Hyperspectral Analysis for Detection of Sea Cucumber Habitat (Holothuria scabra) based on Support Vector Machine

Abstract

Sea cucumbers are sea animals that have a high nutrient and high selling value in both the local market and international markets. In Indonesia, sea cucumbers are found in almost all Indonesian waters, but sea cucumbers in Indonesia have not been grouped by their habitat due to lack of technology that can categorize sea cucumber habitats quickly, precisely and does not damage sea cucumbers. The method is generally destructive and is carried out manually in laboratory tests. In this paper, a classification system from the origin of sea cucumber habitats is introduced using non-destructive Hyperspectral imaging by detecting electromagnetic waves with a spectral range of 400 to 1000 nm. The system algorithm consists of measurement of reflected image profiles, feature extraction, feature selection for spectral and spatial data, object profiles will be combined to select excellent features using the PCA (Principal Component Analysis) method. The data used will be classified into two habitat classes, namely, Pontianak and Belitung using the SVM (Support Vector Machine) method. Data samples will be evaluated with cross-validation to measure system performance. Based on experiments, the accuracy obtained from the classification and evaluation of the SVM method is 92%. The results of this work indicate that this system can be proposed as a classification system for the origin of habitats that do not damage sea cucumbers and are suitable for use in industrial sorting systems.