Sichuan farmers break the traditional wheat testing system, sing a new chapter in Zhennong, Xingxing Township

In order to achieve the goal of "Xingnong Reporting the Country" in summer social practice, the "Wheat Overwatch" team of the School of Information Engineering launched a social practice activity with the theme of strong farmers and farmers from July 1, 2022 to August 10, 2022. The team uses wheat as the starting point, using the method of random sampling of wheat ears intelligent counting and combined with remote sensing image analysis to monitor wheat growth and insect pests, and is committed to improving wheat output. Practice has been carried out in various forms, including field surveys entering the local wheat planting base in Ya'an, online inspection related materials, and the production of intelligent wheat monitoring systems. At present, the practical activities have successfully ended.

Objective: Break the traditional monitoring methods to escort the income guarantee of wheat

Using drones for low -altitude and high -altitude information collection, low -altitude collection mainly collects the details of farmland details, and collects the main collection of macro -information information at high altitude. In the later period, the team uses a semantic segmentation algorithm to accurately divide the crops and abnormal vacancies by using the semantic segmentation algorithm. For low -altitude images to further analyze the details of the crops, it mainly uses the target detection algorithm of wheat in the later stage In the end, enter the model to calculate the production of crop, so as to achieve the ultimate goal of predicting wheat output.

The picture shows the drone shooting wheat pictures

Test: Precision monitoring the invasion crisis of the environmental test in the environment

After doing a good job of epidemic prevention measures, the team entered the Ya'an wheat planting base for field surveys to get real -time data. After that, the team members conducted procedures for the network platform detection system, and the team used the SPNET algorithm to solve the problem of testing the vacancy of wheat farmland.在主干网络层，为了防止因为直连通道上relu函数引起的数据损失，模型消除了直连通道上的ReLU操作，将信号传输分为三个阶段使卷积归一化以不同顺序执行，保证The overall normalization and activation of the signal. In the ENCODER layer, based on the characteristics of most wheat field cracks and vacancies, the model adopts the long distance distribution of capturing the isolated area along the space dimension. Build remote dependence, introduce the attention mechanism, and extract the effective characteristics of wheat fields to achieve better semantic segmentation effects.

The team compiles the network platform detection system

In the field of target testing today, high-precision models usually require a large amount and calculation amount, while lightweight networks generally sacrifice accuracy. The project combines Faster-RCNN and EFFICientDet algorithms The detection framework is performed to calculate the number of wheat ears and the detection of pests and insect pests; compared with the traditional multi -scale fusion method FPN, the network uses BIFPN to learn the importance of different characteristics with the weight of learning. Multi -scale fusion of the upper and lower levels. The traditional network only focuses on the resolution of the zoom trunk network and input pictures. The team proposes a mixed zoom method. At the same time, the resolution, depth, and width of the main network, characteristic network and predictive networks are scaled to achieve the effect of improving accuracy.

Picture shows the detection system screen

The use of science and technology to perform dynamic monitoring of the occurrence and development of crop diseases and insect pests at a large scale, which is of great significance for the timely, efficient and scientific prevention and control of diseases and insect pests. Through artificial intelligence algorithms, the images of the drone are processed in real time, so that the estimated output of wheat is obtained. At the same time, it is monitored at low altitude. Through the algorithm, the current wheat growth status is identified, the pests are recorded, and uploaded to the cloud. At the same time, the wheat detection system provides the corresponding website platform at the same time. Users can watch the monitoring images and results of wheat in real time, greatly saving manpower, accelerating production efficiency, and achieving the goal of "rewarding agriculture to the country". In the discussion, the team members said: "This practice not only made me understand the knowledge outside the book, but also made me feel that I was doing a little power for the country's rural revitalization, which made everyone very excited."

From the preparation of early work to the platform's final test team, the test team lasted more than forty days. The captain said that the team applied the network platform detection system to the farm, which improved the producer's efficiency of the wheat growth of the farm. Inject emerging elements into the construction of smart farms to promote the wisdom of farms and agriculture. "We will make persistent efforts and strive to comprehensively optimize the system. In addition, we can make a force for rural revitalization, which makes our entire team feel relieved. ! "(Li Aimin)

- END -