A Metaheuristic Harris Hawks Optimization Algorithm for Weed Detection Using Drone Images

There are several major threats to crop production. As herbicide use has become overly reliant on weed control, herbicide-resistant weeds have evolved and pose an increasing threat the environment, food safety, human health. Convolutional neural networks (CNNs) demonstrated exceptional results in analysis of images for identification from that captured by drones. Manually designing such architectures is, however, error-prone time-consuming process. Natural-inspired optimization algorithms been widely used design optimize networks, since they can perform a blackbox process without explicitly formulating mathematical formulations or providing gradient information develop appropriate representations search paradigms solutions. Harris Hawk Optimization (HHO) developed recent years identify optimal near-optimal solutions difficult problems automatically, thus overcoming limitations judgment. A new automated architecture based DenseNet-121 DenseNet-201 models is presented this study, which called “DenseHHO”. novel CNN devised classify sprayer drones using algorithm selecting most parameters. Based proposed method capable detecting unstructured field environments with average accuracy 98.44% 97.91% DenseNet-201, highest among optimization-based weed-detection strategies.