Surveillance Approaches to Intrusion Detection, Crop Health, and Disease Prevention in Agriculture
Nguyen Bang Nong
Vietnam Academy of Social Sciences
Mohammad Tarek
Department of Computer Science and Engineering, Rangamati Science and Technology University
Keywords: Agriculture challenges, Machine learning, Computer vision, Crop monitoring, Intrusion detection
Abstract
Agriculture is facing numerous challenges in the 21st century, including crop damage from intrusions, water stress, weed infestation, and disease outbreaks. This study examines the integration of machine learning (ML) and computer vision technologies in addressing these issues, thereby enhancing agricultural productivity and sustainability. Closed-circuit television (CCTV) surveillance systems and drones equipped with computer vision algorithms are employed for intrusion detection, identifying irregularities in large agricultural areas and sending alerts to farmers. This system effectively mitigates crop and livestock damage caused by both domestic and wild animals. The study further delves into the application of computer vision technology in analyzing drone footage for crop growth monitoring and stress detection. High-definition images and sensor data serve as inputs for ML algorithms, enabling the identification, classification, quantification, and prediction of stress factors, such as water stress. Regression and ensemble-based techniques are used to predict leaf water content (LWC), a crucial measure of plant productivity and yield, which is then used to classify water stress levels. Weed management is another critical area addressed in this study. Computer vision systems are used to distinguish weeds from crops in drone footage, enabling real-time weed elimination using lasers and sprays. This approach is crucial in preventing the substantial yield loss caused by weed infestation. Lastly, the study investigates the use of CCTV and drone footage in livestock and crop health monitoring. Notably, multispectral images generated by scanning crops with RGB and near-infrared light are used to identify and treat infected plants promptly. Furthermore, deep learning models like EfficientNet are trained to classify and detect crop/plant diseases, with synthetic data points generated by Deep Convolutional Generative Adversarial Networks (DC-GANs) used to mitigate the challenge of obtaining labeled images of plants/crops. This approach to agricultural management using ML and computer vision technologies presents a promising avenue for improving agricultural productivity and sustainability in the face of increasing challenges.
Author Biographies
Nguyen Bang Nong, Vietnam Academy of Social Sciences
Mohammad Tarek, Department of Computer Science and Engineering, Rangamati Science and Technology University
