Key Laboratory of Quantitative Remote Sensing, Beijing Agricultural Information Technology Research Center
Agricultural Information Technology National Engineering Research Center
Beijing Agricultural Internet of Things Engineering Technology Research Center
Henan Polytechnic University
Leaf area index (LAI) is an important indicator of plant growth and yield. Hyperspectral remote sensing is a rapid, non-destructive monitoring technique that monitors the nutritional status of crops during the growing season without destroying plant tissue structure. The S185 format airborne high-speed imaging spectrometer can be easily loaded on UAVs to quickly obtain hyperspectral remote sensing data. Therefore, S185 can be used to estimate plant LAI quickly, accurately and in real time, for plant growth monitoring, yield estimation and Pest control has an important role to play.
Beijing Agricultural Information Technology Research Center Quantitative Remote Sensing Key Laboratory and other units used S185 airborne high-speed imaging spectrometer to collect soybean airborne hyperspectral data from five growth periods in Jiaxiang County, Jining, Shandong, to retrieve soybean LAI, and obtain good results and analyze The accuracy, accuracy and stability of three regression models of random forest (RF), artificial neural network (ANN) and support vector machine (SVM) are compared, which provides a method and reference for soybean leaf area index (LAI) inversion. It has certain guiding significance and reference value for the precise fertilization of soybean and the rapid and non-destructive growth monitoring.
Jiaxiang County, Jining, Shandong Province is located at the junction of Zhongnan Mountain and North China Plain. It is located in a warm temperate zone with a continental monsoon climate. In this round of experiments, a total of 126 sample squares were collected as research objects, and the variety combination was produced from 46 soybean varieties planted on June 13, 2015. The data was collected five times: August 1, 2015 (flowering period, R1), August 13, 2015 (first pod period, R3), September 1, 2015 (bulk period, R5), 2015 9 On the 17th (Drums expired, R6), September 28, 2015 (first maturity, R7). The planting density in the test area is about 195,000 / hm2. There are 126 plots in each growing season.
Figure 1 Distribution of sampling points - Jiaxiang County, Shandong Province: Each plot is 3 × 5m rectangular (six rows long 5 meters, 0.5 meters apart)
LAI was measured using a LAI-2200C canopy analyzer. Table 1 shows the distribution of LAI at different growth stages. The LAI values ​​were significant during the R6 and R3 periods (p > 0.05), indicating that the LAI distribution was unbalanced during most of the growth period.
Table 1 Soy LAI statistical description: p value is the result of Kolmogorov Smirnov test
Note: * indicates the observed uniform distribution of LAI
For further study of the RF, ANN and SVM models to estimate LAI, sample selection was performed using SRS and STR sampling methods. The calibration set contained 70% of the total sample and the remaining samples were used as validation sets. To explore the accuracy of the model, LAI inversion was performed throughout the growing season and during the single growth period. The drumming period is the most important period for soybean breeding, during which the LAI distribution is uniform. The RF, ANN and SVM models are established by using the first derivative preprocessor and the STR and SRS sampling methods. The results verify the necessary conditions and advantages of the three models. The specific flow chart is shown in Figure 2.
Figure 2 Method flow chart: STR and SRS two sampling methods to obtain 100 calibration sets. The RF, ANN, SVM, and PLS models were used to obtain 100 V-R2 and V-RMSE values, respectively.
Single-length model results:
The results of the entire growth period model:
Through the comparison of the above model results, the optimal RF model is selected for LAI prediction:
in conclusion:
Leaf area index (LAI) is an important indicator of plant growth and yield and can be quantitatively monitored by the S185 Airborne Rapid Imaging Hyperspectral Instrument. Several models were constructed using the SRS and STR sampling methods, and the final model for inverting the entire growth phase of soybean and LAI in solitary long-term was determined. The partial least squares regression (PLS) model was compared with the RF, ANN, and SVM regression models. The RF model produced the highest accuracy, accuracy, and stability throughout the growing season based on STR sampling. Based on the single growth period of STR sampling, ANN has the highest accuracy, accuracy and stability. The accuracy, accuracy and stability of the RF, ANN and SVM models are improved by the STR sampling method. The RF model is suitable for estimating LAI (whole growing season or more than one growing season) when changes such as plots and growth periods are large. When the changes in plot and growth period are relatively small, ANN is more suitable for estimating LAI (single birth).
Easy lifting and flipping out the food by BBQ tongs for prevent hand to touch food straightly.We have some design for special need customers.3 in 1 tongs combined with slotted spatula and bottle opener,tongs with lighter,longer tongs etc. all you can find in our company .If you have other requirement please contact us.
Bbq Tongs,Grill Tongs,Barbecue Tongs,Food Tongs
MEO DEMO CO., LTD. , https://www.meobbqtools.com