To conquer salt fields and vegetable fields with AI, scientific and technological innovation should also step on the "field"

For many people born in “Z generation ” As far as college students are concerned ,“ The field ” It's a very distant picture —— Appear in Textbooks 、 Movies and TV dramas, as well as the descriptions of the elders , It seems that there are few ways for them to understand agriculture .

In the past summer vacation , There are a group of college students who really walk down the field , Heyantian 、 vegetable field “ Intimate contact ”—— however , They do not simply experience the taste of work , It is ​ With artificial intelligence “ Secret book ”, Build a bridge between new technology and traditional agriculture .​

Flashed through my mind idea, To one line of code , And then to the real operation 、 Practical application of problem solving , This summer , Their story begins here ——

​ Building salt field data sets ：“ This has not been done yet ！”​

salt , The necessities of human existence , It is also an important industrial raw material .

at present , Most of the enclosed saltyards in China still adopt the traditional manual inspection method , Every worker should be responsible for the management of ten salt fields , The concentration of brine shall be tested personally 、 The amount of water and the crystallization of brine . Every morning, people still need to plough the fields , Prevent salt from hardening together ; If it rains , Cover the salt field with a covering cloth , Open it after the rain stops . Workers have been in the environment of salt fields for a long time , Skin is easily corroded and damaged .

Feng Huiyu , A junior majoring in communication engineering at Qingdao Agricultural University . He teamed up with his classmates to create “ Intelligent recognition and auxiliary diagnosis of encircling salt drying image ” This project , And the inspiration of the project , Is a member of the project team . This classmate's hometown is Changyi, Shandong , It is one of the important salt drying areas , I have been influenced by the salt drying operation of workers since childhood .“ Can we consider AI Help workers dry salt ？” As soon as the team member comes up with the idea , I got your response very soon .

Although the idea is beautiful , But it is difficult to operate , The first problem that comes up is data . at present ,AI The application of drying salt in salt field is still blank , Huiyu's team doesn't have any data sets that can be used as reference for model training .

“ Then collect it yourself ！” It's summer vacation , Huiyu and the two partners in the team together , I am determined to go to Yantian to collect the necessary picture materials . After communication , They contacted three salt drying farms in Changyi, Shandong Province , Soon the three set off for .

In a week in Changyi , Huiyu and his team members stay in the saltworks every day , From different lighting environments 、 Take photos of salt fields from different angles .7 The sun of the moon is particularly hot , Yantian's light reflection is so strong that you can't open your eyes , Three college students lamented that it was really not easy for salt drying workers , While sweating in the heat, I collected photos .

​​

​ Students go to the salt field to collect data ​

Final , Huiyu and the team members collected thousands of photos , Use the open source industrial level in-depth learning platform of Baidu to create a project of circle salt drying image recognition , And successfully built the relevant data set .“ This may be the first image data set of salt fields in different periods ！”

Different image classification models are used , Huiyu puts the data set into the model , The calculated force of the propeller platform is compared with the actual results , Finally, the model with high accuracy is selected and applied in the project . After establishing the model , Huiyu performed a series of actions such as optimization algorithm processing , Finally, the model of circle salt image recognition is derived .

In the process , Huiyu also learned by chance that the Baidu PaddlePaddle pilot group launched “AI Talent Creation Camp ” Activities , By “30 Sky fix AI The whole process of the project ” Deeply attracted . With the project being done , Hui Yu joins this activity , Soon I joined hands with the official engineer of the propeller , I also know more developers who pay attention to every corner of the society like him . Get the technical guidance from the official , And the help and guidance of like-minded fellow developers , Huiyu's project has taken another step forward . Final , Huiyu and his team have developed... Based on the propeller “ Intelligent recognition and auxiliary diagnosis of encircling salt drying image ” Application project , Through real-time observation of the amount of brine in the salt field 、 Height and crystallinity of salt , The accuracy is up to 70%. The processed data is transmitted to the management personnel , Assist manual judgment and analysis when to add water 、 Raking salt 、 Fish for salt 、 Guituo , Can effectively improve work efficiency , For the salt workers “ alleviate excessive burden ”.

“ I remember when I was doing picture collection in the saltworks , Some workers stood by and watched , It feels very strange , They will also doubt whether what we do will succeed , What can be done .” Regarding this , Huiyu and the workers explained that ： Using artificial intelligence , The salt drying field may be able to realize an integrated intelligent system like the current intelligent greenhouse , Every process has a machine to help control it accurately .

Maybe in the near future , Ride on “ Flying propeller ”, This group of college students can really create a salt field with AI The new future of .

​ Let winter rape “ No calcium deficiency ”  remote sensing +AI Create great value ！​

In recent years , High space 、 The development of remote sensing technology with time and spectral resolution , It provides a valuable opportunity to use remote sensing technology to monitor agricultural disasters and accelerate the development of intelligent agriculture . Every year, , In the key growth period of crops , Our country will develop wheat 、 Remote sensing monitoring of maize growth , In order to keep abreast of agricultural production .

For zhoulongyu, who majored in geographic information science at Huazhong Agricultural University , Using remote sensing images to monitor crops is not uncommon . But when the teacher gave him the collected winter rape data , And let him study how to use deep learning to do scene classification of winter rape , It still makes him a little overwhelmed . After a quick data query , Zhoulongyu ” lock “ Flying propeller , Learn to use now .

Rape is the first of the five major oil crops in China , It is an important source of edible oil and protein feed , At the same time, it is also widely used in industry , Its planting area accounts for... Of the total area of oil crops in China 40% above , The output accounts for% of the total national oil production 30% above , occupy the first place in the world . however , When rape is under nitrogen stress , In case of drought at seedling stage, the fertilizer will be removed , Red leaves are easy to appear ; Lack of nitrogen in the middle and late stage , Branches decrease , The number of pods and grains decreased sharply , Grain weight reduction , The output will decrease significantly . This also became zhoulongyu's participation in “AI Talent Creation Camp ” The main topic of .

In the course of progress , Zhoulongyu encountered many difficulties . The existing image recognition models are difficult to recognize remote sensing images , Zhoulongyu spent a lot of time selecting the appropriate model and modifying and optimizing , Let it adapt to remote sensing images . Later introduced PaddleX Kit , Zhoulongyu witnessed the smooth running of the model . In the next step , Zhoulongyu is still actively seeking opportunities from the school to go into the experimental field to collect more effective data , Make your professional ability and artificial intelligence more possible .

Deep learning 、 The Internet of things 、 Remote sensing and other technologies that many people think can only be applied in science and technology enterprises , Now it appears in the development process of digital agriculture , And this , It is just a microcosm of China's thousands of lines and industries moving towards intelligence . Today's college students are not ignorant youths living in the ivory tower , But is using advanced technology and knowledge , Understand the society with your steps 、 Measure the motherland , While improving yourself , And make greater contributions to social development .

Want to know more AI The story of developers ？ Welcome to scan the code to join AI Talent Creation Camp live broadcast group , Close communication with project authors , Share experiences .

​ Long press and hold the QR code below to immediately ​

​Star​

​

1. The official website of the flying oar ：

   www.paddlepaddle.org.cn/

2. Address of the project ：

   GitHub:

   github.com/PaddlePaddle/Paddle 
   Gitee:

   gitee.com/paddlepaddle/Paddle

3. Welcome to ask questions and exchange in the propeller Forum ~~

4. ​ ​

    http://discuss.paddlepaddle.org.cn

    ​​