Related Experiment Video
Updated: May 28, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.6K
CR-Mask RCNN: An Improved Mask RCNN Method for Airport Runway Detection and Segmentation in Remote Sensing Images
Meng Wan1, Guannan Zhong1,2, Qingshuang Wu1,3
1School of Geography and Tourism, Anhui Normal University, Wuhu 241002, China.
Sensors (Basel, Switzerland)
|February 13, 2025
Summary
This study introduces a new deep learning method for detecting and segmenting airport runways in remote sensing images. The improved approach uses rotated bounding boxes and an attention mechanism for higher accuracy, reducing errors in complex environments.
Area of Science:
- Computer Vision
- Remote Sensing
- Deep Learning
Background:
- Airport runways are critical national infrastructure requiring accurate detection and segmentation in remote sensing images.
- Existing deep learning methods often use horizontal bounding boxes, leading to background interference and inaccuracies.
- Detecting multiple intersecting runways can result in false positives and negatives.
Purpose of the Study:
- To develop an end-to-end method for precise airport runway detection and segmentation in remote sensing data.
- To overcome limitations of horizontal bounding boxes and improve accuracy in complex scenarios.
Main Methods:
- Proposed an improved Mask RCNN (CR-Mask RCNN) incorporating a rotated region generation network.
- Implemented an attention mechanism within the backbone feature extraction network.
- Utilized rotated bounding boxes for more accurate runway localization.
Main Results:
- Rotated bounding boxes demonstrated superior precision compared to horizontal bounding boxes for runway detection, especially in complex backgrounds.
- The attention mechanism significantly enhanced the extraction of local features and detailed information.
- The CR-Mask RCNN method effectively reduced false positives and negatives in runway target detection.
Conclusions:
- The proposed CR-Mask RCNN method offers a highly effective and practical solution for airport runway detection and segmentation.
- Rotated bounding boxes and attention mechanisms are crucial for improving accuracy in remote sensing applications.
- The method enhances the recognition of airport runway targets, proving valuable for infrastructure monitoring.
