How does a tea picking robot work

In recent years, the aging trend of agricultural labor force has significantly intensified, and the difficulty and high cost of recruitment have become bottlenecks restricting the development of the tea industry. The manual harvesting of high-quality tea accounts for about 60% of the total labor consumption in tea garden management. However, when picking the buds and leaves of high-end premium tea, the leaf tips are delicate, and the growth position, posture, and density vary. Especially in unstructured environments with gentle winds and changing lighting, machine harvesting is difficult to achieve. Therefore, researching intelligent tea picking technology is of great significance in promoting the development of the tea industry.

Tea recognition based on image processing

To achieve automated tea picking, the first step is to accurately identify the tender buds of tea trees. In recent years, with the development and application of computer technology, accurate identification of tea buds based on image processing has become a research hotspot.

1. Traditional image processing algorithms based on color space

Due to the significant color differences between tea buds, old leaves, and tree trunks, color features can be used to extract the bud regions in the image. Therefore, early research on tea bud segmentation was mostly based on color features. The traditional image processing algorithm based on color space mainly includes steps such as image preprocessing, color feature selection, and segmentation.

2. Recognition methods based on traditional machine learning

In order to further solve the problem of tea segmentation under natural conditions being easily affected by external environmental factors such as old leaves, branches, and soil, as well as the occlusion and overlap of tea leaves, machine learning methods were introduced in subsequent research. By extracting and synthesizing various feature sample data for training, recognition and detection were carried out. Common methods for identifying tender shoots are based on features such as color, texture, and shape, combined with the use of methods such as K-means clustering, support vector machine, Bayesian discrimination, and cascade classifiers. Traditional machine vision based recognition methods still rely on image preprocessing and data conversion, and improper preprocessing can seriously affect the accuracy of the model.

3. Recognition methods based on deep learning

Algorithms based on deep learning have high accuracy in complex backgrounds, providing a foundation for the research of intelligent picking equipment for tea buds in complex backgrounds. It can be divided into three categories: classification algorithms, object detection algorithms, and semantic segmentation algorithms. The classification algorithm based on deep learning is to classify one image, distinguish whether the image is a tender bud or recognize the state of tender buds in the image, such as the open surface state of buds and leaves, whether it is in a harvestable state, etc. This method has good recognition effect, not only accurately identifying tea tender buds, but also distinguishing the state of different tender buds. It can meet the requirements of tea tender bud recognition under natural light and has good practicality. However, deep learning based methods rely on large samples and have lower detection efficiency. Therefore, further research is needed on tea tree bud and leaf detection, increasing the number of bud and leaf images, and developing algorithms with faster speed, higher accuracy, and better stability.

End picking actuator

The picking object of tea is buds and leaves rather than fruits, and traditional end picking actuators are difficult to apply. Therefore, researchers have developed new end picking actuators for tea buds. In 2021, a clamp type end effector for picking tender tea leaves was designed, which can be controlled to achieve tea picking in tea gardens. The experimental results showed that the missed harvesting rate of one bud and one leaf was 2.8%, and the complete harvesting rate was 91%; The missed harvesting rate of one bud and two leaves is less than 3%, and the complete harvesting rate is 94%. Most of the existing tea picking end effectors use simple mechanical structures, which have little error compensation ability and cannot ensure the success rate of picking and the integrity rate of tender shoots. To solve this problem, a famous tea picking end effector based on negative pressure guidance has been designed. This type of end effector utilizes negative pressure to guide tea buds in a top-down manner, thereby correcting their posture and spatial position. The experimental results indicate that the designed end effector has deviation tolerance performance, which can improve the success rate of picking.

intelligence control system

The functions of the intelligent control system mainly include the control of the driving system and the control of the picking device. Japan has achieved some research results in the control of intelligent mechanical driving systems in tea gardens, such as the development of an “unmanned tea picking machine” by Matsumoto Corporation using artificial intelligence (AI) and sensors to move and harvest tea leaves without driving, which has begun to be sold. In terms of the control of the picking device, a machine vision based passenger intelligent tea picking machine was designed for the traditional reciprocating cutting harvesting device. An automatic recognition of tea buds and automatic leveling control method for the cutting blade were proposed, which can solve the drawbacks of existing tea picking machines that do not selectively cut old leaves and tender buds. At present, the number of end picking actuators controlled by the picking hand control system is single, and the picking efficiency is still not high. In the future, research and development will be needed for multiple end picking actuators and multi robotic arm collaborative control systems