Effectual detection and classification of pests are significant requirements in smart agriculture that directly influence crop productivity. This research offers an enhanced machine learning and IoT-based framework for the precise identification of pests to improve crop health monitoring. Nonetheless, the prevailing detection and classification of pest models often suffer from significant drawbacks. High model complexity, overfitting of data, and poor performance in case of adverse conditions are the major limitations. Also, differentiating the comparable pest species from visual images tends to be difficult. These issues degrade the scalability and reliability in real-world agricultural conditions. The images from the collected datasets are pre-processed initially using a Gaussian Amended Wiener Filter (GAWF) to reduce the noises and enhance the image contrast. The refined images are then directed over the Optimized Cross-Stage Partial connections-enhanced YOLOv8 (OCSP-YOLOv8) model for precise pest region detection. To improve the detection accuracy, hidden neurons in the architecture are optimized using the Adaptive Remora-based Dung Beetle Optimization (AR-DBO) algorithm. The discriminative features from the identified pest regions are extracted using a Dense Residual-based Dilated Pyramid Network (DRDPN). The experimental outcomes indicate the performance of the proposed model on obtaining a Dice coefficient of 0.9916, Jaccard index of 0.993, and Intersection over Union (IoU) of 0.9898 in the case of pest detection. The classification model achieves an accuracy rate of 99.17%, a recall of 98.71%, a precision of 98.82%, and an Formula: see text1-score of 98.76%. From the obtained outcomes, it can be confirmed that the proposed model provides an effective solution for managing pests in smart agricultural systems, thereby enhancing crop productivity.
Singh et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: