Abstract
The effectiveness of insect pest management programs depends on the availability of reliable and updated information about the pest infestation status. Action thresholds derived by captures in monitoring traps are a pillar of modern integrated pest management programs to trigger and optimize the timing and usage of insecticide sprays. However, weekly trap inspections in field may lead to a delayed intervention and imply a certain labour cost. Such issues have led to some early adoption of automatic trap-based moni-toring exploiting new technology. This work aimed to develop an innovative ‘smart’ trap prototype capable to monitor by remote insect pests, selecting codling moth, Cydia pomonella (L.), in pome fruit crops as case study. Smart trap components (hardware) were chosen considering the environmental sustainability and an economic evaluation of the trap prototype cost is provided together with the cost-benefit analysis of the remote pest monitoring. A detection algorithm (software) to automatically identify and count codling moth was developed by using open-source programs. The smart trap prototype was evaluated in field experiments. Quali-tative parameters related to automatic pest identification such as accuracy, sensitivity and precision were calculated according to both false positive and false negative counts. This work describes the different steps necessary to develop smart traps for insect pests monitoring, showing the preliminary field results obtained with the proposed prototype. The smart trap efficiency in capturing codling moth was similar to a standard monitoring trap and the pictures provided a sufficient resolution to manually validate moth captures observing the images by remote. Nevertheless, the detection algorithm failed to automatically provide a trustworthy capture count data by remote because, using deep learning technique, thousands of pictures are usually required for the algorithm training towards the target species in order to reach a sufficient level of reliability. This work provides the basis for a further wider devel-opment of such smart trap prototypes worldwide.