vis18-TopoAngler: fish extraction from body data using interaction topology

essays:TopoAngler: Interactive Topology-based Extraction of Fishes

Authors: Alexander Bock, Harish Doraiswamy, Adam Summers, etc.

Unit: New York University

Conference: Vis2018

I. Background of the paper

1.ScanAllFish Project:

The project started in2016 year, From the University of WashingtonAdam Summers The professor wants people to scan all the 30,000-plus species of fish in total, and put these high resolutionCT Scanning database sharing。 example chart1 as shown.( Isn't it beautiful?)

chart1. has its origins in:

2.ScanAllFish project challenges

So far, the ScanAllFish project has scanned only 600 of these 30,000 species of fish. The reasons are as follows:

Micro-CT is expensive and not widely available

12 hours to complete a single scan

So the scientists decided to wrap multiple fish together for the scanning process. As shown in figure 2:

chart2 (a) Fish binding tags to be scanned (b)12 Tie a fish into a" meat roll" (c) Perform a scan (d) Scan results

Multiple fish are scanned together, and the scan results need to be split to extract the respective results of the different fish. New challenges are posed, such as those shown in subfigure d of figure 2,

The fish can be placed in any direction,

Spatial proximity of the skeletons of different fish ,

The opacity of the rays is similar,

There is no sufficient a priori knowledge of size and shape,

It takes as long as 20 to 40 hours to manually extract the results of the scans of different fish.

3.In order to solve the above challenges, this paper proposes TopoAngler, which combines the computational topology with the viewable bounding surfaces, to perform interactive segmentation on these complex scanned body datasets.

II. System Overview

chart3 system chart a solution.

After reading and downsampling the data, topological analysis is performed, and the results of Join tree and branch decomposition are output, including feature identification and voxel identification.

The user needs to select features at different simplification levels to build meta-features (meta-features)

These meta-features are the output of the post-processing step.

Some of the concepts are sorted out as follows:

Features: connected sub-volume generated after topological analysis, which can be presented at different levels: skeleton, thorax or whole fish.

Meta features: i.e. whole fish. Meta-features are assembled from different layers of features by user interaction. is the final output.

III. hierarchical segmentation

The scalar field data of the microCT is first converted into tetrahedral mesh data, and then the extended connection tree is computed on the tetrahedral mesh (Algorithm: Computing Contour Trees in All Dimensions), and then the tree is decomposed by branching according to the importance measure of each edge in the tree (Algorithm: THE TOPORRERY: computation and presentation of multi-resolution topology).

chart4. Example of tree hierarchy segmentation chart

Requires user input of the number of features presented after simplification

When the number of features is 3, as shown in the above left figure, there are three candidate features ABC

The characteristic number is2 time, edgeBD removed, collaboration withAD The sides converge., constitute a new edgeAE, The final feature is blue with red, as above right chart

IV. application

Figure 5 A~D selection view: user can select candidate features according to the selection view and add or remove them in the meta-feature;

E meta-characteristic view chart

V. result

The case of Figure 6 contains nine South American fishes, mostly water tigers. (A) The entire data set . (B) Use TopoAngler to complete the split in 10 minutes. CD for high-bodied silverplate, EF for high-backed redgill tigerfish

VI. conclude

A paper like this, backed by a project and involving experts in the field, is credible.

1、I came out of coin speculation at 17 and went from having nothing before to being in debt now
2、The end of video encoding is near
3、Super nodes cant have earnings HelloEOS Zizen on EurochainNing Talk Blockchain EOS Super Node Interview Video
4、Finally find the golden cow hall cheat golden cow hall auxiliary seethrough hang
5、The superb artificial intelligence performance of the Honor V10 phone will not be able to get an ID card

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送