Germany: Imaging Technology How AI Monitors the Health of Cannabis Plants
Seed and seedling quality, germination features, and phenotypes of hemp plants are important in research, breeding, product development, and quality control during production. Imaging and image analysis gives invaluable insights into the relevant traits. Lemnatec provide tools that assist in these tasks.
Aachen/Germany — Hemp plants, or Cannabis sativa, are of high interest for both, biobased materials like hemp fibers, and for Cannabis active ingredients in medical use. Producing a medical product requires exceptional quality control mechanisms. Imaging analysis combined with powerful AI technology can help make sure that the plants are healthy.
Lemnatec's Seed AIxpert technology can image hemp/cannabis seeds during germination and seedlings during emergence. Specialized imaging solutions comprising high resolution cameras and customized LED illumination enable high quality imaging of the samples that make subsequent analyses easy. The image processing software can detect germination and measure the dimensions of the emerged seedlings, e.g., root length, or shoot length, as soon as present. Recognizing seed and seedling features, as well as damages or pathogen infestations can be done using our machine learning algorithms. The following learning process ensures that the image analysis workflow recognizes the target features with high accuracy and in a repeatable way.
For soil-germinated seeds, the image processing software can detect germination via seedling emergence, and again the shoots that emerge from the soil can be measured. A range of shoot-related features can be assessed, and quality ratings are possible for the emerging seedlings.
While seedling emergence is closely linked to germination and thus is task of the Seed AIxpert, emerged plants grow and there is a demand to obtain phenotypic information on the growing plants. Such phenotypic data help to understand genetics, to optimize cultivation procedures and treatments, or do carry out plant health studies. Similarly, growth performance data deliver information on the quality of the plant material already during the cultivation process.
Comprehensive plant phenotyping is done by the Pheno AIxpert-family. For example, top view and side view photographs of growing hemp plants can serve to characterize growth performance and phenotypic traits of the plants. This comprises height or width of the plants, projected area, morphological factors (e.g. compactness) or color information such as greenness or stress-induced color changes.
Beyond RGB imaging, the company can apply further imaging methods for phenotyping, including hyperspectral imaging, thermal imaging, and chlorophyll fluorescence imaging. Non-RGB images carry information on physiological features of the plants, such as nutrition status, photosynthetic parameters, or stress responses. The available software package comprises analytical functions that convert optical signals into plant-related information for all types of images used in the phenotyping systems. As with seeds, machine learning algorithms can be used to establish sample-specific analytical workflows. In particular, spectral imaging can provide physiological data with spectral signatures of the plant material relating to biochemical properties via AI-based analyses. Such technology can be used to non-invasively assess active ingredients in growing medical plants.