Greenhouse Monitoring System and Tomato Leaf Disease Classification Using Convolutional Neural Network

The greenhouse automation system helps the workers to maintain and monitor the greenhouse instead of doing manually which leads to human errors, human fatigue, and high expense due to the large-scale variety of plants grown in the greenhouse and labour expenses. Plants in greenhouses can be affected by diseases which reduces income as well as yield. These diseases can be detected using automated systems. In this paper, automation, alert, and monitoring system is designed. The automation system includes sensing of temperature, humidity, soil moisture, and water level. The alert system includes fire alerts and water alerts. When the water level in the tank falls below a threshold point, or in case of an emergency like a fire outburst, the user will be indicated immediately through a text message in his/her mobile phone which is the function of the alert system. The monitoring system includes taking images and uploading them periodically to the cloud as well as identifying and classifying the diseases in leaves. This is done by using sensors and actuators in the Internet of Things (IoT) like Raspberry Pi, DHT22, soil moisture, and water level sensors. Raspberry Pi is programmed to control the actuators accordingly which include fans, heating bulbs, humidifiers and the water pump. Identification and classification of diseases are done by using convolutional neural networks (CNN) architectures like VGG 19, MobileNet V2, ResNet 152 V2, DenseNet 201, and Inception V3 evaluated on the plant village dataset. As a result, this model has achieved a maximum accuracy of 96.53% for DenseNet 201 Therefore, without direct supervision, the workers can monitor and manage the greenhouse by using this automated technology.