Introduction
Mushrooms are very perish after harvest. The post-harvest quality is subject to many internal or external factors during storage or transportation. As a result, it is difficult to determine their post-harvest quality based on the shelf life.
Among all types of mushrooms, pleurotus eryngii has higher demands and economic value in the consumer market of Taiwan. That is the main reason why it is the focus of this project. Hyperspectral technology provides abundant spatial and spectral information than conventional RGB images. The collected information are able to reveal subtle features for detection and classification of substances. Artificial intelligence and machine learning models have been demonstrated to learn the spectral characteristics and to find characteristic bands related to targets.
Based on this concept, hyperspectral platforms established by our team have been used to collect a large number of hyperspectral images of Pleurotus eryngii. A handheld spectrometer was specially designed by this project based on the characteristic bands correlated to the post-harvest quality. Artificial intelligence model running on the cloud platform was also developed by this project to monitor post-harvest quality and to predict decay trend of Pleurotus eryngii in real time. By the end of this year the developed technology will be integrated into the information system of the mushroom packing manufacturers.
Result & Effect
1. Find the characteristic bands related to the post-harvest quality of Pleurotus eryngii
2. Develop a handheld spectrometer with the extracted characteristic bands
3. Establish an artificial intelligence model to automatically learn features and to analyze the data collected by the handheld spectrometer
4. Set up a cloud computing platform to run the artificial intelligence model and to receive the data collected by the handheld spectrometer
5. Integrate the handheld spectrometer and the artificial intelligent model running on the cloud platform with the platforms of the mushroom packing manufacturers.
6. Optimize the collection process and platform to achieve online real-time detection
Contact
