The approaching Qingming Festival heralds the harvest of a specialty product in Hangzhou: West Lake Longjing tea.
Every time this year, tea farmers in the capital of Zhejiang Province begin to pick tea leaves from terraced tea fields in droves, but this year things are a little different.
A tea-picking robot has been deployed to the tea gardens of Fan Village within the West Lake Scenic Area, one of the best tea-growing regions in Hangzhou, to help with the harvest.
The robotic tea picker is composed of solar panels, dual-lens cameras, mechanical arms, and other components. Every second it can pick a premium tea bud with leaves.
“We call it top-quality tea when it comes in the beautiful shape of one bud and one leaf or one bud and two leaves,” said Jia Jiangming, a member of the team behind the robot and associate professor at the School of Mechanical Engineering at Zhejiang Sci-Tech University.
Traditionally, the picking of premium tea relies on manual labor, with workers selecting in person.
But with increasing levels of mechanization in agriculture and ever-higher labor costs, the demand for mechanized tea picking, especially for top-quality tea, has risen in tandem.
Following a survey, the robot developers found that in the main production area for Longjing tea, most tea pickers are over 60, and even those over 50 are rare. Young pickers are few and far between.
“At this rate, this area may face the embarrassing situation of yielding tea but having no one to pick it,” said Jia.
Since 2019, his team has been developing tea-picking robots. The robot now deployed in the tea garden is a fifth-generation variant.
“Generally, the one-bud-and-one-leaf type of early spring tea is about 2 cm long, and the leaf stalk is only 3 to 5 mm long. The precision required for the operation of the robotic arm is extremely high,” said Jia. “If the leaf is cut incorrectly, it will damage the tea plant branches, cause damage or leave behind incomplete, unusable buds, resulting in losses,” Jia added.
To ensure the precision of tea picking, the team introduced a deep convolutional neural network recognition model, using AI’s deep learning capacity to process a large amount of tea tree bud and leaf image data.
This process enables the robot to recognize tea tree buds and leaves. At the same time, the dual-lens camera scans the tea leaves and achieves 3D positioning, accurately locating the bud and leaves.
After accurate positioning, the robotic arm will reach for the leaf stalk of the tea bud. The small scissors mounted on the end of the arm will remove the bud and leaves from the branch.
In the meantime, a negative pressure suction tube attached to the arm will suck the tea bud into a basket.
Yang Yajun, the chief scientist of the National Tea Industry Technology System, said the fundamental way out for agriculture lies in mechanization.
“The tea industry is the same. Reducing the burden of practitioners is the direction in which we should strive as technology workers,” Yang said.
“We hope that the tea industry can also enjoy the convenience brought about by mechanized operations and provide strong support for the industry’s growth,” said Chen Jianneng, a professor at the School of Mechanical Engineering at Zhejiang Sci-Tech University.