Multi-temporal remote sensing image change detection method based on semantic segmentation technology and medium
A technology of semantic segmentation and remote sensing images, applied in the field of intelligent interpretation of remote sensing images, can solve problems such as time-consuming, inefficient, and labor-intensive, and achieve the effect of improving efficiency and saving labor
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0037] Such as Figure 1-3 As shown, the multi-temporal remote sensing image change detection method based on semantic segmentation technology includes the following steps:
[0038] S1. Preprocessing stage: including data preparation, semantic segmentation model building process and training the semantic segmentation model;
[0039] S2. Prediction stage: perform model prediction stage based on the trained semantic segmentation model in S1;
[0040] Compare the multi-temporal remote sensing images, and export the labeling result map after labeling whether it has changed;
[0041] Analyze the pixel pseudo-change of geographic offset in the labeling result map, set the threshold, process the labeling result map based on the morphological operation of the image, and perform threshold screening of the connected regions in the labeling result map.
[0042] Further, the data are remote sensing images and land category labels corresponding to the remote sensing images, and the data ...
Embodiment 2
[0062] 1) Data preparation:
[0063] a) The distribution of vegetation, buildings, crops and other samples in the same city is relatively consistent. The administrative districts and counties of Panzhihua City include: Eastern District, Western District, Renhe District, Miyi County, and Yanbian County. In the embodiment, Panzhihua City is selected as a demonstration site, the remote sensing images of the West District, Renhe District, Miyi County and Yanbian County are used as training samples, and the remote sensing images of the East District are used as test samples. In the training set and the test set, the resolution of the remote sensing images used by each district and county is 1m.
[0064] b) Correspondence between labels and categories in training data:
[0065] Label and category comparison table
[0066]
[0067] Non-building area:
[0068] It consists of landform images related to cultivated land, garden land, woodland, and grassland;
[0069] Construction a...
Embodiment 3
[0092] In the technical effect, for multi-temporal remote sensing images, 1 is the changed area of change detection, 0 is the unchanged area of change detection, and two color coating marks are used to mark the changed area and the unchanged area.
[0093] The traditional multi-temporal remote sensing change detection is mostly to search and compare the image area by area in a manual way. This method has problems such as time-consuming, labor-intensive, and inefficient. This patent aims to assist manual judgment, that is, to find out the changed area through deep learning technology, so that when judging the change, the human will focus more on the area automatically detected by the model, so as to save labor and improve efficiency.
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More