<![CDATA[Citrus vein phloem degeneration (CVD) is a devastating disease of citrus plants and seriously impacts crop quality. Although manual detection is feasible, this method faces many challenges, such as the similarity of early symptoms between healthy and infected leaves. Therefore, manual detection is time-consuming and inefficient. Therefore, an accurate and efficient automatic detection method is needed. This study aims to combine two methods: the K-Means segmentation method and the EfficientNet architecture to build an automatic detection model for CVD in citrus leaves. This method aims to improve the classification accuracy of citrus leaf images. This study is divided into two stages: the first stage uses the K-Means algorithm for image segmentation, and the second stage uses the EfficientNet model for classification. The K-Means segmentation method is used to separate the leaf surface from the background, focusing only on the parts of the leaf that show disease symptoms. The segmentation results are then processed in the second stage using the EfficientNet model. The EfficientNet model is known for its efficient feature extraction and excellent performance in recognizing complex visual patterns. The results showed that combining the K-Means segmentation method with the EfficientNet architecture significantly improved the accuracy of CVPD detection compared to a traditional CNN model without segmentation. This system is expected to assist farmers in detecting CVPD and support the implementation of smart agriculture technology in automated plant health monitoring.]]>
