AI-Based IoT Greenhouse Control System for Environmental Parameters

The environmental parameters of traditional greenhouse which need to be controlled continuously by farmers are not convenient and precise. In this study, an AI-based IoT greenhouse control system for environmental parameters was developed to solve the disadvantages of the traditional greenhouse system. IoT technology was used to help greenhouses connect to the internet that users could monitor greenhouses remotely and conveniently. AI technology was used to detect plant (leaf lettuce) growth and control the greenhouse environmental parameters (soil moisture, light intensity, humidity) automatically. In terms of plant detection, this study used a nonanchor-based object detection model, CenterNet, and anchor-based object detection, YOLOv3 model to detect the growth stage of leaf lettuce, and then recommend a suitable environmental parameter for the corresponding plants. In this study, the growth stage of lettuce was divided into four stages. The results show the mAPs of CenterNet were 1.0, 0.98, 1.0, and 1.0 from growth stage 1 to growth stage 4. The mAPs of YOLOv3 were 0.43, 0.75, 0.85, and 0.99 from growth stage 1 to growth stage 4. Therefore, the performance of CenterNet was better than YOLOv3. In terms of environmental parameters control, this study compared the normal control, typical PID algorithm, and fuzzy PID control algorithm to control current environmental parameters to the optimum environmental parameters. The results show fuzzy PID control algorithm had the lowest overshoot and highest robustness.