Research project to evaluate small scale automated food production systems in semi controlled environments with a particular focus on integrated pest management. Automated greenhouse production has taken huge leaps forward with in the last few years. It is now possible using very cheap micro controller kits such as Arudiono to build simple mechanisms to test soil water, temperature, humidity to consequently activate systems such as sprinklers, fans , heaters etc.
In environments that are not strictly controlled (and even those that are) pests can become a significant problem. This presents two challenges. Firstly detecting outbreaks, and secondly devising a remedial solution. The two are connected in that it can be thought of as the latter being a solution set to the former. The initial phase of this project will concentrate on detection.
Use feature recognition software to analyse images and detect for infestations of a target.
collect (and generate) a training set (images and environmental data)
Train a neural model to recognise the set of features that indicate infestation.
Test and retrain.
Having a model that links infestations to a feature set we can adjust each of the environmental features, (with in an allowable range and weighted priorities to minimise crop damage) to test for a viable solution. In the case that there is no suitable change in conditions, ie there is no solution set, then an insecticide may be used. Ultimately this solution can be fed back in to the model and over time the system could evolve to constantly optimise its environment and minimise insecticide use.