InvertNet Newsletter, No. 2|
InvertDev: Updates from the Front Lines
From the Director (Chris Dietrich, InvertNet lead PI)
Greetings InverNet-izens! Since our last communique, the InvertNet technical team has been making slow but steady progress on improving our website and building our workflow for capturing images of whole drawers of pinned insects. Research programmer Omar Sobh and his team have been upgrading our cyberinfrastructure platform to a new new, improved version of HUBzero. The new version is running on our testbed and we plan to go live within the next week. Our student programmers are also making progress on our image segmentation/annotation and crowd-sourcing tools to be incorporated into the InvertNet.org website as well as on a link to BugGuide, which will allow users of InvertNet to compare images of live arthropods on BugGuide to authoritatively identified specimens in our virtual museum
Our second (and, hopefully, final) prototype insect drawer digitization robot is running and we are now optimizing its performance. InvertNet Co-PI Chris Taylor featured this system in a talk he gave at the recent Wet Collections Digitization Workshop at KU and lead developer, David Raila and I will demonstrate the system at the DROID 2 (Pinned Insect Collection) Workshop at the Field Museum later this month. If things continue to progress as they have over the past few months, we will be in high-throughput digitization mode this summer. Several collaborating institutions have been submitting large numbers of images of scanned slides and vials to our online image repository and we will soon be adding images of whole drawers.
This month we also welcome a new InvertNet partner institution, the University of Iowa, which received a PEN (Partnership with Existing Network) grant and will begin efforts to digitize their insect and crayfish collections in July. Lead PI, Trina Roberts, reports that the collections are small but historically important, having been assembled mostly during the second half of the 19th century. Trina's group will team up with InvertNet collaborator Iowa State U. to share digitization equipment.
Drawer Digitization System
The second Invertnet Digitization parallel robotic system prototype has been optimized in a number of ways and is now undergoing the final round of quality tests. We have optimized the machine's performance in terms of speed, accuracy, dexterity and coverage while significantly reducing the footprint and weight compared to the previous prototype.
The machine control software has been improved with the integration of a customized, real-time operating system that can take advantage of modern 64-bit multi-core computer architectures, unlike the previous system which was limited to obsolete single-core 32 bit computers. The operating system improvements bring high-performance, real-time image processing code that can take advantage of multiple processors for much improved performance.
The camera and positioning systems have been integrated with a high performance computer vision system that enables the system to monitor the images being captured in real-time and adapt dynamically in order to optimize the performance and accuracy of the capturing process. The vision system monitors focus, image quality, drawer position -- automatically recapturing images for problematic areas and adapting to variation in drawer dimensions or positioning.
The real-time adaptation is used to dynamically adjust camera positioning in real-time to adjust image overlap based on specimens near the edge of each image. This assures that stitching errors due to inadequate information in the edges of images are minimized, which improves both the performance and resulting quality of the final composite tray images.
A video of the system in action is available on the InvertNet website; the machine is configured for low-speed and with additional delays for testing purposes. The final working speed will be slightly higher than shown in the video.
The final prototype with exposed parts and wiring. System arrives pre-wired and with pre-assembled, quick-connect components for operation in 20 minutes from opening the box.
Georeferencing Training Webinar
All InvertNet collaborators are invited to participate in one of several upcoming georeferencing training webinars. Each hour long webinar will provide a brief overview of some of the most current and valuable software, web tools, and other resources available for streamlining retrospective georeferencing workflows. If you are interested in participating, please complete and submit the online registration form by May 15th, 2013.
Call for Content: Collaborator Buzz
Want to showcase your institution's progress along the InvertNet journey? Send photos and commentary to InvertNet.email@example.com for inclusion in the Collaborator Buzz column of future newsletters. Submission deadline for August edition is July 25th, 2013.
Follow InvertNet on Twitter: @InvertNetTCN