Artificial Intelligence to identify the ripeness and spoilage of cherries

Artificial Intelligence to identify the ripeness and spoilage of cherries

cherry selection
cherry selection

Without a doubt one of the most important factors to consider by cherry producers around the world is the condition of the fruit. This is a key factor in one of the most successful businesses in the agri-food industry in many countries such as Chile, Australia and the United States.

The international market interested in purchasing cherries not only seeks to satisfy parameters in terms of flavor, but also regarding color, maturity, physiology and appearance of the fruit, etc. This is precisely one of the biggest difficulties involved in the production of cherries, since the fruit is characterized by being especially delicate and sensitive to breakage and decomposition.

Technology as a great ally

In recent years, technology has allowed progress towards new methods that facilitate the care, selection and maintenance of fruit. The use of photographs with Artificial Intelligence, for example, is capable of providing a more detailed view of the fruit and reducing the selection of those that should not be selected.

A recent study made by specialists in electronic and optoelectronics technologies from the Xi’an Aeronautical Institute and Xi’an University of Technology in China proposes a new way of identification based on Swin Transformer, to extract feature information from the image of the cherry and then import that information into classifiers such as multilayer perceptron (MLP) and support vector machine (SVM). for classification.

As they explain, based on the results obtained, thanks to the comparison of multiple classifiers, the optimal MLP classifier is obtained, in combination with the Swin Transformer. Furthermore, performance comparisons are made with the original Swin-T method, traditional CNN models, and traditional CNN models combined with MLP.

The numbers speak for themselves. The proposed method based on Swin Transformer and MLP achieves an accuracy rate of 98.5%, which is 2.1% higher than the original Swin-T model and 1.0% higher than the best performance combination of the traditional CNN and MLP model. On the other hand, the training time required for Swin Transformer and MLP is only 78.43 s, significantly faster than other models. Excellent performance in identifying ripeness and spoilage of cherries.

The successful application of this method provides a new solution to determine cherry appearance, maturity and deterioration. Therefore, this method plays an important role in promoting the development of cherry sorting machines.

Help for the future

As we have seen in previous articles, one of the keys to exporting cherries is post-harvest care and packaging. Today, many continue to use manual methods to select which fruit goes into the boxes or not, leaving this responsibility solely to the human eye. However, evidence has shown that automated selection is much more efficient in terms of time, quality and resources.

Read the full study and conclusions in the following link


Noticias Relacionadas

12% more than in 2023……

Otras noticias