A Hierarchical Deep Learning Crop Remote Sensing Classification Method

Abstract

The invention discloses a hierarchical deep learning (HDL) crop remote sensing classification method. The LLC classification probability of high-level crops can be used as the prior probability of HLC classification of high-level crops. Conversely, the classification probability of HLC can also be used as the prior probability of LLC classification. , the repeated iterations of this process can be expressed by the Markov process model. The HDL model consists of two sub-models (HDL-LLC and HDL-HLC), which are used for the classification of low-level crop LLC and high-level crop HLC respectively, and calculate the classification probability of LLC and HLC classification probability calculations. The invention can realize the remote sensing classification of LLC and HLC at the same time, and incorporates the correlation between LLC and HLC into the classification process, thereby improving the accuracy of remote sensing classification of LLC and HLC, and solving the problem that the current LLC and HLC classifications are separated from each other and cannot be unified.

Description

technical field [0001] The invention relates to a remote sensing classification method, in particular to a hierarchical deep learning remote sensing classification method for crops. Background technique [0002] The agricultural landscape is an ecosystem with distinct hierarchical relationships. According to human knowledge, the agricultural landscape can be divided into macroscopic categories such as fruit trees, forage grass, vegetables, and grains at a broad high level, and can be divided into specific crop types at a low level, such as apple trees, alfalfa, tomatoes, wheat etc. In fact, high-level crops (HLC, High-level crop) and low-level crops (LLC, low-level crop) have their own value in practical applications. For example, the government only needs to know the spatial distribution and area of ​​broad vegetables, but does not need to know the area of ​​specific vegetable types when making agricultural subsidy statistics for cash crops (such as vegetables). The grow...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to provide a method for remote sensing classification of crops with hierarchical deep learning, which can simultaneously realize LLC and HLC remote sensing classification, and incorporate the correlation between LLC and HLC into the classification process, thereby improving the accuracy of LLC and HLC remote sensing classification. LLC and HLC classifications are separated from each other and cannot be unified

Images