Considering that the occurrence and spread of diseases are closely related to the planting environment, a tomato disease diagnosis method based on Multi-ResNet34 multi-modal fusion learning based on residual learning is proposed for the problem of limited recognition rate of a single RGB image of a tomato disease. Based on the ResNet34 backbone network, this paper introduces transfer learning to speed up training, reduce data dependencies, and prevent overfitting due to a small amount of sample data; it also integrates multi-source data (tomato disease image data and environmental parameters). The feature-level multi-modal data fusion method is used to retain the key information of the data to identify the feature, so that the different modal data can complement, support and correct each other, and obtain a more accurate identification effect. Firstly, Mask R-CNN was used to extract partial images of leaves from complex background tomato disease images to reduce the influence of background regions on disease identification. Then, the formed image environment data set was input into the multi-modal fusion model to obtain the identification results of disease types. The proposed multi-modal fusion model Multi-ResNet34 has a classification accuracy of 98.9% for six tomato diseases: bacterial spot, late blight, leaf mold, yellow aspergillosis, gray mold, and early blight, which is higher than that of the single-modal model. With the increase by 1.1%, the effect is obvious. The method in this paper can provide an important basis for the analysis and diagnosis of tomato intelligent greenhouse diseases in the context of agricultural informatization.

Papaver somniferum (opium poppy) is not only a source of raw material for the production of medical narcotic analgesics but also the major raw material for certain psychotropic drugs. Therefore, it is stipulated by law that the cultivation of Papaver somniferum must be authorized by the government under stringent supervision. In certain areas, unauthorized and illicit Papaver somniferum cultivation on private-owned lands occurs from time to time. These illegal Papaver somniferum cultivation sites are dispersedly-distributed and highly-concealed, therefore becoming a tough problem for government supervision. The low-altitude inspection of Papaver somniferum cultivation by unmanned aerial vehicles has the advantages of high efficiency and time saving, but the large amount of image data collected needs to be manually screened, which not only consumes a lot of manpower and material resources but also easily causes omissions. In response to the above problems, this paper proposed a two-stage (target detection and image classification) method for the detection of Papaver somniferum cultivation sites. In the first stage, the YOLOv5s algorithm was used to detect Papaver somniferum images for the purpose of identifying all the suspicious Papaver somniferum images from the original data. In the second stage, the DenseNet121 network was used to classify the detection results from the first stage, so as to exclude the targets other than Papaver somniferum and retain the images containing Papaver somniferum only. For the first stage, YOLOv5s achieved the best overall performance among mainstream target detection models, with a Precision of 97.7%, Recall of 94.9%, and mAP of 97.4%. For the second stage, DenseNet121 with pre-training achieved the best overall performance, with a classification accuracy of 97.33% and a Precision of 95.81%. The experimental comparison results between the one-stage method and the two-stage method suggest that the Recall of the two methods remained the same, but the two-stage method reduced the number of falsely detected images by 73.88%, which greatly reduces the workload for subsequent manual screening of remote sensing Papaver somniferum images. The achievement of this paper provides an effective technical means to solve the problem in the supervision of illicit Papaver somniferum cultivation.

Aiming at low management automation level, heavy workload and other problems that exist in traditional vineyard production, a vineyard production monitoring system is designed based on wireless sensor network technology. The system includes 54 sensor nodes, 3 cluster-head nodes, a sink node and monitoring center. It is free to customize automatic acquisition, processing, transmission, storage of soil moisture rate that from three soil layers separately buried apart from surface 0-20, 20-40, and 40-60 cm, the change of soil temperature and illumination intensity. Each node employs JenOS operation system, and the communication between nodes follows ZigBee protocol. Comprehensively considering the cost, energy consumption, usability and other factors, this system selects the TP-ST series sensors. Experiments of data package transmission rate in budding leafing stage and fruit rapid-expanding stage are taken. Six out of ten sensor nodes have the data package transmission rate above 90%, two sensor nodes below 35%. The reason is that the energy consumption of one solar energy power circuit is excessive, and the other node is deployed where the grape trellis blocking is serious. In addition, the data package transmission rate of fruit rapid-expanding stage is generally lower than that of grape budding leafing stage. This phenomenon shows that the density of branch and fruit has certain influence on the transmission of wireless signal.

Salinization has become one of the most important questions about farmland in China. How to monitor and forecast the salinization trend of the farmland and to carry out valid prevention and cure measures is important for farmland protection in China. The grey model based on the grey system theory is different from the traditional regression forecast analysis. It doesn't need a large number of original data and doesn't require complex computation. It can avoid the fatal weakness of insufficient data and avoid the subjective factors on the forecast precision. By the mechanism of the grey model, the internal rules of the change of the saline-alkali farmland can be scientifically forecasted. It is usually executed with the mode of GM (1, 1). However, the GM (1,1) requires the equal time interval original data sequence. The saline-alkali degree always can't be monitored with the equal time interval. So, a new model by integrating the GM (1,1) with the least square method is proposed. With the new model, the original data sequence, the intervening results and the final forecast results are modified with the least square method. The results showed that the forecast precision was improved with the new model. © 2013 Copyright TSI® Press.