How are “computer vision recognition models” and “lettuce” connected?
The LettUs Grow team, a participant in the 3rd Duoduo Agricultural Technology Competition, provided a solution: using computer vision recognition models to identify and calculate the projection area of lettuce, determining crop status and density change times and values based on leaf occlusion; then predicting plant growth based on plant models, guiding harvest times and yields, arranging planting schedules, and flexibly adjusting production rates according to market demand…
Food is essential to the people. Under the trend of digital transformation, industrial industries are developing rapidly, and modern agriculture is no longer just a task in the agricultural field. Nowadays, the development of modern agriculture involves expertise from multiple disciplines, including computer science, software engineering, engineering, materials science, and mechanical manufacturing and automation. The report from the 20th National Congress of the Communist Party of China pointed out the need to establish a big food perspective, develop facility agriculture, and build a diversified food supply system. Therefore, how to promote the transformation and application of agricultural technology achievements through interdisciplinary technology applications and to promote the high-quality sustainable development of the rural economy has become one of the key development directions in the agricultural industry today.
In this context, Pinduoduo has held the “Duoduo Agricultural Technology Competition” for three consecutive years. Wang Jian, Senior Vice President of Pinduoduo, stated: “We hope to promote technological exchanges, innovation, and development in the agricultural field through the competition, attracting more young agricultural innovation talents and outstanding modern agricultural enterprises to participate in the technological innovation of agriculture, promoting the upgrading and transformation of agricultural technology, and pushing for the high-quality development of agricultural technology in our country.”
So, can technology inject new vitality into traditional planting?
“Cui Tian” is a variety of fruit lettuce named for its crisp sound when chewed. The 3rd Duoduo Agricultural Technology Competition required participating teams to use technological means to grow the new lettuce variety “Cui Tian” in a fully enclosed container with lower energy consumption while achieving better quality and higher yields. The participating teams planted three crops during the competition, with the final crop’s planting results counted towards the finals.
It is reported that this year’s competition used container-style plant factories as the planting scenario, which can maximally block the influence of the external environment. After fierce competition, the Shanghai Academy of Agricultural Sciences team, CyberFamer team, “Endless Life” team, and LettUs Grow team stood out from 15 initial teams worldwide and advanced to the finals.
The “lettuce weight prediction model” mentioned at the beginning of the article is the proposal put forward by the LettUs Grow team. It is reported that their team members include lettuce cultivation experts, lettuce growth model experts, data scientists, plant physiology experts, agricultural business model experts, and AI algorithm engineers, coming from institutions such as Wageningen University and Research in the Netherlands, Beijing Polaris Agriculture Co., Ltd., and Bayer Crop Science. They developed hardware alarm systems, lettuce weight prediction models, etc., in the competition, which can help improve the yield and quality of lettuce in plant factories.
The Shanghai Academy of Agricultural Sciences team provided a more systematic problem-solving approach. They built a relatively mature smart planting decision management system, based on the database construction of a big data cloud platform, integrating algorithms such as plant growth models, light interception models, and transpiration models to collect data returned from various sensors in real-time for planting strategy analysis and decision-making.
“The difficulties we encountered include unfamiliarity with variety characteristics, unfamiliarity with container equipment, uneven LED lighting, and imprecise irrigation control systems,” said He Lizhong, the captain of the Shanghai Academy of Agricultural Sciences team. “We summarized some planting techniques based on experience and developed a decision management system through interdisciplinary teamwork and integration. Based on the precise data acquisition and analysis capabilities of the data platform, the Shanghai Academy of Agricultural Sciences team harvested lettuce first and won the championship of this competition, also receiving the ‘Highest Yield Award.'”
In fact, looking at the backgrounds of the contestants, both the CyberFamer team and the Shanghai Academy of Agricultural Sciences team are agricultural teams among the four finalist teams. They are accustomed to starting from the needs and rules of crop cultivation, combining traditional planting experience with artificial intelligence to achieve better planting results. The CyberFamer team and the “Endless Life” team are engineering teams that take lettuce planting as an application scenario, solving practical agricultural production problems with cutting-edge technology.
According to Zheng Jianfeng, captain of the CyberFamer team, they took the growth rate of lettuce as one of the important bases for adjusting environmental parameters inside the container, so the monitoring method of lettuce growth rate is crucial. To find a simple, feasible, and reliable monitoring method, they used cameras fixed above the lettuce to capture canopy images and designed a single lettuce canopy area recognition algorithm in a community state, establishing a relationship model between canopy area and fresh weight, ultimately achieving biomass monitoring of lettuce based on machine vision recognition.
The intelligent environmental control method based on crop physiological feedback and the dynamic monitoring technology of lettuce net assimilation based on CO2 mass balance, as well as the lettuce canopy area recognition model developed by the CyberFamer team during the competition, allowed them to maintain a high growth rate of lettuce while ensuring the quality of the lettuce. It can be seen that using machine vision recognition technology to grow lettuce and continuously optimizing the model can save labor and improve the accuracy of monitoring data.
The competition between two agricultural teams and two engineering teams was regarded by the judges of the competition, Professor He Dongxian from China Agricultural University, as the biggest highlight of the finals.
Since the competition only counts the planting results of the final crop towards the final score, the planting data of the final crop is crucial. He Lizhong, captain of the Shanghai Academy of Agricultural Sciences team, stated: “Through the first two crops (of lettuce planting), we set up different density experiments, different light quality ratio experiments, different light cycle experiments, growth detection, energy consumption record simulation, nutrient solution formula control, and environmental temperature and humidity regulation. The team members became familiar with the characteristics of the competition variety, mastered and adapted to the use of the container, improved our competition strategy, and controlled problems such as burning and green algae, optimizing the balance between plant growth and energy consumption.”
It is reported that the production efficiency of the last crop of lettuce from the Shanghai Academy of Agricultural Sciences team reached 0.18 kg/m2 / day. “The use of container-style plant factories this time is not production-oriented; only three layers of shelves were installed. If the number of planting shelves is increased to improve space utilization efficiency, based on our planting plan, the production efficiency could completely reach the international advanced level of 0.4 kg/m2 / day,” He Lizhong stated. In terms of quality, the soluble sugar content of the lettuce sent for inspection by this team reached 0.43%, the highest among the four teams.
The engineering team, the “Endless Life” team from Shanghai Jiao Tong University, significantly increased the yield of lettuce from a technical perspective. They constructed the “Endless Life” information platform, which visualizes and remotely collects over 130,000 sets of data, providing “infrastructure” for smart cultivation and solving the limited understanding of new varieties, limited data, and limited expert experience to achieve yield improvement. It is reported that after three rounds of iteration and optimization, the biomass of the “Endless Life” team in the third crop grew by 86% compared to the second crop and by 135% compared to the first crop. This also made many expert judges exclaim, “If more crops were grown, the champion might have been the engineers.”
Captain of the “Endless Life” team, Bao Hua, said: “We spent a lot of effort on hardware adaptation in the early stages, and team members went to the site multiple times for debugging and actively discussed online with engineers from Zhongke Sanan on how to solve problems. We developed the ‘Endless Life Data Platform’ ourselves, achieving automatic data import and subsequent model analysis, and ultimately we basically realized the functions we wanted.”
They are energy-saving oriented, integrating environmental control algorithms to regulate temperature, humidity, and CO2 concentration in indoor planting; by monitoring environmental parameters through the data platform, they preliminarily proved the feasibility and effectiveness of energy-saving-oriented environmental control algorithms in indoor planting scenarios.
There can only be one champion, but both the agricultural teams and engineering teams performed remarkably. “All four teams closely integrated digital technology with agriculture,” said Zhao Chunjiang, an academician of the Chinese Academy of Engineering and representative of the competition judges. “Digital technology is always a tool and method. To solve agricultural problems, we must thoroughly understand agriculture; only then can the combination of engineering and agriculture achieve better results.”
From slash-and-burn agriculture, stone tools, to drone seeding and mechanized harvesting, modern agriculture has developed based on modern industry and modern scientific technology. The evolution of agricultural tools is an important driving force for the development of modern agriculture. From the “Duoduo Agricultural Technology Competition”, we can see that in the AI era, technologies such as cloud computing and big data have become new driving forces for the development of the agricultural industry.
The winning team of the first “Duoduo Agricultural Technology Competition”, “Smart Berry”, saw the market prospects for productizing technology during the competition and established the “Smart Berry” company to help small and medium-sized growers improve efficiency. It is reported that in the key assistance area for rural revitalization, Laowo Village in Nujiang Prefecture, Yunnan Province, “Smart Berry” built a digital strawberry production system for the local area, resulting in a more than 30% reduction in labor costs for the strawberry industry in Laowo Village, including a reduction of 2,500 yuan per mu in fertilizer expenditure and a reduction of 1,000 yuan per mu in pest control expenditure, while strawberry yield increased by 30%. Currently, “Smart Berry” has formed hardware, software, and algorithm products such as intelligent irrigation and intelligent greenhouse environmental control, and by the first quarter of this year, it has output 40 sets of systems in Liaoning, Yunnan, Anhui, Inner Mongolia, Shanghai, and Beijing for assisting strawberry and blueberry production.
Zheng Jianfeng, captain of the CyberFamer team, believes that if their planting scheme is applied to large plant factories, the electricity consumption per kilogram of lettuce can be reduced to 9.5 degrees, “better than the industry high level of 10 degrees of electricity per kilogram of lettuce.” Currently, the dynamic adjustment technology of nutrient solution formulas accumulated by the CyberFamer team during the competition has been written into a popular science paper, and all data has been embedded into the model, forming a standard algorithm that is being applied at the Xiaotangshan base in Beijing.
Wang Jian, Secretary of the Party Committee and Senior Vice President of Pinduoduo, stated: “The final teams showed us the direction and path for the transformation from sweat agriculture to smart agriculture. Pinduoduo will continue to hold innovative competitions, encouraging everyone to write their papers on the land and leave their results in the farmers’ homes.”
Making good use of the “new farming tools” of the AI era, the 3rd Duoduo Agricultural Technology Competition is just the beginning.