Japan New Building Detection - SAR × AI (Annual YoY Diff)

## Overview

This dataset provides AI-detected newly constructed building polygons across Japan, derived from multi-temporal Synthetic Aperture Radar (SAR) satellite observations. Our proprietary AI models compare SAR imagery from multiple time periods and extract structures that have newly emerged between them, delivering the result as query-ready polygons with Snowflake GEOGRAPHY geometry and per-building confidence scores.

<p><br/></p>

Because SAR is independent of cloud cover and daylight, this dataset offers consistent, year-after-year coverage that optical-only sources cannot match — directly queryable inside Snowflake with no imagery download or preprocessing.

<p><br/></p>

## What This Dataset Contains

- Polygon footprints of newly constructed buildings across Japan

- Annual Year-over-Year (YoY) change-detection results

- Per-building metadata: AI confidence score, polygon, centroid, bounding box, area (m²), change-detection window, prefecture codes and names, data version

- Geometry: Snowflake GEOGRAPHY (EPSG:4326), ready for ST_* spatial functions

- Directly usable in Snowflake — no preprocessing required

<p><br/></p>

## How the Data Is Generated

1. Input: multi-temporal C-band SAR satellite imagery (Sentinel-1 constellation)

2. Change detection: proprietary AI models trained on SAR backscatter differences

3. Post-processing: false-positive reduction, polygon shape normalization, and metadata enrichment

4. Geocoding: each polygon is assigned to one or more prefectures via spatial intersection with official administrative boundaries

Note: This dataset represents AI-derived estimates. It is not an official cadastral record or building registry.

<p><br/></p>

## Detection Characteristics

Our models are designed to detect buildings whose appearance produces a meaningful change in SAR backscatter between the two observation periods. Typical examples include:

- Greenfield construction — vacant land or farmland becoming a new structure

- Logistics hubs, factories, and residential developments on previously unused land

- Post-disaster reconstruction where damaged structures were cleared and rebuilt

- Large-scale redevelopment with significant height or volumetric change (e.g., low-rise replaced by a high-rise tower)

<p><br/></p>

A free trial dataset is provided so that customers can evaluate detection behavior against their own areas of interest before purchase.

<p><br/></p>

## Why It Matters

New building emergence is a fundamental signal of urban growth, economic activity, and infrastructure development. With this dataset, you can:

- Quantify regional development trends and identify construction hotspots

- Monitor logistics, industrial, and residential growth

- Track disaster recovery progress through satellite observation

- Enrich insurance and real estate risk models with emerging-asset data

- Drive map and GIS update workflows

<p><br/></p>

## Coverage & Update

**Geographic coverage**

- Area: Japan (46 prefectures; Okinawa not currently covered, support planned for a future release).

**Temporal coverage**

- **Trial dataset** (free 30-day trial, available directly from this Marketplace listing): **v2023 only** — one annual window (2022 → 2023 Year-over-Year diff). After the 30-day period, access to the trial dataset ends — contact [e-mail] to continue via a Private Offer.

- **Full historical dataset** (Private Offer): three annual windows (2022 → 2023, 2023 → 2024, 2024 → 2025), with future annual updates.

- Temporal granularity: annual Year-over-Year diff.

**Refresh**

- Annual updates. Higher-frequency cadences are under active development and available for enterprise engagements on request.

**Cloud / region availability**

- Marketplace direct fulfillment is currently available on **AWS Asia Pacific (Tokyo)**. The listing is discoverable globally, but data is hosted in Tokyo.

- Consumers on other Snowflake regions or clouds (Azure, GCP, AWS US/EU, etc.) can request **cross-cloud delivery via Private Offer** — contact `[e-mail]`.

- Note: government-region clouds (GOV, VPS) are not supported in this delivery model.

**Upgrade path**

- To upgrade from the Trial (v2023) dataset to the full historical / nationwide / enterprise dataset, or for cross-cloud delivery, contact **[e-mail]** for a Private Offer.

<p><br/></p>

## Data Quality Notes

- This dataset is produced by AI analysis; results are statistical estimates and may include false positives and false negatives. The per-building confidence score (PROB) allows downstream filtering by required certainty.

- No building-type classification (e.g., residential / commercial / industrial) is provided.

- The dataset does not aim to match official cadastral records or building registries 1:1.

## <br/>Disclaimer

- This dataset is provided **"AS IS" without any warranties** of accuracy, completeness, fitness for a particular purpose, or non-infringement. AI-derived estimates are inherently subject to error.
- **This dataset is not intended for direct use in business, financial, legal, regulatory, or investment decisions without independent validation by the customer.** Customers are responsible for evaluating the suitability of this data for their specific use cases.
- To the maximum extent permitted by applicable law, Space Shift Inc. shall not be liable for any damages, losses, or liabilities arising from the use of this dataset, whether under the trial offering or any subsequent paid offering.
- For mission-critical use cases, contact `[e-mail]` to discuss enterprise agreements with extended support and validation.

<p><br/></p>

## Usage & Licensing

- Commercial use is permitted under the applicable Listing Terms.

- Redistribution or resale of the raw dataset is not permitted without prior written consent.

- Attribution (Space Shift Inc. / SateAIs) is appreciated when publishing derivative analysis.

<p><br/></p>

## Support & Custom Scope

For customization beyond the standard offering, contact **[e-mail]**:

- **Granularity / bundling**: prefecture-level, regional-block, or nationwide-bundle Private Offers
- **Frequency:** higher-frequency updates (under development; available for enterprise engagements on request)
- **Geographic expansion**: coverage of areas not currently included (e.g., Okinawa Prefecture, select international regions for enterprise engagements)
- **Cross-cloud / region delivery**: see Coverage & Update → Cloud / region availability above
- **Enhanced detection with commercial satellites**: the baseline dataset uses the Sentinel-1 C-band SAR constellation. For customers who require higher spatial resolution, more frequent revisits, or improved detection of smaller structures, we offer tailored analyses combining commercial SAR (e.g., high-resolution X-band) and optical satellite sources. Available for enterprise engagements.

原始信息汇总

数据集名称

Japan New Building Detection - SAR × AI (Annual YoY Diff)

提供方

Space Shift Inc.

定价与访问

• 价格：免费（Trial 数据集）
• 访问模式：Unlimited Access（Trial 期间）
• 升级路径：通过 Private Offer 获取完整历史数据集或企业级服务

数据集概述

该数据集利用多时相合成孔径雷达（SAR）卫星影像（Sentinel-1）和专有 AI 模型，检测日本全境新建建筑物，输出为按年变化的建筑物多边形，并附带 AI 置信度评分。数据集内置于 Snowflake 中，可直接查询，无需下载或预处理影像。

主要内容

• 几何数据：新建建筑物的多边形足迹（Snowflake GEOGRAPHY 格式，EPSG:4326）
• 变化检测：年度同比（Year-over-Year）变化检测结果
• 建筑物元数据：AI 置信度评分、多边形、质心、边界框、面积（m²）、变化检测窗口、都道府县代码与名称、数据版本
• 适用性：可直接使用 Snowflake 的 ST_* 空间函数

数据生成流程

1. 输入：多时相 C 波段 SAR 卫星影像（Sentinel-1 星座）
2. 变化检测：基于 SAR 后向散射差异的专有 AI 模型
3. 后处理：减少误报、多边形形状归一化、元数据丰富
4. 地理编码：通过空间叠加将每个多边形分配到一个或多个都道府县

检测特点

• 可检测因新建引起 SAR 后向散射显著变化的建筑物，如：
  • 空地上新建的物流中心、工厂、住宅
  • 灾后重建（清除受损结构后重建）
  • 大规模再开发（如低层建筑被高层塔楼取代）
• 数据集为 AI 估算结果，非官方地籍或建筑登记记录

覆盖范围与更新

• 地理覆盖：日本（46 个都道府县；不包括冲绳县，未来版本计划支持）
• 时间覆盖：
  • Trial 数据集：仅 v2023（2022→2023 年度同比），免费试用 30 天
  • 完整历史数据集（Private Offer）：三个年度窗口（2022→2023、2023→2024、2024→2025），并提供未来年度更新
• 更新频率：年度更新；更高频率正在开发中，可为企业客户提供
• 云/区域可用性：
  • 直接交付：AWS 亚太（东京）区域
  • 其他区域或云平台（Azure、GCP、AWS 美/欧等）可通过 Private Offer 实现跨云交付
  • 不支持政务云区域（GOV、VPS）

数据质量说明

• AI 分析结果，可能包含误报和漏报；提供每个建筑物的置信度评分（PROB）供下游过滤
• 不提供建筑物类型分类（如住宅/商业/工业）
• 不与官方地籍或建筑登记记录进行 1:1 匹配

免责声明

• 数据集按“现状”提供，不附任何明示或暗示的保证
• 不适用于直接用于商业、财务、法律、监管或投资决策，需客户自行验证

使用与许可

• 允许商业使用（适用于相应的 Listing Terms）
• 未经书面同意，禁止分发或转售原始数据集
• 发布衍生分析时，建议注明来源（Space Shift Inc. / SateAIs）

自定义与扩展支持

• 可根据需求提供：都道府县级、区域级或全国捆绑的 Private Offer
• 支持更高频率更新、地理范围扩展（如冲绳县、国际区域）
• 支持使用商业卫星（高分辨率 X 波段 SAR、光学影像）进行增强检测

适用业务场景

业务领域	应用
地图与地理空间服务	基于新建建筑识别地图修订候选区域
房地产与城市开发	追踪新兴开发区与区域建设趋势
保险与金融	识别新可保资产，支持风险/暴露分析
基础设施与物流	监测建成区扩张以支持服务规划
公共部门与研究	城市规划、区域分析和灾害恢复评估

使用示例（SQL 查询）

• 预览 Trial 数据集：查看新建建筑物多边形及其元数据（例如：FROM SATEAIS.JP_NEW_BUILDINGS_TRIAL）
• 按都道府县统计新建筑：使用 SATEAIS.JP_NEW_BUILDINGS_BY_PREF_TRIAL 视图，按置信度≥0.5 分组统计
• 特定都道府县钻取：查询如 東京都 的新建筑，按置信度排序
• 自定义空间查询：通过经纬度边界框（如 Tsukuba / Ibaraki 物流走廊）检索新建筑

类别

金融、地理空间、政府