Automatic Pest Identification and Counting

EasyChair Preprint no. 7792

Abstract

The aim of integrated pest management is to develop an appropriate, unique plant protection strategy with knowledge of all available parameters. This means that we only release biochemicals into the environment when they are really needed and only those that have an impact on the pests present as well as those that are expected to appear. A number of data are needed to develop a good strategy, including the determination of pest density. The system we have designed aims to provide a mobile, outdoor device capable of detecting the density of insect pests, especially moths, and transmitting this data wirelessly and later displaying it. The data is stored in a central database, from where it can be easily viewed and analyzed on either a web or mobile application interface. Digital image processing algorithms and a convolutional neural network were used to count the pests. The device must operate continuously for a full season (4-5 months) without mains power, so we had to strive for low power consumption throughout the design of the endpoint device. In our work, we have tried to minimize production and operating costs. The system presented in this article can therefore help the planning phase of the integrated pest management process in orchards and verify its effectiveness by being able to provide continuous data on the evolution of pest density with minimal human intervention. The system we have designed can make a major contribution to increasing yields and provide significant support for agricultural engineering.

Keyphrases: digitális képfeldolgozás, kártevő azonosítás, NB-IoT, neurális hálózat

@Booklet{EasyChair:7792,
  author = {Ferenc Braun and Zsolt Molnár},
  title = {Automatic Pest Identification and Counting},
  howpublished = {EasyChair Preprint no. 7792},

  year = {EasyChair, 2022}}