A healthy plant has a significant impact on the environment and human health. To put it another way, early discovery of plant diseases results in a momentous increase in the quality of the finished product. Using IoT-based agriculture, deep learning, and image processing techniques are essential to developing an effective way for detecting and controlling the spread of disease in agricultural products. Leaf drop is a common symptom of many diseases, as is the disappearance of the plant or plants nearby. This means that early detection and prevention are critical to ensuring healthy grape growth and yield. A method for detecting grape disease utilizing a CNN methodology and real-time data from environmental characteristics collected to predict illness likelihood. The image is subjected to multiple layers of CNN processing. The shuffled frog-leaping technique, a meta-heuristic algorithm, is used to optimize the CNN's learning rate (SFLA). The results suggest that the proposed SFLA-CNN outperforms existing approaches in terms of performance.
TopIntroduction
The internet of things aims to reduce human-to-human interaction by automating routine tasks. Data from sensors is collected by a controller, which then processes it and sends it out to actuators, which in turn complete the automation process (Sangeetha et al., 2022). IoT in agriculture aims to automate every aspect of the process, from seed to fork, in order to make it as efficient as possible. Traditional methods of yield management, livestock management, weed control, soil management, and water management, plant management, and animal tracking involve a lot of human touch and high labor costs and power consumption. There has been a lot of interest in agricultural production and efficiency in the development of intelligent systems that monitor and control a variety of field characteristics. Automated measuring systems collect data to produce precise readings, allowing them to make appropriate decisions (Babbar et al., 2022).
Soil moisture, humidity, temperature, and pH levels are some of the environmental characteristics now being tracked by smart farming systems. In machine learning, a new method of calculating intelligence is employed. Smart farming and agriculture also has a wide variety of ML applications. A few examples of how machine learning can be used in this industry include yield prediction, water management, and livestock management. There has been a lot of study done on IoT and ML for smart agriculture, but it has been discovered that there are a number of separate studies. ML and IoT are combined in this study to create better and more precise farming solutions (Popa, Prostean, and Popa, 2019; Medela et al., 2013; Mahdavinejad, 2018).
It is possible, however, that the rapid spread of undesired identifiers in plants may cause major damage to their leaves and fruits producing a loss in economic benefits (Barbedo, 2016). The appearance of leaves and petioles can be used to diagnose the majority of associated diseases, which are characterized by rots and abnormalities (McKinley and Talboys, 1979; Ebihara, Uematsu, & Nomiya, 2010). To prevent the spread of disease, the affected leaves must be removed or treated as soon as possible. Diagnosing plant diseases is therefore critical for both economic gain in agriculture and the development of high-quality crops.
When it comes to plant disease identification, previous literature has mostly concentrated on visual traits and disease characteristics (Harris, 1990). In recent years, plant disease diagnosis has been improved using machine learning approaches (Fujita et al., 2016). Although these methods are computationally expensive, they fail to detect all diseases. Some complex categorization issues can't be generalized by these procedures, which has a restriction. There was a substantial impact on agricultural production in recent years due to the quick development of deep learning techniques in the field. In addition, deep learning methods do not rely on specific features and are independent of real-world variables, such as changing lighting conditions, different-sized objects, and varying backgrounds. By training on a huge number of photos, they have the ability to detect and classify a wide range of objects.