LightGBM is an open-source, distributed and high-performance GB framework built by Microsoft company. LightGBM has some advantages such as fast learning speed, high parallelism efficiency and high-volume data, and so on. Based on the open data set of credit card in Taiwan, five data mining methods, Logistic regression, SVM, neural network, Xgboost and LightGBM, are compared in this paper. The results show that the AUC, F 1-Score and the predictive correct ratio of LightGBM are the best, and that of Xgboost is second. It indicates that LightGBM or Xgboost has a good performance in the prediction of categorical response variables and has a good application value in the big data era.

BACKGROUND: One theoretical advantage of using unmanned aerial vehicles (UAVs) to spray pesticides for maturing corn is that the strong downwash penetrates canopies. However, only few studies have been conducted to examine in-canopy downwash characteristics. This paper investigated the downwash by a six-rotor UAV in mature cornfields. 3D wind speeds in corn canopies and an open area were measured, and comparisons conducted. RESULTS:The downwash by the UAV resulted in in-canopy maximum wind speeds. Z-dimensional downwash was sensitive to all factors, whereas the X-and Y-dimensional downwashes were related to layers and crop positions. Meanwhile, when comparing with the downwash between a 2 m hovering position and the optimal flight parameters, the X-dimensional and Y-dimensional motion time of top-layer downwash generally advanced by 3.8 s and 1.6 s, whereas both motion time and the strength of the Z-dimensional downwash were impeded by ≈2.2-s hysteresis at middle layers and ≈4.5-s time reduction, respectively. Thus, combined with distributions, the corn on the left or right might not be sprayed sufficiently. Furthermore, under the convergence requirement error of 0.01, the overall correlation of the model was ≈0.846 in terms of the Z-dimensional downwash and ≈0.55 and 0.61 for the X-and Y-dimensions, respectively. CONCLUSION: The selection of operation parameters should mainly consider the Z-dimensional downwash. The optimal operation parameters were a height of 2 m with a speed of 4 m s −1 . Meanwhile, the canopy effect could influence the uniformity, motion and strength of downwash. Predictions could be achieved before operation.

BACKGROUND Air‐assisted sprayers is one of the primary fruit tree pest control approaches in agricultural production. It is necessary to study the influence of multiple factors on both wind field and droplet coverage of air‐assisted sprayers. In this article, foliage area volume density (FAVD) and power gradient were considered factors, and field tests were conducted in an orchard to determine such influence. RESULTS The results showed that in‐canopy wind speed was mainly affected by the air‐assisted sprayer. FAVD showed significance to the wind speed and droplet coverage inside canopies (P < 0.001), compared with power gradient (P > 0.05). With the increase of FAVD, the wind speed in the bottom layer of canopies first increased and then decreased, while the wind speed in the middle and top layers first decreased and then increased. Meanwhile, the wind field was mainly concentrated on the surface of canopies and gradually approached the canopy center as the power gradient increased. Furthermore, a Back‐Propagation (BP) neural network prediction model was constructed to predict droplet coverage at any canopy location to avoid repeated experiments. The overall correlation coefficient (R) of this model was about 0.731, indicating good fitting performance. CONCLUSION FAVD has a significant effect on wind speed and droplet deposition inside the canopy, and the air‐assisted sprayer parameter setting should consider the effect of FAVD. The prediction model can predict droplet deposition inside the canopy without repeating. The study can provide a reference for selecting operating parameters of air‐assisted sprayers and help reduce droplet loss and environmental pollution. © 2022 Society of Chemical Industry.

Air-assisted sprayers are widely employed in orchards, but inappropriate spray parameters can lead to large droplet losses, pesticide waste, and environmental pollution. To investigate the factors affecting the droplet loss of an air-assisted sprayer behind canopies, a two-factor, five-level full experiment was conducted in an actual orchard, where the two factors were the power gradient and foliage area volume density (FAVD). In addition, the location of the sampling point was also considered in the data analysis, including horizontal distance, forward distance, and height. The results show that all factors significantly affected droplet coverage (p-value < 0.01). The droplet coverage showed an increase and then a decrease with an increasing power gradient, and the maximum coverage was measured at power gradient P3 (forward speed: 0.49 m/s, spray pressure: 0.30 MPa, and spray flow rate: 7.13 L/min) or P4 (forward speed: 0.58 m/s, spray pressure: 0.35 MPa, and spray flow rate: 8.44 L/min). The effect of FAVD on droplet coverage had obvious regularity, and this regularity did not change with the power gradient. At different positions behind canopies, the droplet coverage had great differences. The droplet coverage gradually decreases with increasing horizontal distance and height, while increasing with forward distance. This study provides a reference for the air-assisted sprayers to reduce droplet loss, and data support for subsequent research on precision spraying based on FAVD.

The classification of plug seedlings is important work in the replanting process. This paper proposed a classification method for plug seedlings based on transfer learning. Firstly, by extracting and graying the interest region of the original image acquired, a regional grayscale cumulative distribution curve is obtained. Calculating the number of peak points of the curve to identify the plug tray specification is then done. Secondly, the transfer learning method based on convolutional neural network is used to construct the classification model of plug seedlings. According to the growth characteristics of the seedlings, 2286 seedlings samples were collected to train the model at the two-leaf and one-heart stages. Finally, the image of the interest region is divided into cell images according to the specification of the plug tray, and the cell images are put into the classification model, thereby classifying the qualified seedling, the unqualified seedling and the lack of seedling. After testing, the identification method of the tray specification has an average accuracy of 100% for the three specifications (50 cells, 72 cells, 105 cells) of the 20-day and 25-day pepper seedlings. Seedling classification models based on the transfer learning method of four different convolutional neural networks (Alexnet, Inception-v3, Resnet-18, VGG16) are constructed and tested. The classification accuracy of the VGG16-based classification model is the best, which is 95.50%, the Alexnet-based classification model has the shortest training time, which is 6 min and 8 s. This research has certain theoretical reference significance for intelligent replanting classification work.