Deep Learning-based Crops and Weeds Classification in Smart Agriculture
Hassina Ait Issad, Rachida Aoudjit, Malika Belkadi, Mustapha Lalam and Mehammed Daoui
Today, agriculture remains the most important sector in the world. However, it faces a huge challenge: producing more and better with fewer resources while reducing the negative impact on the environment. Thus, to face this challenge, agriculture must become intelligent. In intelligent agriculture, the images collected from the monitored environment by various equipment’s play an important role. In fact, aerial images of drones, for example, can be a valuable source of information. High-quality and real-time images can be used to correctly recognize and classify crops to monitor their growth, to prevent diseases, weeds and pests that can damage them. By monitoring crops, by targeting areas to be treated and by accurately managing quantities, farmers are able to reduce input consumption (pesticides and water), resulting in higher yield, reduced costs and environmental impact. As a result, image processing has received a lot of attention because of its strong ability to extract information from images and develop decision support tools. Convolutional Neural Networks (CNNs) as a particular type of Deep Neural Network have gained popularity in recent years as a powerful means of classifying or categorizing images. In this paper, we propose a vision-based classification system for identifying weeds and crops using AlexNet and ResNet Convolutional Neural Networks. Evaluation and simulation results showed that the proposed crop and weed detection method is effective.