About CRIM

AutoCarto, a tool for the automatic detection of hydrographic contours satellite images


Definition of the client's problem 

Natural Resources Canada (NRCan) must maintain its geospatial database (GDB) up-to-date, most notably the data pertaining to the Canadian hydrographic network. Satellite images, taken on a regular basis, allow NRCan scientists to monitor the progress of waterbodies (rivers, lakes) by detecting any changes in their contours. Through this analysis, maps are updated. Although certain detections are computer-assisted, several of the processing steps must still be performed manually. The overall process, in addition to being lengthy and costly, severely limits the update frequency of maps.

In order to curtail the number of manual steps involved and hence reduce costs and efficiently manage information updates, 

NRCan sought to further automate the overall detection process of hydric zones from satellite imagery. At the time, no commercial tool was capable of adequately meeting the needs of the ministry. It is under these circumstances that NRCan appealed to CRIM to develop AutoCarto, a program for the automatic updating of existing contours and the detection of new water objects.

Mandate Challenge

The main challenge of this mandate was to manage the large volumes of data required for this type of analysis, namely large-size images and the vast amount of hydric zones that must be detected and processed as a result of the millions of waterbodies that make up Canada’s landscape.

CRIM’s solution to the problem

The program AutoCarto, developed by CRIM, allows for the semi-automatic updating of contour maps of Canadian hydrographic zones (lakes and rivers) from images taken from the SPOT satellite. A Web version working within the context of a server has also been developed (http://geo-ti.crim.ca/).

This mandate includes a second phase that will extend the processing capabilities of AutoCarto to RADARSAT-2 images, by appealing to CRIM’s high-end expertise in the area of polarimetric image processing.

Impact on the enterprise

For NRCan, AutoCarto is a customized solution that provides a quality solution, both in terms of time and money, for the process of updating maps of the Canadian hydrographic network. Thanks to the automatic detections offered by the software, NRCan scientists can henceforth readily view any changes to the contours of waterbodies and neglect zones that remain unchanged. This has led to a 50% reduction of the number of objects that require processing. This work also had an unforeseen but positive upshot, namely the software developed by CRIM also assists in the systematic error detection of contour localizations on maps. In addition the software improves on the process of image browsing.

Positive fallout for the enterprise

AutoCarto improves on the quality and frequency of cartographic updates that are available to the general public through GéoBase (http://www.geobase.ca/). It thus enables the NRCan to fulfill its mission in a more efficient manner. These improvements, in both precision and efficiency, will benefit all Canadian organizations that access the relevant data.

CRIM stands out

Starting from a very specific need having no satisfactory commercial solution, CRIM succeeded in developing a completely operational program that efficiently automates several of the steps involved in hydrographic mapping from computerized analysis of satellite images. The solution allows for treating large-size images and detecting the millions of water contours of the Canadian landscape. It offers gains in performance, and also in quality, most notably in regards to image browsing.

Teams

Upcoming event

  • Vision Géomatique 2017
    19 October 2017 0:00
    Delta Sherbrooke
    Vision Géomatique est de retour en 2017 pour sa 5e édition qui se déroulera du 17 au 19 octobre 2017 au Delta Sherbrooke.
    +
  • Nous y serons !! https://t.co/ypF4iESZ07
  • Tourisme Montréal RT @MeetMontreal: Nov 8 - 18, Influential players in #creativity and #culture will flock to the first edition of @hubmontreal https://t.co/…

Recent Publications

  • PeopleBot

    +
  • Fine-grained domain classification of text using TERMIUM Plus

    +