遥感图像变化检测

# 遥感图像变化检测数据集

## 数据集描述

专门用于遥感图像变化检测研究的数据集，包含了完整的图像处理流程和标注信息。该数据集包含24组配准对齐的遥感图像样本，每组样本包含5种不同类型的图像文件和对应的标注文件。

## 数据集特点

- **数据规模**: 24组图像样本
- **图像类型**: 光学图像、SAR图像、二值变化图
- **文件格式**: TIF（原始图像）、PNG（变化黑白二值图）、JSON（标注文件）
- **预处理状态**: 已配准对齐裁剪，尚未按固定分辨率分割
- **标注完整性**: 包含像素级变化检测标注

## 文件结构

数据集包含以下6个目录：

### 图像文件
- **A/**: 高分二号（Gaofen-2）事前光学图像（.tif）- 变化检测的参考基准图像
- **B/**: 高分三号（Gaofen-3）事后SAR图像（.tif）- 合成孔径雷达图像
- **C/**: 哨兵二号（Sentinel-2）未处理的事后光学图像（.tif）- 原始光学图像
- **D/**: 哨兵二号（Sentinel-2）相对辐射校正后的事后光学图像（.tif）- 经过预处理的光学图像
- **E/**: 黑白二值像素变化图（.png）- 变化检测

### 标注文件
- **json/**: 变化图对应的JSON标注文件，可以使用[LabelmeCD-AI](https://github.com/Mriris/labelme_cd_AI)读取和修改

## 数据集用途

### 主要应用场景
1. **变化检测算法研究** - 开发和测试新的变化检测方法
2. **多模态融合** - 研究光学图像与SAR图像的融合技术
3. **图像预处理评估** - 比较不同预处理方法的效果
4. **深度学习** - 作为训练和测试数据

### 研究方向
- 时序遥感图像分析
- 多光谱图像处理
- 城市建筑变化监测

## 技术规格

- **处理状态**: 已配准对齐
- **通道数**: 3

## 注意事项

1. **文件完整性**: 确保A、B、C、D、E目录中的文件数量一致
2. **预处理需求**: 根据具体应用需求，必须进一步统一分辨率
3. **去重**：虽然每组图像是人工单独标注的，但是为了避免重叠区域导致的验证集和训练集混淆，可根据坐标自行去重

## 引用信息

如果您在研究中使用了这个数据集，请引用：

```bibtex
@dataset{remote_sensing_change_detection_2025,
title={remote-sensing-change-detection},
author={Tingxuan Yan},
year={2025},
publisher={Model Scope},
howpublished={\url{https://www.modelscope.cn/datasets/Mriris/remote-sensing-change-detection}}
}
```

## 许可证

本数据集采用 CC BY 4.0 许可证发布，允许自由使用、修改和分发，但需注明出处。

## 联系方式

如有任何问题或建议，请通过以下方式联系：
- modelscope: [@Mriris](https://www.modelscope.cn/Mriris)
- 邮箱: [e-mail]

# Remote Sensing Image Change Detection Dataset

## Dataset Description

A dataset specifically tailored for remote sensing image change detection research, encompassing a complete image processing workflow and annotation information. The dataset comprises 24 registered and aligned remote sensing image samples, where each sample includes 5 distinct types of image files and their corresponding annotation files.

## Dataset Characteristics

- **Data Scale**: 24 sets of image samples
- **Image Types**: Optical images, Synthetic Aperture Radar (SAR) images, binary change maps
- **File Formats**: TIF (raw images), PNG (binary black-and-white change maps), JSON (annotation files)
- **Preprocessing Status**: Registered, aligned and cropped, yet not segmented with fixed resolution
- **Annotation Completeness**: Contains pixel-level change detection annotations

## File Structure

The dataset includes the following 6 directories:

### Image Files
- **A/**: Pre-event optical images (Gaofen-2, .tif) - Reference baseline images for change detection
- **B/**: Post-event SAR images (Gaofen-3, .tif) - Synthetic Aperture Radar images
- **C/**: Unprocessed post-event optical images (Sentinel-2, .tif) - Raw optical images
- **D/**: Post-event optical images after relative radiometric correction (Sentinel-2, .tif) - Preprocessed optical images
- **E/**: Black-and-white binary pixel change maps (.png) - For change detection

### Annotation Files
- **json/**: JSON annotation files corresponding to the change maps, which can be read and modified using [LabelmeCD-AI](https://github.com/Mriris/labelme_cd_AI)

## Dataset Applications

### Main Application Scenarios
1. **Change Detection Algorithm Research** - Develop and test novel change detection methods
2. **Multimodal Fusion** - Investigate fusion technologies for optical and SAR images
3. **Image Preprocessing Evaluation** - Compare the performance of different preprocessing methods
4. **Deep Learning** - Serve as training and testing data

### Research Directions
- Temporal remote sensing image analysis
- Multispectral image processing
- Urban building change monitoring

## Technical Specifications

- **Processing Status**: Registered and aligned
- **Number of Channels**: 3

## Precautions

1. **File Integrity**: Ensure consistent file counts across directories A, B, C, D and E
2. **Preprocessing Requirements**: Further unify the resolution based on specific application requirements
3. **Deduplication**: Although each image set is manually annotated individually, to avoid confusion between the validation and training sets caused by overlapping regions, deduplication can be performed using coordinate information as needed

## Citation Information

If you use this dataset in your research, please cite:
bibtex
@dataset{remote_sensing_change_detection_2025,
title={remote-sensing-change-detection},
author={Tingxuan Yan},
year={2025},
publisher={Model Scope},
howpublished={url{https://www.modelscope.cn/datasets/Mriris/remote-sensing-change-detection}}
}

## License

This dataset is released under the "CC BY 4.0" license, which allows free use, modification and distribution, provided that proper attribution is given.

## Contact Information

For any questions or suggestions, please contact via the following channels:
- ModelScope: [@Mriris](https://www.modelscope.cn/Mriris)
- Email: [e-mail]

搜集汇总

背景与挑战

背景概述

该数据集是一个专门用于遥感图像变化检测研究的资源，包含24对已配准和对齐的遥感图像样本，涵盖光学图像、SAR图像和二进制变化图等多种类型。数据集提供了完整的像素级变化检测标注，适用于算法研究、多模态融合和深度学习等多种应用场景。