【2021 Solutions】 Fish farming is successful. Haisheng Technology uses AI to help fishermen raise good fish.
For Lien Weizheng, the founder and CEO of Haisheng Technology, who used to be an IC design engineer at MediaTek, he faces computer work all day long in Zhuke. Although he is busy, he has a sense of uncertainty and anxiety about his life goals. He couldn't help but think about whether the technological capabilities he had learned could be directly connected with people and land, solve various social issues, improve human life, and then change the world.
So, he resolutely left the enviable semiconductor industry and founded Haisheng Technology. He stayed at the fish farm in Qigu, Tainan, his hometown for a long time, and developed relationships with the fishermen. Finally, he was able to talk about "fishing secrets" and gain the trust of the fishermen. . The "Good Fish Aquaculture Monitoring and Breeding System" developed by Haisheng Technology by combining artificial intelligence, Internet of Things, organizational design and other technologies, uses a fish-friendly and environment-friendly method to reduce risks and increase yields, achieving sustainable farming production.
▲ Schematic diagram of Haisheng Technology’s AIoT aquaculture breeding management system for raising good fish
Wearing a sunhat, Lian Weizheng walked among the fish ponds under the scorching sun. When asked: Why did he switch from a technology person to an entrepreneur who helps fishermen raise fish? He replied with a shy smile: MediaTek's Chairman Cai Mingjie also asked him the same question. "I want to have more contact with people, directly understand people's needs, and find places where technology is needed to improve people's lives," he said, so he resolutely left the enviable semiconductor industry and became a "half-technical, half-fisherman" ”.
The annual output value of Taiwanese fishermen is low, but there is great potential to increase output value through technology
Why did you choose aquaculture fishery as the theme of entrepreneurship? Citing a report from "Tianxia Magazine", he pointed out that to show the value of the industry, we can observe whether the industry has prospects based on the per capita annual output value of each employee. For example, Qualcomm has an annual output value of NT$23 million per person, and Microsoft has an annual output value of NT$23 million. 22.5 million yuan, the annual output value of each Norwegian fishermen is as high as approximately 24 million yuan, which is three times the annual output value per person of Taiwan's electronics industry (8 million yuan). The current average of Taiwan's breeding industry is only about 400,000 yuan. From 400,000 to 24 million yuan, there is considerable room for growth and progress.
Thinking about it from another perspective, Taiwan’s electronic hardware and IoT equipment technology are unique in the world. If smart farming can be fully utilized with the help of these technologies, the average annual output value of fishermen is expected to increase significantly. The expertise of the Haisheng Technology team lies in artificial intelligence image processing technology and data analysis technology. They use these technologies to solve the risks and labor problems caused by the invisibility of fish and shrimp farming underwater, from being able to see, to quantity, and to control. Departure allows fishermen to quickly grasp the growth status of fish and shrimp using their mobile phones, and can also be combined with automated equipment to reduce labor and risks.
Lian Weizheng liked raising ornamental fish and watching birds when he was a student in the Department of Electrical Engineering of Tainan National Cheng Kung University. He often went to Qigu to watch birds. Through observation of the environment, he found that the fish ponds in Qigu were becoming more and more deserted. This is mainly affected by two major factors. Firstly, it is difficult for young people to enter the industry without experience; secondly, it is difficult for the older generation to pass on the inheritance without records. Therefore, the population of Taiwanese fishing villages is aging and the breeding output cannot be increased. The most effective way to solve the above problems is to import An automated system to assist farming.
The Haisheng Technology Fish Breeding System uses self-designed IoT underwater cameras to monitor farmed fish groups, and cooperates with Edge-AI image defogging technology to make the images of fish in the farmed ponds clearer, allowing fishermen to see the fish with their mobile phones. The images are synchronized to the cloud for group activities and feeding, determining whether they are sick, etc. Cloud-AI automatically identifies, samples, and measures biological information such as the length and weight of each fish, and automatically records and counts the growth status of the entire fish group for reporting and The visual presentation of images can also analyze the feed-to-meat ratio at different breeding stages. The AI expert system can also predict harvest dates based on growth data, control sales and shipment schedules, and ensure stable and sufficient supply.
▲A good fish raising system can save about 75% of manual work time
The system for raising good fish is like a ruler, taking into account every detail of fish farming
The system of raising good fish creates an objective ruler for the aquaculture industry. Using this ruler, fishermen can measure and compare the effects of different farming methods, see parts that they have not paid attention to before, and create more usage and value. In the past, traditional manual fishing sampling was labor-intensive and time-consuming. It usually took two people an hour to sample only 30-40 fish, and often caused fish casualties accidentally. Field verification results show that Haisheng Technology’s fish farming system takes 5 minutes. It can automatically sample and measure the biological information of 300 fish without disturbing the activities of the fish. It is extremely efficient and can ensure zero damage to the fish.
The farmers use systematic AI growth statistics as a basis for decision-making in determining the size of fish entering the pond and shipping. It replaces the original traditional method of manually fishing and measuring and recording 30 fish, which can save about 75% of manual work. Time, for the breeding industry where manpower is extremely scarce, the benefits of saving manpower are very significant. Haisheng Technology has also gradually expanded the AI technology it has developed into other fields. For example, Haisheng Technology participated in the AI+ Rising Talents Competition of the AI Program of the Industrial Bureau of the Ministry of Economic Affairs, "Problems Presented by Major Companies, Solved by New Innovations", and will use its exclusively developed fully automatic intelligent The underwater image defogging algorithm was integrated into the artificial intelligence model and applied to AU Optronics' solution for removing rainbow film from LED panel screens, and won the grand prize for meeting all technical thresholds.
▲Haisheng Technology’s fully automatic artificial intelligence underwater monitoring application can be extended to other industries
Lian Weizheng pointed out that the defogging technology can automatically estimate the degree of color distortion in water and automatically adjust the contrast of each area to improve image quality. Comparing AI models that consider underwater physics and methods based on adversarial network models, the method proposed by Haisheng is clearer in image details, and has also achieved results in multiple objective image quality evaluation standards such as NIQE and PSNR. Better scores.
Underwater image defogging algorithm is technically superb and favored by major manufacturers
This defogging technology combined with artificial intelligence technology can effectively improve the accuracy of fish identification. On groupers whose colors are similar to the background and difficult to distinguish, the accuracy can be significantly increased by more than 3%, even if the features are obvious. Moreover, the recognition rate of swallow fish is already as high as 98%, and the accuracy can be increased by nearly 1%. Using this technology to make the images of farmed fish clearer can be used as a basis for judging fish diseases. If the technology related to underwater defogging is built into the artificial intelligence model recognition layer, it can also be used to solve the problem of rainbow patterns appearing during screen shooting, causing panel defects to be undetectable due to shielding. This will effectively solve the problem of AU Optronics The problem of rainbow pattern interference was encountered during panel inspection.
As for the challenges of Haisheng Technology in business promotion? Lian Weizheng believes that the biggest difficulty is how to break into a new industry, because Haisheng often faces customers from traditional industries. When recommending or cooperating with traditional industry customers on a new technology product, how to gain the trust of customers? Achieving the goal of customers being partners is the key. We need to visit frequently and gradually build up a sense of trust before we can understand our needs and have the opportunity to further assist customers in solving their pain points.
Hisun Technology’s excellent image recognition and AI technologies have gradually gained recognition from customers and established empirical relationships with major manufacturers. In the short term, it will continue to gain a foothold in the market with its own image processing products such as the Fish Breeding System. In the medium term, Development hopes to expand the scope of products and business services, from underwater to onshore, from the production end to the consumer end to provide one-stop production and marketing integration services; in the long run, Haisheng Technology will copy various successful experiences in Taiwan and integrate partners resources to attack the international market.
▲The founder and CEO of Haisheng Technology, Lian Weizheng, works in Yubao
▲Lian Weizheng and Qiu Yushao, a couple, join the ranks of smart breeding with their scientific and technological expertise
「Translated content is generated by ChatGPT and is for reference only. Translation date：2024-05-19」
