Semantic segmentation method and system for remote sensing image fusing GIS data

1. A semantic segmentation method for a remote sensing image fusing geographic information system (GIS) data, comprising:

obtaining a first remote sensing data training set and first GIS data, wherein the first remote sensing data training set comprises n remote sensing training images, and n is a positive integer greater than 1;

preprocessing the first remote sensing data training set to obtain a second remote sensing data training set;

preprocessing the first GIS data based on the second remote sensing data training set to obtain second GIS data;

performing data enhancement on the second remote sensing data training set to obtain a third remote sensing data training set;

training a semantic segmentation model based on the third remote sensing data training set and the second GIS data; and

performing semantic segmentation on the remote sensing image to be segmented based on the first GIS data and the trained semantic segmentation model to obtain a semantic set.

2. The semantic segmentation method for the remote sensing image fusing GIS data according to claim 1 , wherein the preprocessing the first remote sensing data training set to obtain a second remote sensing data training set comprises:

setting the following condition: i=1;

performing geometric correction on an i th remote sensing training image to obtain an i th corrected remote sensing image;

performing coordinate system conversion on the i th corrected remote sensing image to obtain an i th coordinate remote sensing image; and

determining whether i is greater than or equal to n; and if i is less than n, setting the following condition: i=i+1, and returning to the step of “performing geometric correction on an i th remote sensing training image to obtain an i th corrected remote sensing image”; or if i is greater than or equal to n, outputting the second remote sensing data training set, wherein the second remote sensing data training set comprises n coordinate remote sensing images.

3. The semantic segmentation method for the remote sensing image fusing GIS data according to claim 2 , wherein the preprocessing the first GIS data based on the second remote sensing data training set to obtain second GIS data comprises:

setting the following condition: i=1;

aligning the ith coordinate remote sensing image with the first GIS data based on latitude and longitude coordinates to obtain an ith aligned image;

tailoring the first GIS data in the ith aligned image based on the ith coordinate remote sensing image to obtain an ith GIS image; and

determining whether i is greater than or equal to n; and if i is less than n, setting the following condition: i=i+1, and returning to the step of “aligning the ith coordinate remote sensing image with the first GIS data based on latitude and longitude coordinates to obtain an ith aligned image”; or if i is greater than or equal to n, outputting the second GIS data, wherein the second GIS data comprises n GIS images.

4. The semantic segmentation method for the remote sensing image fusing GIS data according to claim 2 , wherein the performing data enhancement on the second remote sensing data training set to obtain a third remote sensing data training set comprises:

setting the following condition: i=1;

zooming in/out the i th coordinate remote sensing image to obtain an i th zoomed-in/out image;

segmenting the ith coordinate remote sensing image to obtain an ith segmented image;

performing mirror flipping on the ith coordinate remote sensing image to obtain an i segmented image;

rotating the zoomed-in/out image, the segmented image, and the flipped image by a predetermined angle for k times to obtain an ith rotated image, wherein k is a positive integer greater than 1; and

determining whether i is greater than or equal to n; and if i is less than n, setting the following condition: i=i+1, and returning to the step of “zooming in/out the ith coordinate remote sensing image to obtain an i th zoomed-in/out image”; or if i is greater than or equal to n, outputting a zoomed-in/out image set, a segmented image set, a flipped image set, and a rotated image set, wherein

the zoomed-in/out image set comprises n zoomed-in/out images, the segmented image set comprises n segmented images, the flipped image set comprises n flipped images, the rotated image set comprises n rotated images, and the third remote sensing data training set comprises the zoomed-in/out image set, the segmented image set, the flipped image set, and the rotated image set.

5. The semantic segmentation method for the remote sensing image fusing GIS data according to claim 2 , after the performing semantic segmentation on the remote sensing image to be segmented based on the first GIS data and the trained semantic segmentation model, further comprising:

optimizing the semantic set based on a conditional random field to obtain an optimized semantic set.