Cultivated land monitoring method and system based on convolutional neural network fused with residual correction

Abstract

The invention discloses a cultivated land monitoring method and system based on a convolutional neural network fused with residual correction. The method comprises the steps: acquiring different-time-phase high-resolution remote sensing images in the same region, carrying out the registration preprocessing, and then carrying out the image cutting; extracting a corresponding cultivated land area vector labeling true value to obtain a cultivated land sample and a background sample in the image area; constructing a convolutional neural network fused with residual error correction, including adding a hole convolution operation based on a Unet network, expanding a receptive field to extract global features, then fusing a residual error correction model to correct a cultivated land extraction result, and taking a result after residual error correction as a cultivated land extraction result; training the convolutional neural network fused with residual correction; for a to-be-extracted high-resolution remote sensing image, cultivated land extraction is carried out according to a training result; the method comprises the steps of performing difference operation and morphological filteringpost-processing on a grid image corresponding to a cultivated land extraction result according to the cultivated land extraction result in remote sensing images of different time phases in the same region to obtain a cultivated land region change detection result.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a cultivated land monitoring scheme based on a convolutional neural network with fusion residual correction. Background technique [0002] Cultivated land refers to the land used for planting crops and farming, which is the resource that people rely on for survival, while basic farmland is the unoccupiable cultivated land planned according to the needs of the country, population and agricultural product production. [0003] Remote sensing imagery can be used to detect changes in land surface types and conditions, that is, change detection using remote sensing imagery. Using multi-temporal high-resolution images acquired at two or more times to detect changes in basic farmland in cultivated land reserves is an effective way to implement permanent basic farmland protection policies. However, most of the change detection using remote sensing im...

AI Technical Summary

Problems solved by technology

[0007] In view of the deficiencies in the detection of existing cultivated land protection areas, the purpose of the present invention is to provide a cultivated land monitoring scheme based on the convolutional neural network with fusion residual correction to extract cultivated land, and then compare the results of cultivated land extraction in images of different time phases

Images