Automated weed and crop recognition and classification model using deep transfer learning with optimization...

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Weed and crop competition for resources leads to reduced crop yields and increased farming costs.
• Conventional weed control methods, such as extensive pesticide use, cause environmental pollution and contribute to herbicide resistance.
• There is a growing demand for organic farming and pollutant-free produce, necessitating innovative weed management solutions.

Purpose of the Study:

• To develop an Automated Weed Recognition and Classification using a Deep Learning Model with Lemurs Optimization (AWRC-DLMLO).
• To accurately detect and classify weeds alongside crops using advanced artificial intelligence techniques.
• To enhance the efficiency and effectiveness of weed management in agriculture.

Main Methods:

• Image pre-processing using Gaussian filtering (GF) to reduce noise.
• Plant segmentation using Residual Attention U-Net (RA-UNet).
• Feature extraction with ShuffleNetV2, hyperparameter optimization using Lemurs Optimization Algorithm (LOA), and classification with Cascading Q-Network (CQN).

Main Results:

• The proposed AWRC-DLMLO method demonstrated superior performance in weed detection and classification compared to existing models.
• Simulations confirmed the effectiveness of the deep learning model with lemurs optimization in agricultural applications.
• The integration of GF, RA-UNet, ShuffleNetV2, LOA, and CQN achieved high accuracy in identifying weeds and crops.

Conclusions:

• The AWRC-DLMLO technique offers a promising AI-driven solution for automated weed management in smart agriculture.
• This approach can significantly contribute to sustainable farming practices by minimizing crop loss and reducing reliance on chemical herbicides.
• The study highlights the potential of deep learning and optimization algorithms for advancing precision agriculture and ensuring food security.