Download presentation
Presentation is loading. Please wait.
Published byLoreen Singleton Modified over 9 years ago
1
U.S. Department of the Interior U.S. Geological Survey Exploring New Ground Data Sources GFSAD30 April 2015 Meeting Justin Poehnelt, Student Developer jpoehnelt@usgs.gov
2
Topics Mobile Application Image Classifier Digital Globe High Resolution Imagery Google Street View Imagery
3
Mobile Application Release Updates Achieved Apple Store Compatibility Google Play Store In Progress Issues Some issues with slow GPS capture on older devices. Need to work on versioning server application so that updates do not break mobile application. Android does not ask for compass calibration on compass initialization unlike IOS devices.
4
Mobile Application: Collect Data
5
Classifying Images Created a functional prototype to classify any type of imagery through crowdsourcing. High Resolution Satellite Imagery Photos from Field Data Mobile Application Photos Lucas Photos Other Sources Google Street View Images
6
Prototype
7
Prototype Feedback Class Definitions Size of Area to Focus Change to a 3x3 grid Note: Most imagery currently displayed is from Africa and of lower quality. http://dev.croplands.org/classify
8
Digital Globe Enhanced View Programmatic access to partial stack of images through web map tiles. No access to specific bands through this method. Tiles are served at 256 * 256 pixel size in epsg:3857 Zoom levels divide the layer into a 2^n by 2^n grid where n = 0..18. Max zoom of 18 corresponds to 6.8719476736 x 10^10 tiles. Can extract acquisition dates. Images are quickly and quickly downloaded.
9
Google Street View Goal: Get Google Street View images into application for classification. Issues Where to request imagery? API’s automatically snap to nearest image if available, but where is that? What is the acquisition date?
10
Google Street View Goal: Get Google Street View images into application for classification. Issues Where to request imagery? API’s automatically snap to nearest image if available, but where is that? When is the acquisition date?
11
Solving the Where Two Solutions Easy Way: Use Google Directions API to extract polyline of Google’s road layer. Limited to 2500 requests per day. Incomplete data set. Difficult Way: Extract data from Google Street View Coverage Map. Complete data set.
12
Solving the Where with Directions 1. Query Google Directions API with two Locations 2. With polyline, query Google Street View API with configurable spacing 3. If image exists, add location to database with location and heading of image. 4. Use different Google Street View API to extract acquisition date.
13
Solving the Where with Directions Map showing image locations to arbitrary location east of Flagstaff, AZ.
14
Solving the Where with Directions Issues Limited to where Google Directions creates route. Need algorithm for creating different transects on scales such as county, state, country and continent.
15
Solving the Where with Coverage http://gmaps- samples.googlecode.com/svn/trunk/streetview_landing/stre etview-map.html
16
Solving the Where with Coverage Issues Map may not be up-to-date There are 4,294,967,296 tiles to search for at zoom level 16 where each pixel is approximately 2.38 meters at the equator. Have already extracted nearly all of zoom level 15 tiles that are not empty. More significant processing needed but no usage restrictions on obtaining locations.
18
End
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.