The application of advanced optical multispectral control technology, color correction system and AI Artificial Intelligence deep learning system in automatic identification of changes in physiological indicators of ripening of vegetables and fruits is highlighted as follows:
  1. Establish the database of ripeness, color, physiological characteristics and antioxidant capacity of standardized agricultural products.
  2. The non-invasive optical system was used to detect the color value of fruits and vegetables. The changes of internal characteristics of fruits and vegetables were detected and the physiological indexes such as maturity, ethylene production rate, respiration rate, fruit firmness, starch content, total soluble sugar content and antioxidant composition were mastered in time.
  3. The convolution neural network based on deep learning is applied to image recognition and analysis. The convolution neural network of long short-term memory (LSTM) is an "end-to-end" process of two stream three-dimensional convolution neural network, which can automatically identify the image content of any size and length, so as to improve the image recognition performance of fruit maturity. Combining the spatial and temporal information of fruit image sequence, we can get the best fruit maturity discrimination and classification. The neural network based on deep learning can achieve both classification and instance segmentation. At the beginning of the study, we developed algorithms with specific fruits and vegetables, and then expanded them to other fruits and vegetables with transfer learning. In the future, AI will be used to integrate a large number of physiological data and RGB data of fruits and vegetables, so as to more accurately estimate the real-time physical, chemical and physiological characteristics of fruits and vegetables.