No abstract

Feature extraction and machine learning for classification plays an essential role in Automated Insect Identification (AII) because of its capability of insect classification in different taxonomic levels. Part Separating algorithm (PS) feature extraction integrated into Support Vector Machine (SVM) classifier could not demonstrate general results. The performance of SVM combined with PS was high only in Family Sphingdae of Order Lepidoptera but its accuracy rate may be dropped when used for classifying in other families. Therefore, this paper applied Extreme Learning Machine (ELM) having PS as the feature extraction process for AII for the butterfly family identification of Order Lepidoptera. In the pattern recognition process of image processing, the recognition ability of ELM classification with various activation functions and SVM were also investigated and compared. The experimental results showed that the classification in ELM using five insect features via the PS algorithm can be improved the as well as the ability to generalize every butterfly family of ELM performance, showing higher recognition rates than the SVM method in every family of order Lepidoptera.

No abstract