研究目的
To propose a road extraction method that can overcome the problem of incomplete or fractured road due to shadow occlusion of buildings and vegetation, by extracting roads via road centerline extraction, road width extraction, broken centerline connection, and road reconstruction.
研究成果
The proposed method improves centerline accuracy and integrity through tensor voting and road reconstruction, with high extraction accuracy for road width (average 95.68%), effectively addressing issues caused by shadows and vegetation. Future work will focus on enhancing segmentation, shadow removal, and connection algorithms.
研究不足
The tensor voting algorithm may not connect very large centerline fractures, and the segmentation accuracy can be affected by complex backgrounds and occlusions, leading to potential misjudgments of non-road areas as roads.
1：Experimental Design and Method Selection:
The methodology includes preprocessing of experimental data, initial road extraction using multiscale segmentation and feature extraction, use of FMM to obtain boundary and source distance fields, branch backing-tracking for initial centerline extraction, road width calculation, tensor voting for broken centerline connection, and road reconstruction with final centerline and width matching.
2：Sample Selection and Data Sources:
Three sets of QuickBird images (512x512 pixels,
3：61m resolution) from urban areas, sourced from http:
//www.cse.iitm.ac.in/~vplab/satellite.html.
4：List of Experimental Equipment and Materials:
High-resolution remote sensing images, computational tools for image processing (specific software not mentioned).
5：Experimental Procedures and Operational Workflow:
Steps involve image segmentation with SRM algorithm, feature extraction to remove non-road interference, FMM for distance fields, branch backing-tracking for centerlines, tensor voting for connection, and reconstruction.
6：Data Analysis Methods:
Evaluation using completeness (Com), correctness (Cor), and quality (Q) metrics for centerline extraction, and accuracy assessment for road width extraction.
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
