The greenhouse nowadays is a specialized high tech production system for a wide variety of crops. Due to its large scale, it is impossible for a grower to monitor and control all plants individually. This, combined with other factors like an inhomogeneous climate, complex IPM in which the biological balance is vulnerable and where no individual plant information is available, leads to sub-optimal plant conditions.
In this project, the goal of BU Greenhouse Horticulture of Wageningen University & Research, is to give the grower plant scale information and support with a decision support system (DSS) for local measures to get a more sustainable crop production. The focus of the project is the detection of pests and diseases in a Gerbera crop. Detection of plant load, defined as all buds and flowers in the crop, and detection of powdery mildew. The sensor system for detection will consist of different camera types, which will be mounted on a measuring cart which can drive autonomously in the greenhouse.
Monitoring plant load
The plant load is a parameter which growers use a lot to determine crop status and expected yield. The plant load is monitored using a camera giving images from top view of the crop. In these images, all buds and flowers are detected using a deep learning algorithm. The number of buds and flowers in the greenhouse will then be shown in a heat map, indicating weak and good performing spots in the greenhouse.
Detecting powdery mildew
Powdery mildew in the crop will be detected using a hyperspectral camera. The camera will take images with a high spectral resolution of the crop. Images of all the different wavelengths contain the information to detect the powdery mildew. The detection algorithm is still under development, but will be introduced to the measurement cart in short term. For the presence of powdery mildew in the greenhouse, a heat map will be generated as well.