Pests’ Attacks Prediction Using Sensor Fusion in Green Houses

Tomato is considered one of the most important crops not only because of its nutritional value but also due to its critical role in many fields of industry. Considering all of the crops’ losses’ factors like plants’ diseases or stresses of weather conditions, pests and insects are considered one of the major threats that attacks the tomato crops, causing a lot of losses. This work considers the use of sensor fusion techniques combined with convolutional neural networks (CNN) on field programmable logic gate arrays (FPGAs). The use of FPGAs facilitates the reconfiguration of suitable CNN, adapting to the environmental conditions based on the sensors fusion results. A case study has been devised, consisting of two blocks, first the sensor fusion block by using sensors’ integrated data from the soil pH, the temperature and the humidity level. The second block is the convolutional neural network block comprising both AlexNet and VGG16 models. On comparing the two models, the VGG16 proved to be more efficient in disease detection. Consequently, for the detection of the tomato mosaic virus disease, the yellow leaf curl tomato virus and the healthy tomato leaves, seven VGG16 neural network models were trained on different combinations of these three classes. By taking advantage of the sensor fusion, a higher accuracy for tomato diseases’ detection has been achieved meanwhile reducing the workload of the neural networks’ computations.