The market for AgTech, which combines agriculture and cutting-edge technology, is growing. The AI wind is blowing through the agricultural industry, and AI technology is being applied to various agricultural applications such as monitoring, quality and defect inspection, automatic harvesting, and smart farms.

Accordingly, we looked at the development trends of agricultural applications using vision AI technology at the 8th K-Farm Exhibition held recently at KINTEX in Ilsan.

■ Automation of agriculture, automation of harvesting with vision AI
Collaborative robots, which are widely adopted in industrial automation such as manufacturing, are now expanding their application to agricultural fields. In order to harvest non-standard agricultural products compared to products produced in manufacturing, it is essential to install vision AI using deep learning.

Gogofarm CEO Ko Won-seok, who unveiled the harvesting robot solution at the K-Farm site, put a lot of effort into the vision artificial intelligence that was developed in-house. CEO Koh first cited the lack of data as the most difficult part in developing vision awareness in the agricultural sector.

He said, “We trained the data of green, unripe tomatoes for one month targeting tomato varieties, and we need to accumulate at least 300,000 pieces of data per variety to see the effects of vision AI.” We are currently conducting research and development on automated harvesting of tomato varieties by securing 1,000 sheets of data per day at the verification center in Sangju, and we are aiming for commercialization of strawberry varieties by the first half of next year.

“Because we are collecting so much data, there are significant difficulties in data labeling,” said CEO Koh, adding that there is a need for an auto-labeling solution that can automatically classify various types and massive amounts of collected data.

■ Non-destructive vision inspection to catch defects, cutting-edge introduction trend in the poultry industry
Machine vision has traditionally been a solution used primarily in manufacturing industries such as electronics, parts, automobiles, and food packaging. Recently, as deep learning-based vision AI has developed and begun to be applied to various applications, the poultry industry is also introducing AI inspection systems to upgrade the quality inspection process for eggs.

Hanbit IoT is providing a solution that applies AI vision to the egg quality inspection system to identify whether eggs produced are defective and select defective eggs. △Normal △Blood spots △Cracks △Micro cracks △Potential cracks, etc. 7 types of defects can be identified and analyzed, and information that can help with the health status and production process of laying hens can be obtained.

Lee Chang-hyeok, a senior researcher at Hanbit IoT, said, “We are developing and providing a total solution, from sorting equipment to AI solutions,” and confidently stated, “With AI vision technology, we can determine defects in 38 different items, including detailed items in defect inspection.”

The researcher emphasized, “The existing sonic vibration egg sorting inspection system has disadvantages such as the risk of damaging eggs due to tapping and the occurrence of cross-contamination,” adding, “On the other hand, the vision AI inspection system has no risk of damage and can sort with a high accuracy of 90%.”