研究目的
To develop a novel fine registration technique for very high resolution remote sensing images to estimate and mitigate residual local misalignments.
研究成果
The proposed fine registration approach effectively reduces local residual misalignments in VHR images through superpixel-based RN estimation and local rectification, improving overall registration accuracy as confirmed by experimental results.
研究不足
The method assumes images are standardly registered; performance may vary with image scenes, particularly in repetitive or homogeneous textures. Future work needed for more robust estimation objects.
1：Experimental Design and Method Selection:
A two-step strategy involving superpixel segmentation and frequency filtering to generate sparse superpixels for RN estimation, followed by local rectification using optimized control points.
2：Sample Selection and Data Sources:
Two datasets from Chinese GF2, GF1, and ZY3 satellites, with subsets of 6000*6000 pixels containing various terrains.
3：List of Experimental Equipment and Materials:
Not specified in the paper.
4：Experimental Procedures and Operational Workflow:
Superpixel segmentation using SLIC, frequency filtering with DBOP filter, RN estimation via spatial correlation, control point optimization, and local affine transformation using TIN.
5：Data Analysis Methods:
Visual interpretation and correlation coefficient analysis for RN estimation and registration accuracy assessment.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容-
GF-2 satellite
China
Acquisition of very high resolution remote sensing images
-
GF-1 satellite
China
Acquisition of very high resolution remote sensing images
-
ZY-3 satellite
China
Acquisition of very high resolution remote sensing images
-
SLIC algorithm
Superpixel segmentation for image processing
-
DBOP filter
Frequency filtering to generate bandpass frequency maps
-
RANSAC algorithm
Mismatch detection for control points
-
TIN
Triangulated irregular network for local affine transformation
-
登录查看剩余5件设备及参数对照表
查看全部
