A method of rice extraction from multi-temporal remote sensing images based on rice index

Abstract

The invention relates to a method for extracting rice from multi-temporal remote sensing images based on rice index. Firstly, the remote sensing images of each time phase are geometrically registered, and then the normalized difference vegetation index NDVI of the registered remote sensing images of each time phase is calculated to obtain the time series NDVI image, which is then calculated from the time series NDVI image and the rice growing season. Obtain multiple normalized differential time series indices NDTI, and fuse the NDTI to the minimum value to obtain the rice index RI, then perform threshold segmentation on the RI feature image to obtain the rice extraction result, and finally use the spectral constraints and geometric constraints to perform the rice extraction results. After post-processing, more accurate rice extraction results were obtained. The invention can realize the automatic extraction of rice from multi-temporal remote sensing images without manual selection of training samples, and the invention designs a rice index for rice extraction based on the characteristics of multiple growing seasons of rice, and can be widely used for the extraction of multi-ripening rice.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a method for extracting rice from multi-temporal remote sensing images based on rice index. Background technique [0002] Rice is one of the main food crops in my country, and its growth status affects national food security, agricultural economic development and social stability. The extraction of rice planting areas is the basis for the implementation of rice monitoring, the rational allocation of rice planting resources, and the development of rice production. Satellite remote sensing technology has become an effective method for rapid and automatic extraction of large-scale rice planting areas due to its advantages of periodic and large-scale coverage. . [0003] Multi-temporal remote sensing images can reflect the changes of ground objects at different times, and crops have significant phenological characteristics that change periodically...

AI Technical Summary

Problems solved by technology

However, in the existing methods, the unsupervised method often sets multiple threshold conditions based on multiple vegetation indices or a combination of vegetation indices, which requires a lot of labor, time and energy to get an optimal rice extraction model

However, supervised classification methods, especially those based on deep learning, need to select training samples through a large number of field surveys and manual labeling, and the degree of automation is low.

In addition, the characteristics of multiple growing seasons and more than one ripening of rice in one year are one of the remarkable phenological characteristics that distinguish it from other crops and ground features, and the existing methods rarely make use of this "multi-season" feature

Images