Digital Maize Crop Guardian: Automated Identification of Fall Armyworm Infestation Using Computer Vision

Fall armyworm (FAW) infestation poses a significant threat to maize cultivation worldwide, leading to substantial yield losses. Timely identification of FAW presence is crucial for implementing effective management strategies to mitigate its impact. This study explores the use of computer vision techniques to automatically detect FAW infestation in maize crops by analyzing visually observable patterns indicative of FAW damage amidst various biotic and abiotic stresses. Leveraging deep convolutional neural networks (DCNNs) and transfer learning, an algorithm is proposed to identify FAW-infected areas in maize fields. The algorithm exhibits high accuracy rates, achieving 98.47% training accuracy and 93.47% validation accuracy. Evaluation on actual clear FAW images yields an average classification accuracy of 82%, with 70.8% accuracy on augmented images and 32.1% accuracy on false positives. By analyzing images and identifying affected spots, our proposed algorithm serves as a digital guardian for maize crops, aiding in timely intervention and management strategies thereby aiding in the preservation of maize yields and ensuring food security.