Written by  John Muehlebach

Teaching computers to recognise pests

Machine vision technologies are now capable of identifying birds, animals and any other object better than expert humans.

The development of this technology started in the 1950s but serious results only started to be achieved in 2012.
Improvements have been accelerated by improvements in graphics processors and computing power.
Today using commonly available computer hardware, security cameras and customised open source software, growers can automatically recognise and deter pests from their orchard.
Software used to detect objects is called a deep neural network (DNN) because of its resemblance to the brain.
Much like a brain the DNN learns to recognise objects through training. For example, if the DNN were trained to recognise the difference between crows and starlings, thousands of images of both would be input to the DNN, which would be progressively optimised to make the correct decision, either starling, crow or neither at its output.

For contact details, see Tree Fruit Nov 2018