*Article* **Neural Visual Detection of Grain Weevil (***Sitophilus granarius* **L.)**

**Piotr Boniecki <sup>1</sup> , Krzysztof Koszela 1,\*, Krzysztof Swierczy ´nski ´ <sup>1</sup> , Jacek Skwarcz <sup>2</sup> , Maciej Zaborowicz <sup>1</sup> and Jacek Przybył <sup>1</sup>**


Received: 23 November 2019; Accepted: 14 January 2020; Published: 20 January 2020

**Abstract:** A significant part of cereal production is intended for agri-food processing, which implies a necessity to search for and implement modern storage systems for this product. Stored grain is exposed to many unfavorable factors, particularly caryopsis macro-damage caused mainly by grain weevil (*Sitophilus granarius* L.). This triggers a substantial decrease in the value of the stored material, thus resulting in serious economic losses. Due to this fact, it is necessary to take steps to effectively detect this pest's presence when grain is delivered to storage facilities. The purpose of this work was to identify the representative physical characteristics of wheat caryopsis affected by grain weevil. An automated visual system was developed to ease the detection of damaged kernels and adult weevils. In order to obtain the empirical data, a decision was made to take advance of SKCS 4100 (the Perten Single Kernel Characterization System). The measurements obtained were used to build the training sets necessary in the process of ANN (artificial neural network) learning with digital neural classifiers. Next, a set of identifying neural models was created and verified, and then the optimal topology was selected. The utilitarian goal of the research was to support the decision-making process taking place during grain storage.

**Keywords:** artificial neural network (ANN); *Grain weevil* identification; neural modelling classification
