At a time when the sustainability of animal diets is gaining attention, introducing insects as a substitute feed option addresses the expanding need for feed. The authorization of insects as livestock feed in Europe opens up a profitable market, highlighting the chance and chance to craft effective and automated quality control techniques for these feed types. Insects are an important source of protein, especially given the growing world population and the associated increase in food production. Hyperspectral imaging emerges as a pivotal technology, offering the capability for early detection of diseases, accurate identification of growth phases, and ideal harvesting periods.
Pupal stages – Analysis of spectral information
For the identification and differentiation of the relevant pupal stages, experts defined three distinct pupal stages (emergence stage, middle stage, initial stage) and selected representative pupae for each group. Measurements for these pupal stages were conducted using the ULTRIS X20. The spectral information collected from multiple measurements demonstrates that the individual pupal stages exhibit different spectral behaviors, allowing them to be fundamentally identified and distinguished. The course of the averaged values with the standard deviation across wavelengths is depicted. The standard deviation represents the variability within the individual pupal stages due to the full-area capture with the hyperspectral camera.
Pupal Stages – Classification of spectral measurements
The application of hyperspectral imaging techniques for classifying the three pupal stages of mealworms has achieved significant success. For this purpose, various pupal stages were measured under defined and uniform conditions in the laboratory, and based on these data, applied to an independent dataset that contains the three pupal stages. Through the analysis of spectral signatures using machine learning methods, each defined stage could be identified. The mixed stages for the individual pupae arise because these animals are intermediate stages of the three groups, and thus certain areas can be assigned to one class or another.
Harvest stages of mealworms – Analysis of spectral information
The spectral signature of frozen mealworms shows the most significant differences when compared to pre-harvest and ready-to-harvest stages, particularly in the range of 600 to 800 nm. The distinction between ready-to-harvest and pre-harvest stages is less pronounced, yet it is most notable within this specific wavelength range. This analysis of spectral information highlights the potential of using spectral signatures to accurately identify the optimal harvesting time and to ensure the quality of mealworms for consumption or processing.
Mealworms – Classification of spectral measurements
For the extraction of spectral information, areas were selected in the measurements with the respective mealworm stages (pre-harvest, ready-to-harvest, frozen) where these stages are likely to be represented with high probability. The selection was based on expert knowledge, allowing for the capture of representative spectral information and the underlying variability within the stages. These measurements were used as the basis for training the machine learning algorithms. Further measurements with the individual stages were labeled (pre-harvest = blue, ready-to-harvest = green, frozen = yellow) and used for evaluation. The evaluation itself shows red areas with incorrectly assigned stages and green areas where the stages were correctly assigned. The identification of the three stages was successfully performed, as the majority of the labeled areas (97.4%) were marked as correctly assigned stages (green areas).
Mealworms – Pre-harvest, ready-to-harvest, frozen (from left to right); Colored Infrared images (upper row) and classification performance (lower row) with incorrect classified (red) and correct classified (green) areas of the three stages.
What is crawling into the future?
Overall, it has been demonstrated that the hyperspectral camera ULTRIS X20 can successfully distinguish parameters relevant to the insect industry. This camera enables non-invasive, rapid, and reliable detection of defined pupal stages and identifying animals ready for harvest. Thus, the first step towards optimizing breeding processes and enhancing production efficiency has been achieved. The findings underscore the potential of hyperspectral measurements in mealworm farming as an alternative food source. Further research is needed on early-stage disease detection or varying quality due to different feed sources for the insects. With advancing technology, future investigations could be conducted using ULTRIS SWIR, which could reveal new insights regarding relevant parameters, as its wavelength range provides information about deeper layers of the insect body.
About the Author
Dr. Viktoriya Tsyganskaya is the Head of Project Management at Cubert GmbH and has been leading research and customer projects since 2018. She earned her PhD in Remote Sensing from Ludwig Maximilian University of Munich, specializing in radar remote sensing and environmental monitoring. Viktoriya has extensive experience from her scientific work, including the project “Dikes under Pressure,” and expertise in sustainable environmental solutions. Her deep knowledge in remote sensing makes her a key contact for innovative hyperspectral technologies at Cubert.
