Machine Vision Approach Using Multi Features for Detection of Oil Palm Stem Disease

Indonesia is a significant palm oil producer, widely cultivated due to its status as a prominent vegetable oil-producing plant. Palm oil is a highly sought-after crop in agriculture due to its profitability and the demand for substantial quantities of high-quality oil. Early diagnosis of oil palm diseases is crucial for prompt prevention and eradication measures essential for maintaining high-quality palm oil production. Hence, there is a requirement for a machine vision-based method to classify diseases on plant stems. The proposed method consists of several main processes: pre-processing, feature extraction, and classification. Several machine learning algorithms and feature extraction using color and texture features were performed to develop a machine vision for detecting oil palm stem disease. It applied cross-validation with a K-fold value of 10. Performance evaluation was carried out using three parameters: precision, recall, and accuracy. The Linear Discriminant Analysis (LDA) method achieves the optimum results in testing scenarios combining color features based on HSV color space and textures features, achieving an accuracy value of 90.50%.