|
 |
|
|
|
Leveraging Pattern Recognition based Fusion Approach for Pest Detection using Modified Artificial Hummingbird Algorithm with Deep Learning |
|
PP: 509-518 |
|
doi:10.18576/amis/190303
|
|
Author(s) |
|
Arwa Alzughaibi,
|
|
Abstract |
|
Pest detection is a vital feature of agriculture and ecosystem organization, intended to classify and mitigate the power of harmful organisms on crops and the atmosphere. By automating the recognition procedure, researchers and farmers can improve accuracy in pest organization, enhance resource allocation, and finally donate to sustainable and strong farming practices. Leveraging innovative technologies like computer vision (CV) and machine learning (ML), pest recognition techniques can examine images and sensor information to recognize the presence of pests in agricultural areas. Pest classification utilizing pattern detection and DL contains the growth of sophisticated methods to mechanically recognize and classify numerous pests. This incorporated technique connects the powers of traditional pattern detection approaches and the learning abilities of deep neural networks (DNNs). DL, particularly Convolutional Neural Networks (CNNs), has shown amazing victory in learning complex patterns straight from raw data. By using DNNs, the method can mechanically learn hierarchical representations, enabling it to discern complex features and relationships in pest-related data without clear feature engineering. In this aspect, this study introduces a fusion of the Modified Artificial Hummingbird Algorithm with Deep Learning-based pest detection and classification (MAHADL-PDC) technique. The MAHADL- PDC technique aims to effectually recognize distinct pests’ types. The input image quality is enhanced by the adaptive median filtering (AMF) approach. In addition, feature extraction using the EfficientNet-B4 model is performed to learn complex features, and its hyperparameters were chosen by utilizing MAHA. The MAHADL-PDC method uses the deep belief networks (DBNs) model to detect and classify pests. To highlight the significant performance of the MAHADL-PDC method, a series of experiments were made. The performance validation of the MAHADL-PDC approach portrayed superior outcome over existing models.
|
|
|
 |
|
|