Machine Learning-Based Grasshopper Species Classification using Neutrosophic Completed Local Binary Pattern

Abstract

Locusts are seen as a major threat to the ecosystem because they devastate crops and contribute to thousands of tons food lost every year. Numerous well-trained agents are needed for the efficient control of these insects. However, this is a challenging process. Grasshopper detection methods are being developed using traditional forecasting methods by expert entomologists. The maximum potential of these methods has not yet been completely realized. Hence the majority of work is still done manually. In this paper, a neutrosophic CLBP (completed local binary pattern) based grasshopper species classification framework is proposed. Our proposed system comprises a novel grasshopper species database of over 7.392 images for grasshopper species classification. The grasshopper image is first converted to a neutrosophic field. These discriminatory features are merged with rotation invariant LBP. Our proposed system could achieve up to 99.7% classification accuracy even while working with challenging datasets of wide image quality and size range. The proposed methodology involved diagnosing 11 species and subspecies. It demonstrates the impracticability of conventional diagnostic techniques in the later stages. It could have a big impact on data analysis, enabling more effective handling of global pest.

Keywords

• neutrosophic set
• local binary pattern
• locust
• deep learning
• machine learning