当前位置：
X-MOL 学术
›
Comput. Electron. Agric.
›
论文详情
Our official English website, www.x-mol.net, welcomes your
feedback! (Note: you will need to create a separate account there.)
Hyperspectral imaging using multivariate analysis for simulation and prediction of agricultural crops in Ningxia, China
Computers and Electronics in Agriculture ( IF 7.7 ) Pub Date : 2020-05-01 , DOI: 10.1016/j.compag.2020.105355 Guan Hong , Hazem T. Abd El-Hamid
Computers and Electronics in Agriculture ( IF 7.7 ) Pub Date : 2020-05-01 , DOI: 10.1016/j.compag.2020.105355 Guan Hong , Hazem T. Abd El-Hamid
Abstract Crop identification and classification are the basis of fine agricultural development. Hyperspectral remote sensing has been utilized as a cost-efficient approach. Hyperspectral remote sensing data sensing technology has achieved breakthroughs in modern technologies such as long-term dynamic monitoring of crop growth, crop species damage, and acquisition the agricultural information accurately. The novelty of this paper appears in the simulation and prediction of some crops according to wavelength of bands. In the present study, the hyperspectral curves and sensitive bands of 400 samples for four varieties crops (Red medlar or wolfberry, jujube, watermelons and grape) were collected using a hyperspectral instrument. Some multivariate analyses were used for identifying optimal bands, simulation and prediction of different agricultural crops. Principal component analysis (PCA), load coefficient method (LCM), successive projection algorithm (SPA) partial least-squares regression (PLS), and competitive adaptive reweighting sampling (CARS) methods were used to select characteristic wavelengths using Euclidean distance. High value of Euclidean distance was 15.13, 14.095, 16.305 and 14.10 between medlar & jujube, jujube & watermelons, jujube & watermelons and jujube & grape, respectively; it is appears high in CARS among other multivariate analysis. The results show that these characteristic bands can distinguish four kinds of characteristic crops well, but the effect of CARS extraction for selected crops is the best. For prediction of optimal bands, PLS was proven to be the most accurate model of crop prediction, a prediction was developed by applying the CARS-PLS-DA model based on each pixel at the optimal wavelengths. All data and points that selected for wolfberry prediction is very accurate not missing any points for its prediction; on the other hand, foe melon one point of wolfberry appear in the prediction of melon. For Jujube, some points from other crops interface in its prediction; on the other hand grape doesn’t miss any points in its prediction. Prediction of selected crops shows that grape and watermelon have the best crops in specificity and sensitivity, respectively. Cross validation and calibration confirm the possibility of hyperspectral imagine for prediction of all crops especially grape and watermelon. Different multivariate analysis proved its ability to identify optimal bands of diffident species of crops under any conditions. The overall results indicated that hyperspectral imaging could be used for crop identification and prediction for effective classification technique.
更新日期：2020-05-01
京公网安备 11010802027423号