Spatial Computing

Spatial Sciences Institute, University of Southern California

who we are

projects

publications

join us

View Our GitHub Profile

Hosted on GitHub Pages — Theme by orderedlist

WHO WE ARE

We are a research team at Spatial Sciences Institute, University of Southern California. We develop computer algorithms and build applications to solve real world problems in spatial sciences.

 

current

  • Yao-Yi Chiang
    Assistant Professor (Research), USC Spatial Sciences Institute
  • Vijayan Balasubramanian
    Graduate Student, USC Data Informatics
  • Weiwei Duan
    Graduate Student, USC Computer Science
  • Yuan Gao
    Associate Professor, Department of Information and Management, Northwest University, China
  • Narges Honarvar Nazari
    Graduate Student, USC Computer Science
  • Saarthak Khanna
    Graduate Student, USC Data Informatics
  • Sumedha Kucherlapati
    Graduate Student, USC Data Informatics
  • Zexuan Luo
    Graduate Student, USC Computer Science
  • Sanjay Singh
    Graduate Student, USC Computer Science
  • Jakapun Tachaiya
    Graduate Student, USC Data Informatics
  • Tian Xiang Tan
    Graduate Student, USC Computer Science
  • Jianhua Wu
    Visiting Scholar, USC Spatial Sciences Institute
  • Ronald Yu
    Undergraduate Student, USC Computer Science
  • Ying Zhang
    Assistant Professor, North China Electric Power University, School of Control and Computing Engineering
  • affiliate USC faculty

  • Craig Knoblock
    Director of Information Integration, USC Information Sciences Institute
    Research Professor, USC Computer Science
  • alumni

    2015
  • Robin Franke
    Undergraduate Student, USC GeoDesign
  • Alex Chen
    Undergraduate Student, USC GeoDesign
  • Leonard Ngo
    Undergraduate Student, USC GeoDesign
  • Kuangyu Xiong
    Undergraduate Student, USC Architecture
  • Yi Hou
    Undergraduate Student, USC GeoDesign
  • Rashmina Ramachandran Menon
    Graduate Student, USC Computer Science
  • Zebao Zhang
    Visiting Scholar, USC Spatial Sciences Institute
  • Nilesh Gupta
    Graduate Student, USC Electrical Engineering
  • Yamini Goyal
    Graduate Student, USC Computer Science
  • Andrew Hsu
    Palos Verdes Peninsula High School
  • Wilson Franca De Souza
    Undergraduate Student, Humboldt State University, Environmental Resources Engineering
  • Yang Meng
    Graduate Student, USC Data Science
  • Woojin Park
    Visiting Scholar, USC Spatial Sciences Institute
  • Sima Moghaddam
    Graduate Student, USC Computer Science
  • Haohan Yang
    Graduate Student, USC Computer Science
  • 2014
  • Akshay Anand
    Esri (Previously Graduate Student, USC Computer Science)
  • Ramtin Boustani
    Graduate Student, USC Computer Science
  • Renuka Fernandes
    Undergraduate Student, USC Electrical Engineering
  • Sanjauli Gupta
    Yahoo (Previously Graduate Student, USC Computer Science)
  • Jizhe Zhou
    Undergraduate Student, Electronics and Information Engineering, Beihang University (Beijing, China)
  • 2013
  • Ketan Akade
    Yahoo (Previously Graduate Student, USC Computer Science)
  • Cathy Ji
    Undergraduate Student, USC Computer Engineering
  • Parin Jogani
    Ebay (Previously Graduate Student, USC Computer Science)
  • Shrikanth Narayanan
    Graduate Student, USC Data Science
  • Ashish Shirode
    Intel (Previously Graduate Student, USC Computer Science)
  •  

     

    PROJECTS

    Geospatial Data Integration

  • Data Analytics with Knowledge Graph

    A domain expert can process heterogeneous data to make meaningful interpretations or predictions from the data. For example, by looking at research papers and patent records, an expert can determine the maturity of an emerging technology and predict the geographic location(s) and time (e.g., in a certain year) where and when the technology will be a success. However, this is an expert- and manual-intensive task. In this project, we are building an end-to-end system that leverages data collected from public sources to predict the (geographic) center(s) of a technology and when the center(s) will emerge. In our pilot study, we built a system to predict the future (geographic) center(s) for fuel cell technologies. The system extracts and cleanses data from public sources including research papers and patent records. After data extraction and cleansing, the system uses an ontology-based data integration method to generate knowledge graphs in the RDF (Resource Description Framework) format and enables users to switch quickly between machine learning models for predictive analytic tasks.
  • Spatiotemporal Data Mining Using Heterogeneous Geospatial Sources

    Given the various representations of geographic entities in heterogeneous data sources (e.g., a building can be a point or a polygon; building names can be store in a column called “name” or “name_eng”), a challenging problem is how to efficiently provide semantic descriptions for datasets from a large variety of sources to support machine learning and/or data mining tasks. In this joint effort with the BAE Systems, we are building an end-to-end approach that enables efficient modeling of spatiotemporal datasets from heterogeneous sources for performing analytical tasks on the modeled datasets

  • Modeling and visualizing geospatial data: Karma


  • Efficient cleaning and transformation of geospatial data attributes: ArcKarma (an Esri ArcGIS plugin)


  •  

    Text Recognition in Maps (OCR for Maps)

  • Overall approach:


  • Generating geonames from map images: ArcStrabo (an Esri ArcGIS plugin) (OCR for Maps)

  • Generating named road vector data from map images (sample results)

  •  

    Road Vectorization from Maps

  • Overall approach:


  • Sample results:


  •  

    Symbol Recognition in Maps [Repository]

  • Overall approach:

    1. Take a scanned map...(here shows an USGS historical topographic map)

    and a symbol example


    2. Automatically identify map symbols that look like the symbol example (the blue boxes)


  •  

    Linking Maps to Other Spatiotemporal Datasets

  • Linking Map Symbols to DBPedia:

    1. Take a scanned map...


    2. Automatically identify hotel symbols and link the symbol locations to DBpedia


    3. Linked locations in a GIS

  •  

    Linking Historical Maps to USC Shoah Foundation Visual History Archive:

  • The Visual History Archive (VHA) in the USC Shoah Foundation contains a large digital life story collection of survivors before, during, and after the Holocaust and other genocides. Currently, location information (e.g., place names) mentioned in the VHA is indexed by keywords. For example, using “Poland” as the keyword for place search on the VHA Online returns 5,325 indexing terms in which the indexing terms (place names) with verified locations are displayed in a Google Maps web interface. Since place names and administration boundaries can change significantly over time, displaying search results on a current map would not provide the best visualization tool for navigating the VHA digital collection through space and time. In addition, a number places mentioned (indexed) in the testimony could not be located due to the lack of historical sources for verifying the location information of these places. This limits the opportunity for researchers, educators, and the general public to access valuable VHA materials and prevents the VHA collection from being indexed and searched by advanced spatial queries (e.g., finding the testimonies mentioned cities or towns in Poland between 1930 and 1945).

    Historical maps are a great source of detailed place information in the past. For example, during the World War II (WWII), the US Army Map Service (AMS) created around 40,000 maps covering a significant amount of the earth. Other map sources provide detailed historical pre- and post-WWII maps, such as the Polish mapping company, Centrum Kartografii, which offers pre-WWII maps of Poland with a comprehensive list of place names including towns, manufacturing plants, monuments, etc. These historical maps can be found in either paper or scanned (digital) format in map archives such as the David Rumsey Map Collection or libraries including the USC Libraries, UCLA Map Libraries, Western Michigan Libraries, and the Library of Congress. The problem we are addressing here is how to systematically and effectively link places mentioned in the VHA collection to relevant historical maps and other historical materials.



  • Created by Andrew Hsu

    Created by Robin Franke

    Created by Alex Chen
  •  

    PUBLICATION

    Journal Articles Peer-Reviewed Conference/Workshop Articles

     

    JOIN US

    Hacking the SPACE!

    We are always looking for students and summer interns to work on interesting problems in spatial science, data science, computer science. Please feel free to send us an email if you are interested to join the team.

    USC Graduate Students

    Credit or non-credit (Computer Science, Data Informatics, or GIST studetns), we simply ask you to put down at least 10 hours a week so that you will have enough time to finish a cool project. You can also take the geospatial data integration course from CS department to learn more about our research.

    USC Undergraduate Students

    We love to work with undergraduate students. Join us and gain experience in research and build some awesome applications!

    Visitors and Students from Other Schools (including international scholars and students)

    We welcome visitors and students from other schools. We had great experience with international summer interns in the past. Come work with us in Los Angeles and enjoy the nice weather!

    Potential Research Topics

    Integrating, modeling, visualizing, and mining spatiotemporal information from heterogeneous data sources
    We are building automatic techniques to identify spatiotemporal patterns from heterogeneous data sources. Check the projects under Geospatial Data Integration for more information.
    Type of position: directed research, potential PhD students
    Requirement: excellent programming skills in Python, familiarity with machine learning and data mining, familiarity with Elasticsearch, Spark, spatial databases is a plus.

    Exploiting online data sources to build accurate gazetteers from map scans
    We are building automatic techniques to convert text labels in scanned maps into machine-readable text. Our current system utilizes Tesseract for the task of optical character recognition (OCR), and the results could contain missing or incorrectly recognized characters. We are investigating scalable, accurate methods to improve the character recognition results using existing geographic names. The idea is to automatically compare the recognition results with existing geographic names and find the best match efficiently. These geographic names can come from public online sources such as OpenStreetMaps and DBPedia.
    Type of position: directed research, potential PhD students
    Requirement: excellent programming skills in C#, familiarity with fuzzy string comparison algorithms, familiarity with Elasticsearch and/or image processing is a plus.

    Building an accurate character recognition system
    Optical character recognition (OCR) software such as the open source Tesseract uses a combination of techniques in machine learning, pattern recognition, and image processing. We are using Tesseract to convert text labels in hundreds of historical map scans into machine-readable text. We are investigating the best practice to train the underlying algorithms of Tesseract for efficiently recognizing specific font types and document layout.
    Type of position: directed research, potential PhD students
    Requirement: excellent programming skills in C#, familiarity with image processing, familiarity with pattern recognition and/or machine learning is a big plus.

    Symbol recognition from map scans
    We are investigating methods to automatically extract locations of cartographic symbols from map images.
    Type of position: directed research, potential PhD students
    Requirement: excellent programming skills in C#, familiarity with image processing, familiarity with pattern recognition and/or machine learning is a big plus.

    Image segmentation
    Grouping similar colors in an image is very often the first step for object recognition from images. Existing image methods applied on document images do not always result in a clear representation of text in images. We are investigating efficient and effective methods that utilize both color and spatial distances between pixels to separate text from other object in a document image.
    Type of position: directed research, potential PhD students
    Requirement: excellent programming skills in C#, familiarity with image processing and machine learning, familiarity with computer vision is a big plus.