0,tmm/courses/547-17/slides/alistair-traffic-4x… · the image cannot be displayed. your computer...

2
Visualiza(on, Selec(on, and Analysis of Traffic Flows Roeland Scheepens, Christophe Hurter, Huub van de Wetering, and Jarke J. van Wijk IEEE Transac@ons on Visualiza@on and Computer Graphics, Vol 22, No 1, January 2016 Presenta@on by Alistair Wick Overview 2 Mo(va(on Individuals move, but groups make traffic Traffic is unpredictable, but not random – it forms paPerns These paPerns are important, but difficult to analyse This tool helps us examine these paPerns, and how they change Takes a novel approach, using anima@on as a key visual channel 3 Data Trajectories over @me 10s of thousands of journeys 100s of thousands of sample points 4 Data Case Studies One day over Paris: 17,841 flights, 424,546 samples Vessel data near Dutch coast: 16,421 vessels, 420,335 samples 10s of thousands of journeys 100s of thousands of sample points 5 Data Large numbers of overlapping trajectories High density varia@on At a minimum, data includes: Posi@on Heading Velocity May also include: Vehicle type Al@tude for aircra^ Approach is designed for ordered traffic, not random movements! 6 Approach Density map shows where the trajectories lie Like a high-resolu@on heatmap, simple lightness encoding High-level overview of the paths taken 7 Approach Animated par@cle flow shows direc@on Low cluPer (unlike glyphs, arrows) Leaves a wide range of visual channels for other uses 8 Approach Important to realise the par@cles are not marking individual vehicles They are a visual aid to show direc@on of flow Not part of the data: vis layer added by the applica@on Uses some form of edge bundling (may be op@onal) on trajectories 9 Alterna(ves Animated textures Doesn’t work for thin trajectories Color maps Low resolu@on, can’t overlap Arrows/Glyphs CluPered; obscure the view Could just show the vehicle loca@ons? Terrible for iden@fying dis@nct trajectories 10 Par(cle Display Method Incorporated to density map Hard to combine with color Imprecise Gaussian Bells Alpha-blended on top of map Easy to color High visibility 11 Par(cle Flow - Pros Can resolve opposing direc@ons Low cluPer Intui@ve; mo@on encodes mo@on Low-priority channel – color channels remain free 12 Par(cle Flow - Cons (Very) high overhead Requires guaranteed performance Real-@me rendering at high resolu@on and framerate Thousands or millions of par@cles with constant turnover 13 Par(cle Flow - Cons (Very) high overhead Requires guaranteed performance Real-@me rendering at high resolu@on and framerate Thousands or millions of par@cles with constant turnover *Challenging – but not intractable 14 Selec(on/Filtering Both representa@ons respond to selec@on This is how the user explores the data Selec@ons are localized and filter all paths passing through them: 15 Selec(on/Filtering Selec@ons are also direc.onal The user picks a direc@on range to filter – paths which match the range within the selec@on area are selected 16

Upload: others

Post on 26-May-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 0,tmm/courses/547-17/slides/alistair-traffic-4x… · The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted

Visualiza(on, Selec(on, and Analysis of Traffic Flows RoelandScheepens,ChristopheHurter,HuubvandeWetering,andJarkeJ.vanWijkIEEETransac@onsonVisualiza@onandComputerGraphics,Vol22,No1,January2016

Presenta@onbyAlistairWick

Overview

2

Mo(va(on

•  Individualsmove,butgroupsmaketraffic•  Trafficisunpredictable,butnotrandom–itformspaPerns•  ThesepaPernsareimportant,butdifficulttoanalyse

•  ThistoolhelpsusexaminethesepaPerns,andhowtheychange•  Takesanovelapproach,usinganima@onasakeyvisualchannel

3

Data

•  Trajectoriesover@me

10softhousandsofjourneys 100softhousandsofsamplepoints4

Data CaseStudies• OnedayoverParis:17,841flights,424,546samples• VesseldatanearDutchcoast:16,421vessels,420,335samples

10softhousandsofjourneys 100softhousandsofsamplepoints5

Data

•  Largenumbersofoverlappingtrajectories• Highdensityvaria@on• Ataminimum,dataincludes:•  Posi@on•  Heading•  Velocity

• Mayalsoinclude:•  Vehicletype•  Al@tudeforaircra^

• Approachisdesignedfororderedtraffic,notrandommovements!

6

Approach • Densitymapshowswherethetrajectorieslie•  Likeahigh-resolu@onheatmap,simplelightnessencoding•  High-leveloverviewofthepathstaken

7

Approach • Animatedpar@cleflowshowsdirec@on•  LowcluPer(unlikeglyphs,arrows)•  Leavesawiderangeofvisualchannelsforotheruses

The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

8

Approach

•  Importanttorealisethepar@clesarenotmarkingindividualvehicles•  Theyareavisualaidtoshowdirec@onofflow• Notpartofthedata:vislayeraddedbytheapplica@on

• Usessomeformofedgebundling(maybeop@onal)ontrajectories

9

Alterna(ves

• Animatedtextures•  Doesn’tworkforthintrajectories

• Colormaps•  Lowresolu@on,can’toverlap

• Arrows/Glyphs•  CluPered;obscuretheview

• Couldjustshowthevehicleloca@ons?•  Terribleforiden@fyingdis@ncttrajectories

10

Par(cle Display Method

Incorporatedtodensitymap• Hardtocombinewithcolor•  Imprecise

GaussianBells• Alpha-blendedontopofmap•  Easytocolor• Highvisibility

11

Par(cle Flow - Pros

• Canresolveopposingdirec@ons•  LowcluPer•  Intui@ve;mo@onencodesmo@on•  Low-prioritychannel–colorchannelsremainfree

12

Par(cle Flow - Cons

•  (Very)highoverhead• Requiresguaranteedperformance• Real-@merenderingathighresolu@onandframerate•  Thousandsormillionsofpar@cleswithconstantturnover

13

Par(cle Flow - Cons

•  (Very)highoverhead• Requiresguaranteedperformance• Real-@merenderingathighresolu@onandframerate•  Thousandsormillionsofpar@cleswithconstantturnover

*Challenging–butnotintractable

14

Selec(on/Filtering

• Bothrepresenta@onsrespondtoselec@on•  Thisishowtheuserexploresthedata•  Selec@onsarelocalizedandfilterallpathspassingthroughthem:

15

Selec(on/Filtering

•  [email protected]•  Theuserpicksadirec@onrangetofilter–pathswhichmatchtherangewithintheselec@onareaareselected

16

Page 2: 0,tmm/courses/547-17/slides/alistair-traffic-4x… · The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted

The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

•  Selec@onscanberefinedinotherdimensions• Here,theuserfiltersbyal@tude:

17

Selec(on/Filtering

• Differentmodesofselec@oncanbecombined• Compoundselec@onsuseset-likelogic:

• Where⊕canbeeither⋃or⋂(unionorintersec@on),and∖issetsubtrac@on

•  Extremelypowerful• Usercanadd,intersect,andsubtract(inverse)selec@ons

𝑆=⊕ 𝑆↓𝑖 ∖⋃ 𝑅↓𝑖 

18

Selec(on/Filtering

Summary:• Polygonalselec@onareas•  Selectstrajectoriespassingthrougharea

•  Selectforheading/al@tuderanges•  Customselec@onwidget

• Powerfulcompoundselec@ons• Usercanenable/disablespecificselec@ons• Canhidetheunselectedtrajectories• Parametric–selec@onscanbemodifiedatany@me

19

Examina(on

• Detailwindowscanbedisplayedon-demandforanyselec@on

20

Fig.6(page6)

Examina(on

•  Zoomedviewofselec@onarea• Histogramoftrajectoriesover@me•  Polarareahistogramofdirec@ons•  Allsub-plotsarelinked•  Pickableselec@oncolor

21

Comparison

22

• Differentwindowsarealsolinked

Comparison

• Windowscanbecombinedformoredirectcomparison

23

Use-Case: Infographics

• Mul@pleselec@onscreateafacetedviewsuitableforinfographics• Hereasta@cviewcontainsalltheinforma@onneeded

24

Fig.10(page8)

Use-Case: Infographics

• BoattrafficontheDutchcoast• Antwerp(boPom)histogramshowswavepaPern–@descominginandout!

25

Fig.11(page9)

Use-Case: Air Traffic

•  FlightsinandoutofCharlesdeGaulleairport•  Take-offinblue•  Landingingreen• Partofevalua@onwith2professionalflightcontrollers

26

Fig.9(page8)

Use-Case: Air Traffic

•  Sameviewrotatedtoshowal@tude• Alterna@ngincoming/outgoingflows• Differentflowscanbeselectedandcompared

27

Fig.9(page8)

Thanks! Ques(ons?

28