CCTV News: Mobile phone unlocking, face payment, nowadays, face recognition technology has been widely used, but have you heard of "insect face recognition"? In Anhui, there is such a group of young doctors, who hold "selfie sticks" and take "photos" of insects in the field for "insect face recognition". What kind of technology is this, and how can it help agricultural production?
A long "selfie stick" pokes into the wheat, deep or shallow, and moves quickly, like a "minesweeper". In Xuanzhou District, Xuancheng City, Anhui Province, Dr. Du Jianming of the Hefei Institute of Intelligent Machinery of the Chinese Academy of Sciences is busy taking photos of pests on the wheat with his colleagues in the wheat field.
Dr. Du Jianming, Hefei Institute of Intelligent Machinery, Chinese Academy of Sciences: Generally, aphids have a gathering in this position, and they look quite good now. Then if the whole pole goes down, it sweeps over like this, there are no pests, there seems to be no wheat root, and the whole stem is quite clean under it.
How does the "selfie stick" achieve "insect face recognition"? Du Jianming's "selfie stick" is equipped with a high-definition camera and a smart terminal, just extend the device probe into the wheat field and gently click the shooting button of the mobile phone to complete the picture collection.
Dr. Du Jianming, Hefei Institute of Intelligent Machinery, Chinese Academy of Sciences: The role of mobile phones is not only to collect photos, but also to upload the photos we take anytime, anywhere to the server for storage.
On the dedicated mobile phone software, the bugs taken by Du Jianming are automatically framed by the intelligent system. At the same time, the system analyzes the type, number and occurrence level of pests in the background. These pests and diseases have a huge impact on agricultural production. In order to "take food from the pests", crop pests and diseases need to be monitored regularly. However, the traditional calculation method is to count the land one by one, which is time-consuming and laborious, and cannot guarantee the accuracy of the data.
How to combine artificial intelligence and agriculture to make agriculture smarter? This is a question that Du Jianming and his scientific research team have been thinking about. However, insect face recognition can be different from face recognition. Some pests are so similar that it is difficult to distinguish and classify them using ordinary artificial intelligence techniques.
Dr. Du Jianming, Hefei Institute of Intelligent Machinery, Chinese Academy of Sciences: Bugs are generally very small, and what we really do is not only to identify whether it is an aphid, but also to know how many of these aphids there are, in order to determine the average number of pests and diseases in this field later, which is used to assess the disaster level of the disease here. On the whole, it is a level more difficult than face recognition.
In order to gradually break through the difficulty of "insect face identification", Du Jianming and his team traveled to almost every county and city in the province to collect data from field pests, continuously accumulate samples, and optimize databases. At present, the "insect face" database has collected more than 7 million pictures, more than 500 kinds of pest and disease "appearances", covering 29 kinds of crops and cash crops such as wheat, rice, rape, etc., with an identification accuracy of more than 80%. Huge databases keep pests nowhere to hide.
Hu Haiying, senior experimentalist, Hefei Institute of Intelligent Machinery, Chinese Academy of Sciences: We have a variety of pests, including some common and uncommon, which can also provide some assistance to experts, as well as a reference suggestion, that is, how to control, what kind of medicine or compound fertilizer to use, etc., from production to final harvest a series of management.
Source: CCTV