Introducing a smart monitoring system (PHLIP) for integrated pest management in commercial orchards

Abstract

In this study a developed modularized mobile system has been introduced in the framework of the research project PHLIP that enables spatiotemporally high-resolution population monitoring of insects (pests) in orchards, using deep learning (DL) object detection, which can be used as the basis for implementing a site-specific application of insecticides. As a case study, an image annotation database was built with images taken from yellow sticky traps and annotated cherry fruit flies. A faster Region-based Convolutional Neural Network (R-CNN) DL model was applied. The results showed average precision of 0.88 which, indicates that the DL model can perform as a component of an automated system for assessing pest insects in orchards. An important outcome of PHLIP will be the creation of application maps for site-specific insecticide application. Therefore, decreasing the amount of insecticides applied in orchards - which are critically assessed in terms of their environmental impact - should be possible, while the yield efficiency would not be changed. The spatial monitoring will create desirable conditions for a sustainable pest management in horticultural management.