Plant diseases are a critical problem in modern agriculture, responsible for significant productivity losses of crops and economic damages. This paper describes the PlantShield system that employs both texture-based as well as deep learning methods to enhance the robustness and precision of plant disease identification. The PlantVillage dataset, which has approximately 54,000 photos of both healthy and damaged leaves from different plant species, serves as the foundation for the system. In this approach, textural properties like contrast and homogeneity are extracted using the Gray Level Co-occurrence Matrix, while deep visual features are extracted using CNN. This KNN method is used for fusing and classification of the features. The system would be supposed to offer a better performance when it comes to differentiation between subtle disease patterns through fusing low-level texture features with higher-level CNN features. Evaluation for the accuracy, recall, precision, and F1-score shows that the proposed solution by PlantShield is robust and reliable in early plant disease detection, promising its applications in precision agriculture and real-time field monitoring.