Detection of sunn pest-damaged wheat grains using artificial bee colony optimization-based artificial intelligence techniques.

ABSTRACT

BACKGROUND: In this study, artificial intelligence models that identify sunn pest-damaged wheat grains (SDG) and healthy wheat grains (HWG) are presented. Svevo durum wheat cultivated in Konya province, Turkey is used for the process, with 150 HWG and 150 SDG being used for classification. Thanks to the constructed imaging setup, photos of the 300 wheat grains are obtained. Seventeen visual features of each wheat grain are extracted by image-processing techniques and evaluated in three different groups of dimension, texture and pattern as visual parameters. Artificial bee colony (ABC) optimization-based artificial neural network (ANN) and extreme learning machine (ELM) algorithms are implemented to classify the damaged wheat grains. RESULTS: A correlation-based feature selection (CFS) technique is also utilized to find the most effective among the 17 features. In the classification process using five selected features, the mean absolute error (MAE) and root mean square error (RMSE) values for ABC-based ANN are calculated as 0.00174 and 0.00433 respectively. The proposed technique is integrated into graphical user interface (GUI) software to construct an effective detection system for practical use. CONCLUSION: The results indicate that, thanks to the modified ANN algorithm and implemented CFS algorithm, the detection accuracy of damaged wheat grains is considerably increased. © 2019 Society of Chemical Industry.