PENGENALAN BIBIT PEPAYA CALIFORNIA MENGGUNAKAN TEKSTUR URAT DAUN DENGAN METODE JST-PB DAN GLCM

Abstract

California papaya plants have three sex types: female, male and perfect. Early identification of the sex of papaya plants is very important to improve production efficiency. However, this process is difficult to do because sex characteristics only appear 4-6 months after flowering. This study aims to identify California papaya seedlings as male or female by analyzing the texture of leaf veins using the Back Propagation Artificial Neural Network (JST-PB) method and Gray Level Co-occurrence Matrix (GLCM). The JST-PB method is used to model complex patterns, while GLCM measures the spatial distribution of pixel intensity in leaf images. The dataset consists of 300 images of young California papaya leaves, with the number of each class as follows:  100 images for the female class, 100 images for the male class, and 100 images for the perfect class. All images were cropped to 200x200 pixels, focusing on young leaves, while old or dry leaves were not used. The results showed that the JST-PB method was able to achieve an overall accuracy rate of 71% in California papaya sex recognition. The accuracy for female papaya reached 76%, while that for male papaya was 66%. However, further testing showed that the perfect class could not be identified significantly, so this study concludes that two-class classification (female and male) is more reliable than three-class classification.These findings suggest that JST-PB has potential in California papaya sex recognition and classification, although improvements are needed especially in male papaya identification. Further research is recommended to increase the amount of training data, explore variations in JST architecture, and use different cross-validation methods or test datasets.