Hybrid architecture for crop detection and leaf disease detection with improved U-Net segmentation model and image processing

Abstract

Agriculture stands as a cornerstone of India's economy, supporting the livelihoods of millions and feeding a vast population. Enhancing crop production is imperative, given the substantial portion of land dedicated to agriculture. However, the multifaceted nature of farming, influenced by variables like soil composition, climate, and diseases, poses significant challenges. Embracing technological advancements is pivotal to augmenting crop yields and ensuring sustainable agricultural practices. This study introduces an innovative hybrid architecture that addresses significant challenges in agriculture by identifying leaf diseases and detecting crops using deep learning and sophisticated image-processing techniques. Here, the proposed model comprises two phases namely, Crop Prediction and Leaf Disease Identification. To improve its suitability for analysis, the input image is first preprocessed. An improved U-Net segmentation algorithm has been employed to identify areas of interest in the image. Features pertinent to shape, color, and texture, including an enhanced Local Gabor XOR pattern (LGXP), are then extracted to capture comprehensive information about the crops and potential diseases. The core of our approach lies in a hybrid architecture, integrating elements of Improved Linknet and LeNet architectures. This model first determines the kind of crop in the image by using features that have been extracted. In the following step, deep features and statistical characteristics extracted from the segmented image are used to identify numerous prevalent diseases affecting the foliage. Implemented in Python, our approach is rigorously evaluated against conventional models, showcasing superior performance across various metrics. Consequently, the model has achieved a higher detection accuracy of 0.982 and the F-measure of about 0.956, indicating that the model operates better and identifies the leaf disease more successfully than other existing techniques. This research endeavours to empower farmers with actionable insights, fostering smarter agricultural practices and contributing to food security and economic prosperity.