نوع مقاله : مقاله پژوهشی
نویسنده
گروه مهندسی برق، دانشکده مهندسی برق و کامپیوتر، دانشگاه صنعتی سیرجان، سیرجان، ایران.
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسنده [English]
This study proposes an efficient and novel method for recognizing pistachio varieties by leveraging Gabor filters for feature extraction and a genetic algorithm for feature selection, aiming to enhance classification accuracy and speed. Unlike conventional approaches that focus on individual processing of pistachio kernels—which, despite high accuracy, are time-consuming and computationally demanding—the proposed method employs holistic processing of images containing multiple pistachios, thereby significantly accelerating the recognition process. Key textural and shape-based features are initially extracted using Gabor filters. Subsequently, the genetic algorithm is applied to select optimal features and reduce data redundancy. The refined feature set is then classified using the K-Nearest Neighbor (KNN) algorithm. Experiments conducted on a dataset of 1,000 sub-images covering five common pistachio types (Akbari, Ahmad Aghaei, Kaleh Ghouchi, Fandoghi, and Badami) yielded an average classification accuracy of 99.5%. In comparison with deep learning models, the proposed method demonstrates competitive performance while requiring no extensive training phase or high computational resources. This results in a faster and more resource-efficient implementation, making it particularly suitable for industrial applications, especially in automated pistachio processing and packaging systems. The proposed method contributes to the automation of agricultural workflows, reducing both processing time and labor costs, and offers a practical solution for real-time classification in resource-constrained environments.
کلیدواژهها [English]
Pistachio is one of Iran’s most significant non-oil export products, contributing substantially to the country's economy. Iran ranks among the top global pistachio producers, with several varieties, including Ahmad Aghaei, Fandoghi, Akbari, Badami, and Kalleh Ghouchi, dominating the market. Given the economic and industrial importance of pistachios, accurate classification of different types is crucial for improving quality control, optimizing processing efficiency, and minimizing waste. Traditional classification methods face challenges in identifying pistachio types with high accuracy. Therefore, leveraging advanced image processing, deep learning, and machine vision techniques has become a promising research direction. Among various texture analysis techniques, Gabor filters have proven effective in extracting meaningful patterns from images, offering a computationally efficient alternative to deep learning-based approaches.
This study proposes a machine vision-based method for pistachio classification using texture analysis. The approach consists of several steps:
The proposed method achieved an overall classification accuracy of 99.5%, significantly outperforming a reference method (Shamsi-Goshki et al., 2013) with 94.8% accuracy. The genetic algorithm effectively reduced computational complexity while maintaining classification performance. Performance metrics, including precision, recall, and F1-score, confirmed the superiority of the proposed approach over previous methods. Additionally, the proposed bulk classification approach is more suitable for industrial applications compared to conventional methods that classify individual pistachios.
This study demonstrates the effectiveness of combining Gabor filters, genetic algorithms, and KNN classification for pistachio type recognition. The approach achieves high accuracy with significantly lower computational costs than deep learning models, making it suitable for large-scale industrial applications. Future research can explore integrating deep learning with texture-based methods for further improvements.
Conceptualization, methodology, software, validation, formal analysis, investigation, resources, data curation, writing—original draft preparation, writing—review and editing, visualization, supervision, and project administration by Asma Shams-Kermani. Author has read and agreed to the published version of the manuscript.
Data available on request from the authors.
The authors would like to thank all participants of the present study.
Ethical approval was not required for this study as it did not involve human or animal participants, or any sensitive personal data.
The author declares no conflict of interest.