Optimal pollination requires just the right number of insect pollinator visits to flowers. Too few or too many visits, or visits by ineffective insect pollinators, can diminish the quality of food a flowering plant produces.
Typical techniques for monitoring insect pollination use direct visual observation or pan trapping, which are labour-intensive and take many days.
Additionally, without a very large number of trained observers it is impossible to collect simultaneous data across large farms. Yet such data are needed to provide time-critical evidence of the extent of crop pollination, before a season’s pollination window is closed.
With our digital system, however, a farm manager could obtain same-day data on crop pollination levels.
How fine-grained analysis of insect pollinator movement enables better food production
Our pollination monitoring system was set up at Sunny Ridge farm in a strawberry greenhouse open to insects. The array of cameras monitored insect activity among the strawberries, recording honeybees, hover flies, moths, butterflies and some wasps.
Managing big (insect) data with advanced software
The volume of data our system collects requires custom software to reliably track individual insects flying among complex foliage.
A key issue our software overcomes is identifying insect movements within a video sequence, so an individual insect on a single path isn’t accidentally counted multiple times. This enables accurate assessment of the number of insects in a region during a day, an analysis of their type (e.g. species), and monitoring of their flower visits.
Our custom software uses a hybrid detection model to detect and track insects and flowers in videos. This model combines the AI-based object-detection capabilities of deep learning using a convolutional neural network, together with separate foreground detection algorithms to identify the precise positions of insects and the flowers they visit in the recorded videos.
The insect paths our software produces are computed using a method called the Hungarian algorithm. This examines the positions of insects in each video frame in a sequence, and enables the identification of a match between the locations of the insects across a sequence of video frames.
By recording and visualising these paths, we gain an understanding of insect behaviour and the efficiency of pollination in a greenhouse.
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.
Any cookies that may not be particularly necessary for the website to function and are used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies.