This Small Business Innovation Research (SBIR) Phase I is directed toward high-throughput robotic imaging and data capture of 3D physical insect specimens with applications to national intelligence and security, vector epidemiology, agriculture, biodiversity research, and museum curation. Natural history museums are vast, largely untapped sources of biological data for addressing complex questions ranging from tracking terrorists to climate change. Yet ~1 billion U.S. specimens remain in museum drawers and cabinets, their characteristics, associated data, and even their existence hidden. Insects preserved on pins represent a significant proportion, with the ten largest collections housing over 100 million. While the scope of collections is immense, the expertise needed to identify insects of importance is in decline as taxonomists retire and university positions close. For some insect groups, few or no experts remain capable of identifying species. The proposed technology will radically accelerate the pace of insect specimen digitization. It will capture critical data from museum specimens and revitalize taxonomy by providing datasets for computer-automated insect identification efforts. The project demonstrates robotic handling of pinned specimens is feasible, enabling a high-throughput system capable of manipulating pinned specimens, partitioning the 3D scene, photographing specimens, and extracting database information with OCR rendering of specimen labels.
The broader impact/commercial potential of this project spans myriad applications. Insects vector important diseases of humans, livestock, and crops. Annually, malaria is implicated in a million human deaths worldwide and aphids and their viruses cost US farmers over one billion dollars. New invasive insects regularly arrive on our shores undetected. Insects also have forensics applications, locating events in time and place and calculating times of death. Tools to quickly and efficiently identify insects have immediate application in vector epidemiology, agricultural and environmental monitoring, border security and port inspection, law enforcement, and intelligence. Spatial-temporal data associated with field collection of insect specimens are crucial for geolocation forensics, tracking climate change and range expansion of invasive species, and locating geographic sources of potential biological control agents for exotic pests. Current technologies and specimen digitization practices take too much time, cost too much money, and simply cannot image most specimens. Semi-automated systems address only microscope slide- or paper-mounted specimens. Currently no technology exists to even partially automate the digitization of individual pinned insect specimens. Potential customers include insect collections, museums, agribusinesses, epidemiologists, and agents in homeland security, port inspection, environmental monitoring, and national intelligence. Enormous secondary commercial potential exists with value-added online databases.
This SBIR Phase 1 project was directed toward high-throughput robotic imaging and data capture of 3D physical insect specimens. Our work results can potentially enable automated processing systems capable of photographing and documenting hundreds of insects in the time it currently takes a skilled human to document just one. For the first time, producing a digital image catalog of the millions of insect specimens closeted in national collections such as the Smithsonian may become feasible. Natural history museums are vast, largely untapped sources of biological data. Roughly 1 billion U.S. specimens preserved in institutional drawers and cabinets have never been digitally documented in a way that makes this information readily available. The ten largest insect collections alone house over 100 million specimens. These collections, in some cases built up over centuries of field research, contain a huge variety of specimens. Meanwhile, the expertise needed to identify insects has become increasingly spread out; typical institutions employ one or two specialists, expert in only a few species. Addressing these twin problems requires development of an automated digitization system that can (1) catalog specimens at an unprecedented rate and (2) image specimens in enough detail to facilitate identification by physically remote experts. No previous technology exists to even partially automate the digitization of individual pinned insect specimens; it has remained a slow, laborious and expensive process. Our project demonstrated that a robotic solution is feasible. Over the course of our work, we developed a prototype robotic system capable of identifying and accurately locating the various parts of a pinned insect specimen, such as the pin, the label, and the insect itself. Our system can then safely, quickly, and autonomously capture label data and photograph the insect from any number of user-specified angles. This process represents a significant challenge for an automated system: each specimen is unique in size and shape and in how it is mounted and labeled. Nonetheless, the algorithms we developed are able to position the correct target areas of a pinned specimen in front of a close-up camera lens with better than millimeter precision. During Phase 1 we also established important engineering details that will guide our development of a practical field-deployable system, including application-specific solutions for key aspects of machine vision, robotic control, optimal photo imaging, mechanical mounts and manipulators, and safe handling parameters for the delicate specimens. Enormous potential exists in the creation of online databases that render millions of specimens readily accessible to students, scientists, and others. Accordingly, the broader impact of this project spans myriad applications. Insects vector important diseases of humans, livestock, and crops; annually, malaria is implicated in a million human deaths worldwide and aphids and their viruses cost US farmers over one billion dollars. Global travel and climate shifts are driving unprecedented spread of insects. New invasive species regularly arrive on our shores undetected. Insects also have important forensics applications in tracking the movements of containers and people, in locating events in time and place, and in calculating times of death. Tools to efficiently identify insects and cross-reference them with a large specimen database have immediate application in vector epidemiology, agricultural and environmental monitoring, border security and port inspection, as well as in law enforcement and intelligence. All of this begins by digitizing the data. Contact: www.sr2group.com
