Cultivated land change detection method based on depth learning

Abstract

The invention discloses a cultivated land change detection method based on depth learning, which is applied to cultivated land census task and has higher convenience and efficiency than manual on-the-spot investigation, at the same time, the remote sensing detection result has higher authenticity compared with the results reported by various levels of departments. Traditional remote sensing interpretation is mainly in the form of artificial visual, is high in human resources cost, but also cannot meet the growing amount of remote sensing data interpretation needs. The invention firstly accurately matches the remote sensing images of the same area taken at different times, then extracts the cultivated land area by adopting the full convolution network, then directly obtains the preliminarycultivated land change result by using the difference method, and finally obtains the final change detection result by adopting the morphological methods such as corrosion and expansion. The inventionalso discloses a method for extracting the cultivated land area from the remote sensing images. The invention adopts the method of depth learning semantic segmentation network to realize automatic extraction and change detection of cultivated land resources, which has higher detection accuracy than the method of the invention in the prior art.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image intelligent ground object monitoring, and more specifically, relates to a cultivated land change detection technology based on a deep learning method under the demand of ground object monitoring. The method realizes the large-scale remote monitoring of cultivated land by remote sensing. Background technique [0002] Cultivated land is the foundation of agriculture. With the advent of the information age, the rapid acquisition and detection of cultivated land information by using spatial information technology has become an inevitable development of the times. Especially in recent years, with the deepening of my country's reform and opening up, accurately grasping the distribution and changes of cultivated land in the region is not only the need of technological development, but also the necessity of macro-management of agricultural development. At present, the extraction of farmland ...

AI Technical Summary

Problems solved by technology

[0004] (1) Most of the traditional methods mainly use the method of manual visual calibration. This method has the characteristics of a large amount of tasks, and the accuracy gradually decreases with the amount of tasks. Moreover, as the amount of remote sensing image data increases, this method can no longer meet the current tasks. need

[0005] (2) The more commonly used automatic change detection method is mainly the pixel difference method, in which the pixel difference method mainly directly differentiates the pixel values ​​of the two images to obtain the result of pixel-by-pixel comparison, and then passes through low pixel value filtering and other processing methods This method is suitable for the situation where the difference between the acquisition conditions of the two images is small, but with the increasing number of sources of remote sensing images, this method is less suitable for shooting images under different loads and lighting conditions

The effect of this method is limited by the accuracy of ground object classification. The previous object comparison method is mostly based on the traditional object classification results, and the low classification accuracy leads to poor performance.

Images