Golden papayas (Carica papaya L.) cannot be mechanically harvested, although fruit growers have always wanted to find a way. Among the reasons for the difficulty of designing a reliable automated method is the need to identify fruit at different stages of ripening, from mature to almost mature to unripe, with several intermediate variations.
In search of a solution to this problem, a group of researchers at the University of Campinas’s School of Food Engineering (FEA-UNICAMP) in São Paulo State, Brazil, have investigated the use of algorithms and computer vision to analyze agricultural products.
“The idea is to automate harvesting using non-invasive imaging technologies such as computer-aided analysis of digital images in visible light as well as infrared,” said Douglas Fernandes Barbin, principal investigator for a research project on the subject with FAPESP’s support. Findings from the study were published in the February issue of the journal Computers and Electronics in Agriculture.
Common sense says a papaya is ripe when the peel is yellow, whereas a green peel is a sign the fruit is far from sufficiently mature. This is often correct, but there are many exceptions.
“Peel color usually points to the stage of fruit maturation in the case of papayas, but this criterion isn’t always reliable, and there’s a great deal of variation,” Barbin said. “Yellow peel doesn’t always mean the fruit is ripe in the sense that its compounds have been completely converted into sugar, so that it’s sweet and soft to the palate. Sometimes the peel is partly yellow and partly green, and it can be hard to decide whether a variegated papaya is ready for picking.”
The researchers at FEA-UNICAMP, in partnership with a group at Londrina State University (UEL) in Paraná, led by Sylvio Barbon, Jr., investigated the use of a portable sensor that illuminated and analyzed the fruit on the tree. The device emitted a luminous signal that was reflected by the fruit and captured for spectral analysis.
The electromagnetic spectrum comprises all forms of electromagnetic radiation, which can be highly energetic and dangerous (e.g., gamma rays and X-rays) or inoffensive (e.g., radio, TV and cell phone waves).
Visible light, the form human eyes have adapted to recognize, occupies a small portion of the electromagnetic spectrum. Infrared radiation is invisible to humans but can be seen by many animals, including several bird species.
“In the case of papayas, the infrared portion of the spectrum supplies important information about the biochemical stage of fruit maturation,” Barbin explained. This information can be used to supplement the data obtained from digital images, further enhancing the precision of the method.
In the study, golden papayas were purchased in a retail market in the city of Campinas and then measured and weighed. Peel color was determined using a colorimeter. The researchers also analyzed physicochemical properties such as pulp firmness, pH, soluble solids, total carotenoids, and ascorbic acid content.
The samples were classified into three maturity stages (MSs) according to pulp firmness, using a texture analyzer to measure compression force (N).
Papayas with the firmest flesh (higher than 33 N) were classified as MS1, while those with pulp firmness between 33 N and 20 N were classified as MS2, and those with less than 20 N – soft and ready for consumption – were classified as MS3.
“The device used to acquire data on fruit characteristics was a compact digital imaging camera positioned in a structure designed to provide suitable lighting and optimal representation of the fruit’s surface,” Barbin explained.
Two color images of each papaya were recorded, one of each side. The images were digitally processed to separate three color channels (red, green and blue) and also for evaluation in terms of hue, saturation, lightness, and feature extraction.
“In the digital images for each sample, we explored several color spaces to represent it with the aim of obtaining a better performance during fruit classification,” Barbin said. “The color spaces that we studied included RGB, HSV, CIELAB and their mean values on the surfaces analyzed as descriptive attributes.”
Barbin explained that these descriptive features were fed into a random-forest decision tree algorithm to model the fruit ripeness classification system.
Computational processing followed by statistical analysis of all properties and features produced indicators of pulp firmness for each papaya.
“We compared these predictions with the previous mechanically obtained values and found an accuracy rate of up to 94.7%,” Barbin said.
Brazilian growers are keen on mechanizing the papaya harvest, he added. They stand to gain many benefits. In the irrigated plantations of the São Francisco River Valley in the Northeast region, for example, it would be possible to identify and pick papayas so they would reach the ideal stage of ripeness only some days later, upon arriving at markets in the Southeast (Rio de Janeiro and São Paulo).
“Less-ripe fruit could be selected for export, and mature papayas could be shipped to markets closer to the farm, such as cities in the Northeast,” he said.
Next steps for the researchers include adapting the sensor to a portable manual device that can be used in the plantation to point the luminous signal directly at papayas on trees.
Source : By Peter Moon | Agência FAPESP