SBIR-STTR Award

The project's goal is to enable machine learning ("AI") enabled pollinator ID and behavior detection on agricultural producer lands where detection units send insect IDs and behavior counts back to the producer. Our objective is to build on an existing USDA open-source automated insect identification device that can already send machine vision results back to the farmer. All camera-based computer vision devices have limited usefulness in the field due to the very heavy power consumption of cameras. To get past this barrier our primary technical objective is to use new LiDAR on a chip technology to reduce and sometimes eliminate the need to use device cameras when detecting pollinators and recording pollination behavior events. The plan to accomplish this goal is to integrate the LiDAR chip onto the existing open-source board. The LiDAR can provide data-rich information to the device without turning on a camera if we develop specialized software detection models and hardware. We will develop an "insect on the end of a wire" test fixture to mimic flight and pollination patterns of live insects with dead specimens. Using that fixture we will conduct thousands of identical trials to check our LiDAR integration success. Then we will use field trials with a rented bee hive to verify the value of this technology to insect detection and pollination even recording. This work aligns with section 8.13.3 a through c of the RFA enabling measurement of native pollinator abundance and behavior mapping pollinator deficits and artificial systems for improving pollination outcomes.