A multi-temporal and spatial scale early warning method and system for drought stress of forest plants

Abstract

The multi-temporal and spatial scale early warning method and system for drought stress degree of forest plants provided by the present invention arrange multiple forest meteorological monitoring nodes and multiple plant physiological parameter monitoring nodes according to the results of feature fusion and cluster analysis. Real-time collection of forest meteorological parameter data and plant physiological parameter data. Feature extraction is performed on the collected forest meteorological parameter data and plant physiological parameter data after preprocessing to obtain feature information, and multiple fused feature vectors are output after feature fusion of the feature information. After the fused feature vector is input into a bidirectional double-layer LSTM model for forward iteration and reverse iteration, the bidirectional double-layer LSTM model outputs a feature vector to be analyzed, and after the feature vector to be analyzed is normalized and classified and expressed, the output layer outputs the drought stress degree of the plant. The advantage is that the error generated between iterations is reduced, and the water content in the stem is used as a direct characterization index to carry out drought stress degree early warning with a large coverage, high accuracy and strong timeliness.

Images