularization item that constrains the image deformation field and the point cloud deformation field, and the image deformation field and the consistency constraints of the point cloud deformation field. The method of the invention can improve the ability of boundary preservation in image registration, and can also improve the problem of wrong matching caused by the small structure hidden in the image background when the fine structure point cloud information is obtained before registration.

Description

technical field [0001] The invention relates to the field of image registration, in particular to a method based on deep learning and used for non-rigid registration of medical images. Background technique [0002] The registration algorithm is a method for aligning image data or point cloud data by obtaining the spatial correspondence between image data or point cloud data. In the medical field, it is widely used in image-guided interventional surgery, image acquisition, quantitative analysis of sports fields, and digital human image generation. [0003] In the medical field, registration algorithms are mainly divided into image registration algorithms and point cloud registration algorithms according to the different data used. Traditional medical image registration methods generally consist of three main parts: objective function, coordinate transformation model, and iterative optimization strategy. Among them, according to the definition method of the image similarity ...

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
no application Login to view more
Patent Type & Authority Patents(China)
IPC IPC(8): G06T7/33G06T7/11G06T7/136G06T7/155
CPCG06T2207/10028G06T2207/10081G06T2207/10088G06T2207/10104G06T2207/10108G06T2207/10132G06T2207/20081G06T2207/20084G06T7/11G06T7/136G06T7/155G06T7/337
Inventor 朱闻韬饶璠杨宝陈凌张铎申慧叶宏伟王瑶法
Owner ZHEJIANG LAB
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap