Optimized heterogeneous bi-directional recurrent neural network for early leaf disease detection and pesticides recommendation system

In this manuscript, optimized heterogeneous bi-directional recurrent neural network for early leaf disease detection and pesticides recommendation system (HBDRNN-ELD-PRS) is proposed. Initially, the input images are collected from plant dataset. To execute this, the collected input image is pre-processed using multimodal hierarchical graph collaborative filtering (MHGCF) for removing the noise, then the pre-processed images are fed to the feature extraction using second-order synchrony-extracting wavelet transform (SOSEWT) to extract the geometric features, such as area, slope, cancroids and perimeter. Then the extracted images are fed to the heterogeneous bi-directional recurrent neural network (HBDRNN) for effectively categorize Leaf Disease Detection as pepper bell bacterial spot, pepper bell healthy, potato late blight, potato early blight, potato healthy. Generally, HBDRNN does not adapt any optimization methods to compute optimal parameters to ensure accurate leaf diseases classification. Hence, the harbor seal whiskers optimization algorithm (HSWOA) is proposed to optimize the heterogeneous bi-directional recurrent neural network which accurately classifies the leaf disease. The proposed HBDRNN-ELD-PRS is implemented in Python. The performance metrics, such as accuracy, precision, specificity, recall, F1-score, computation time, ROC are analyzed. The proposed HBDRNN-ELD-PRS approach achieves 99.87% accuracy, 98.09% precision, and 97.83% recall when compared to the existing techniques.