We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

The most accurate cameras to generate map data from street-level imagery

00:00

Formal Metadata

Title
The most accurate cameras to generate map data from street-level imagery
Title of Series
Number of Parts
351
Author
Contributors
License
CC Attribution 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Identifiers
Publisher
Release Date
Language
Production Year2022

Content Metadata

Subject Area
Genre
Abstract
Mapping is time-consuming and requires a high volume of a workforce when it comes to keep maps up to date periodically. This brings the need of finding alternative approaches to keep maps up to date. Mobile mapping is the process of collecting geospatial data from a mobile vehicle using a 360º camera, laser scanner, GPS/IMU positioning system, and other sensors. Many devices now include a geotag for every photo captured, and GPS accuracy can have major effects on the quality of street-level imagery and derived data. Join us in an exploration of the different accuracy levels of GPS-enabled cameras, where we will take a look at how different devices compare, and what varied levels of GPS accuracy look like both for image location and for data extracted using computer vision and structure from motion. Understanding the differences between devices is an important step in planning street-level imagery capture, as it will align your expectations with the advantages and limitations of the hardware you use. We tested various devices and will share the results of our investigation, with the aim of equipping you to capture street-level imagery with the tools and methods that fit your needs.
Keywords
202
Thumbnail
1:16:05
226
242
Euclidean vectorOpen setTexture mappingMeta elementMappingProjective planeDigital rights managementObservational studyCASE <Informatik>Computer animation
Scale (map)Euclidean vectorLevel (video gaming)Computing platformTexture mappingScaling (geometry)Source codeJSONUML
Capillary actionCone penetration testTexture mappingWorkstation <Musikinstrument>Total S.A.Motion capturePressureComputer-generated imageryPole (complex analysis)Texture mappingSoftwareSmartphoneMedical imagingMappingShared memoryObject (grammar)Total S.A.Coma BerenicesBuildingWorkstation <Musikinstrument>Computer animation
Capillary actionData structureForm (programming)Medical imagingGame controller
Data structureCapillary actionMedical imagingObject (grammar)Computer animation
Standard errorRadiology information systemHand fanLevel (video gaming)Complete metric spaceMappingResultantMedical imagingView (database)Suite (music)
Object (grammar)Social classUtility softwarePole (complex analysis)InformationComplete metric spaceObservational studyFile formatProcess (computing)Level (video gaming)MappingCASE <Informatik>Computing platformPosition operatorComputer animation
AreaMappingText editor
Complete metric spaceMereologyJSONComputer animation
Capillary actionData analysisProcess (computing)Complete metric spaceMappingPosition operatorAreaMedical imagingLevel (video gaming)Formal verificationMereologySuite (music)XMLProgram flowchart
Capillary actionMedical imagingLevel (video gaming)MappingOrder (biology)Maxima and minimaComputer animation
Digital photographyMaxima and minimaCurvatureData typeLengthImage resolutionLatent heatComputer animation
Maxima and minimaComputer animation
Arrow of time
Task (computing)Process (computing)Object (grammar)Formal verificationFile formatFormal verificationMedical imagingPosition operatorComputer animation
Capillary actionoutputLevel (video gaming)Virtual machineComputer animation
Complete metric spacePerformance appraisalMathematical analysisCapillary actionComplete metric spaceVirtual machinePosition operatorComputer animation
Complete metric spaceCapillary actionCurvatureMaxima and minimaDigital photographyPosition operatorTotal S.A.NumberData typePositional notationMedianStandard deviationAverageVarianceComplete metric spaceTerm (mathematics)Position operatorMaxima and minimaPopulation densityComputer animation
Digital photographyData typeAreaPopulation densityMotion captureFrequencyComputer-generated imageryComplete metric spaceMaxima and minimaImage resolutionMappingScaling (geometry)Position operatorLevel (video gaming)Electric generatorMaxima and minimaMetreComplete metric spaceComputer animation
Transcript: English(auto-generated)
Yeah, hello everyone, this is Saitruk Sarvaj, I'm with META, I'm community project manager I support META's open open community mapping effort In this lightning talk, I will address the issue that Maply regenerates map data and then I will in this case study. I will
Highlight, what is the quality of this map regenerated data? So in short Maply is a street level imagery platform scales and automates mapping and basically Maply brings a novel solution to legacy mapping solutions such as like total station GPS and mobile mapping and
You can basically contribute to a Maply with any camera, which is geotagged So we have 1.7 billion imagery with creative commons share alike a license Which can be used even for commercial purposes and then images comes from Contributor networks and then we
detect objects and build 3d reconstruction from those images and extract map features from those images and Once the imagery has been captured images can be uploaded with desktop uploader Maply tools and smartphone and I have short
Okay Can we play with video please? All
Right, so basically Cool, so images comes to our pipeline and we process those images and then generate a 3d 3d reconstruction from those images and then identify all the objects in the image and we convert Those street level imagery to map data, but the question is how this map data is accurate and complete
So in this lighting thought I want to just share quick result on the completeness and then position accuracy of this data So here we have a 3d reconstruction from suite level imagery and then we are converting this data to map data, which you can See in a second that in the map view and yeah, as I mentioned that we are going to also
analyze quality of this data So in this particular case study I addressed position accuracy and completeness of Street lights, but all of those data available on map lee proud platform
Basically, you capture street level imagery and then without doing any processing on your end. You can extract this data in georgism format and Maple really generated map features are available on ID editor rapid and joe SM. So mapler has plug-in on joe SM as well and
In this part we are focusing on like quality and completeness of maple data and This is my methodology how I analyze quality and completeness of map legends and map data here I collect a suite level imagery in the area of interest and then grantor data is provided by
or the municipality in the area of interest and then images are upload to my flurry and then I Coordinated like verification to remove like false positives and then extract map data in the last part I prepare data set to be analyzed and then I analyze completeness and positional accuracy of this data
So street level imagery is collected with iPhone iPhone 11 and then GoPro hero and GoPro max and then ground truth data is provided by order municipality and then Images uploaded to mapler and then generated by mapler so this is the specification of the cameras and Then GoPro max is 360 which gives you better visibility of the of the places you can see on the bottom side and
Then he is grand true data and arrow of interest it's three by three and then three twenty kilometers of data has been collected for this project and for data processing images are uploaded to mapler II then false positives are removed with the verification tool and then
297 streetlight imagery has been Generated so basically input data is street level imagery and then you get machine-generated streetlight data and then I pair ground truth data and machine generated data and I calculate completeness and positional accuracy and then the most completeness
Comes with the GoPro hero because of density of imagery and then in terms of horizontal positional accuracy GoPro max got the higher positional accuracy and so basically map
Collecting map data is time-consuming and that really gives you a chance to collect map data in scale and then automates map generation effort and then basically with up like capturing seat level of imagery data you can just generate tons of map data and then this map data will be available for your
Basically processing and then two meter two meter is position like you said you with GoPro max and then the max completeness was with GoPro hero. Yeah. Thank you for listening. That was quick one. But if you have any question happy to answer