Research On Extraction Method Of Maize Flood Damage Area Based On Multi-Source Characteristics

Maize is an important food crop,and its yield stability is of great significance to national food security.Flood disasters are an important factor affecting corn yields.Therefore,quickly and accurately obtaining the area of corn flood damage can help agricultural management departments assess the impact of flood disasters on corn yields and formulate corresponding emergency response plans.Remote sensing images have the advantages of wide coverage and rich information,but the extraction accuracy is not ideal due to the low degree of discrimination between the affected corn and the unaffected corn in the image after the flood disaster.Therefore,how to extract the area of maize flood damage with high precision from post-disaster remote sensing images has been a problem that researchers have been paying attention to.Aiming at the advantages of remote sensing images,this paper conducts research on the basis of deep learning technology,proposes a corn flood-affected area identification and area extraction model,and designs and implements a corn flood-affected area identification and area extraction system.The main work done is as follows:(1)Dataset creation and multi-source feature selection.Collected the Planet remote sensing image data of Tengzhou City,Zaozhuang City,Shandong Province in July and September 2020,and preprocessed the data with remote sensing data processing software,used the preprocessed results to create a Planet image dataset,and selected semantic features,NDVI as features.(2)A model for identification and area extraction of maize flood-affected areas was constructed.Based on the convolutional neural network,this paper constructs a corn floodaffected area identification and area extraction model.The model is divided into two parts: corn waterlogging identification and corn waterlogging segmentation.The data are preprocessed by two groups and sent to the two parts to complete the recognition and segmentation tasks.In the segmentation part of the model,the Involution operator and the PPM module are added to improve the reasoning speed and segmentation accuracy of the model;in the identification part of the model,an attention mechanism is added to improve the feature extraction ability of the model.The experimental results show that the accuracy,precision,recall rate,and F1 value of the model have increased by 1.9%,3.3%,2.2%,and 2.8% respectively compared with those before the improvement,and the average accuracy rate has increased by 3.1%,and after adding the Involution operator The average model inference time was reduced by about 35%.(3)Establish a maize flood-affected area identification and area extraction system.The system has designed multiple functional modules,including remote sensing image dataset production module,dataset preprocessing module,network model training module,and network model testing module.The system is written in C++ language,and a visual interface is designed to facilitate user operation and improve the efficiency of area extraction.These works have practical application value in quickly and accurately extracting the spatial distribution of corn flood damage and accurately obtaining the area of corn flood damage,and provide a basis for using Planet data to extract the area of corn flood damage.