Development of a Harvest Time Prediction Algorithm of Strawberry Using RGB Data-Based Ripening Level Decision
摘要
We intend to lay the foundation for the development of a robot that can harvest strawberries by devQueryeloping an algorithm that can measure and predict the maturity period using RGB data of the strawberry maturity process.
MethodRGB data was used to analyze the process of changing the color of strawberries to bright red by photographing 100 samples once a day in one of the strawberries’ harvesting periods, and the rate of change in length was analyzed by dividing bright red (ripe) and light green (unripe). Based on the obtained experimental RGB data, the area of discoloration of strawberries was calculated using a particle filter to derive the rate of change in ripe and unripe areas. The area ratio is calculated using the derived areas from particle filtering, and a linear function is derived for harvest time prediction with actual harvest data.
ResultsThe time that it takes to harvest strawberries depends on various factors, and there were many differences in the timing of maturity of strawberries grown under the same conditions in the same place during the same period. Harvest time took at least 7 days from the time it began to turn bright red, and it took up to 12 days. In addition, the harvest time of strawberries can be determined when the bright red ratio is about 80%, and it was confirmed that the length and area ratios were similar in value during this period.
ConclusionThe average target RGB values for tracking the ripe region of strawberry and area derivation were about 146, 74, and 41, and the unripe area was about 168, 163, and 95, respectively. It was found that it takes 7 to 12 days to harvest strawberries under the same condition, and most of them took about an average of 10 days. Based on the proposed prediction method, a reasonable accuracy of harvest time prediction was obtained, but more frequent data acquisition and large amount of data are needed to improve the accuracy of harvest time prediction because various factors such as ambient temperature and humidity can have an impact on harvest time.