Wyvern 开放卫星数据源：一次探索性分析

Wyvern's Open Satellite Feed

Source | HN Comments

文章介绍了对 Wyvern 公司开放卫星数据源的探索性分析。Wyvern 是一家加拿大初创公司，其 Dragonette 6U CubeSat 卫星拍摄了高光谱影像。文章作者使用 Python、GDAL、DuckDB 等工具，分析了 Wyvern 开放的 GeoTIFF 数据，并利用 Mapbox 进行地理编码。文章还展示了卫星位置的估计，并提供了数据分析的示例代码和结果。

Mark Litwintschik

我拥有 15 年的咨询和实践经验，服务过英国、美国、瑞典、爱尔兰和德国的客户。以往的客户包括 Bank of America Merrill Lynch, Blackberry, Bloomberg, British Telecom, Ford, Google, ITV, LeoVegas, News UK, Pizza Hut, Royal Mail, T-Mobile, Williams Formula 1, Wise & UBS。我持有加拿大和英国护照。我的简历, Twitter & LinkedIn. 首页 | 基准测试 | 分类 | Atom Feed 发表于 2025 年 3 月 10 日星期一，分类于 GIS

Wyvern's Open Satellite Feed

上个月，位于加拿大埃德蒙顿的 Wyvern 公司（一家拥有 36 名员工，融资 1600 万美元的初创公司）针对其 VNIR（可见光和近红外）的 23 到 31 波段高光谱卫星影像，启动了一项开放数据计划。

这些影像由他们三颗 Dragonette 6U CubeSat 卫星中的一颗所拍摄。这些卫星由 AAC Clyde Space 公司制造，该公司在英国、瑞典和其他一些国家设有办事处。它们在地球表面上方 517 - 550 公里的轨道上运行，星下点空间分辨率 (GSD) 为 5.3 米。

SpaceX 从加利福尼亚州的 Vandenberg Space Force Base 发射了他们的全部三颗卫星。它们的发射时间分别为 2023 年 4 月 15 日、6 月 12 日和 11 月 11 日。

他们位于蒙特利尔的 AWS S3 存储桶中托管了大约 130 GB 的 GeoTIFF 文件。发布的 25 张图像拍摄于 6 月至两周前。

在这篇文章中，我将研究 Wyvern 的开放数据源。

我的工作站

我使用的是 5.7 GHz 的 AMD Ryzen 9 9950X CPU。它有 16 个内核和 32 个线程，以及 1.2 MB 的 L1 缓存，16 MB 的 L2 缓存和 64 MB 的 L3 缓存。它配备了一个液冷散热器，并安装在一个宽敞的全尺寸 Cooler Master HAF 700 电脑机箱中。

该系统具有 96 GB 的 DDR5 RAM，时钟频率为 4,800 MT/s，以及第五代 Crucial T700 4 TB NVMe M.2 SSD，其读取速度高达 12,400 MB/s。SSD 上有一个散热片，以帮助降低其温度。这是我系统的 C 盘。

该系统由一个 1,200 瓦的全模块化 Corsair 电源供电，并安装在 ASRock X870E Nova 90 主板上。

我通过 Microsoft 的 Ubuntu for Windows 在 Windows 11 Pro 上运行 Ubuntu 24 LTS。如果您想知道为什么我不使用基于 Linux 的桌面作为我的主要工作环境，那是因为我仍然在使用 Nvidia GTX 1080 GPU，它在 Windows 上具有更好的驱动程序支持，并且我不时使用 ArcGIS Pro，它仅原生支持 Windows。

安装先决条件

我将使用 GDAL 3.9.3、Python 3.12.3 和其他一些工具来帮助分析本文中的数据。

$sudo add-apt-repository ppa:deadsnakes/ppa
$sudo add-apt-repository ppa:ubuntugis/ubuntugis-unstable
$sudo apt update
$sudo apt install \
gdal-bin \
jq \
libimage-exiftool-perl \
libtiff-tools \
python3-pip \
python3.12-venv

我将设置一个 Python Virtual Environment 并安装一些依赖项。

$python3 -m venv ~/.wyvern
$source ~/.wyvern/bin/activate
$python3 -m pip install \
astropy \
geocoder \
pystac \
rich \
shapely \
sgp4

我将在此帖中使用我的 GeoTIFFs 分析实用程序。

$git clone https://github.com/marklit/geotiffs \
~/geotiffs
$python3 -m pip install \
-r ~/geotiffs/requirements.txt

我将在此帖中使用 DuckDB，以及它的 H3, JSON, Lindel, Parquet 和 Spatial 扩展。

$cd ~
$wget -c https://github.com/duckdb/duckdb/releases/download/v1.1.3/duckdb_cli-linux-amd64.zip
$unzip -j duckdb_cli-linux-amd64.zip
$chmod +x duckdb
$~/duckdb

INSTALL h3 FROM community;
INSTALL lindel FROM community;
INSTALL json;
INSTALL parquet;
INSTALL spatial;

我将设置 DuckDB，以便每次启动时都加载每个已安装的扩展。

$vi ~/.duckdbrc

.timer on
.width 180
LOAD h3;
LOAD lindel;
LOAD json;
LOAD parquet;
LOAD spatial;

本文中的地图是使用 QGIS 3.42 版本渲染的。QGIS 是一个桌面应用程序，可在 Windows、macOS 和 Linux 上运行。该应用程序近年来日益普及，每月约有 1500 万次来自世界各地用户的应用程序启动。

我使用了 QGIS 的 Tile+ plugin，通过 Bing 的 Virtual Earth Basemap 以及 CARTO 添加了地理空间上下文到地图。深色的非卫星影像地图主要由来自 Natural Earth 和 Overture 的矢量数据组成。

Dragonette 卫星

以下是 PulseOrbital 针对 2025 年 3 月 8 日和 9 日 Wyvern 的 Dragonette 星座对塔林上空的估计飞越列表。

下面我将尝试估计他们的三颗卫星各自的当前位置。我在 n2yo 上找到了 Two-line elements (TLE) 详细信息。

我在 2025 年 3 月 10 日运行了以下代码。它生成了一个 CSV 文件，其中包含他们三颗卫星的名称和估计位置。

$python3

from  datetime      import datetime, UTC
import json
from  astropy       import units as u
from  astropy.time    import Time
from  astropy.coordinates import ITRS, \
                 TEME, \
                 CartesianDifferential, \
                 CartesianRepresentation
from  sgp4.api      import Satrec
from  sgp4.api      import SGP4_ERRORS

tles = '''AAC-HSI-SAT1
     1 56225U 23054AZ 25068.93354999 .00004447 00000-0 28073-3 0 9998
     2 56225 97.4339 315.2974 0008749 291.1678 68.8616 15.09548670105253
     AAC-HSI-SAT2
     1 56995U 23084BX 25069.13857882 .00006243 00000-0 33207-3 0 9998
     2 56995 97.7590 200.5235 0013312 352.0602  8.0417 15.15565025 95939
     AAC-HSI-SAT3
     1 58848U 23174DH 25068.89805865 .00005229 00000-0 30564-3 0 9999
     2 58848 97.4007 137.8776 0011809 171.1785 188.9655 15.12246107 73641
     '''.strip().splitlines()
with open('locations.csv', 'w') as f:
  while tles:
    name = tles.pop(0).strip()
    line1 = tles.pop(0).strip()
    line2 = tles.pop(0).strip()
    satellite = Satrec.twoline2rv(line1, line2)
    t = Time(datetime.now(UTC).isoformat().split('+')[0],
         format='isot',
         scale='utc')
    error_code, teme_p, teme_v = satellite.sgp4(t.jd1, t.jd2) # in km and km/s
    if error_code != 0:
      raise RuntimeError(SGP4_ERRORS[error_code])
    teme_p = CartesianRepresentation(teme_p * u.km)
    teme_v = CartesianDifferential(teme_v * u.km / u.s)
    teme = TEME(teme_p.with_differentials(teme_v), obstime=t)
    itrs_geo = teme.transform_to(ITRS(obstime=t))
    location = itrs_geo.earth_location
    loc = location.geodetic
    f.write('"%s", "POINT (%f %f)"\n' % (name,
                       loc.lon.deg,
                       loc.lat.deg))

以下是在 QGIS 中渲染的上述 CSV 数据。

开放数据源

Wyvern 有一个 STAC 目录，其中列出了影像位置和拍摄元数据。我将下载此元数据，并使用 Mapbox 的反向地理编码服务获取每个图像的地址详细信息。Mapbox 的免费套餐每月提供 10 万次地理编码搜索。

$python3

import json
import geocoder
from  pystac      import Catalog
from  rich.progress  import track
from  shapely.geometry import shape

mapbox_key = '...' # WIP: Replace with your key
root = Catalog.from_file(href='https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/catalog.json')
seen = set()
with open('enriched.json', 'w') as f:
  for item in track(list(root.get_items(recursive=True))):
    if item.assets['Cloud optimized GeoTiff'].href in seen:
      continue
    seen.add(item.assets['Cloud optimized GeoTiff'].href)
    centroid_ = shape(item.geometry).centroid
    resp = geocoder.mapbox([centroid_.y, centroid_.x],
                key=mapbox_key,
                method='reverse')
    assert resp.ok is True, resp.status
    f.write(json.dumps(
      {'properties':  item.properties,
       'geom':     shape(item.geometry).wkt,
       'id':      item.id,
       'bbox':     item.bbox,
       'assets':    {k.lower().replace(' ', '_'): v.href
                for k, v in item.assets.items()},
       'mapbox':    resp.current_result.__dict__,
       'collection_id': item.collection_id},
      sort_keys=True) + '\n')

上面生成了一个 33 行的 JSONL 文件。以下是一个示例记录。

$head -n 1 enriched.json | jq -S .

{
"assets":{
"cloud_optimized_geotiff":"https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/wyvern_dragonette-001_20240703T171837_4c406dd3/wyvern_dragonette-001_20240703T171837_4c406dd3.tiff",
"data_mask":"https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/wyvern_dragonette-001_20240703T171837_4c406dd3/wyvern_dragonette-001_20240703T171837_4c406dd3_data_mask.tiff",
"overview_image":"https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/wyvern_dragonette-001_20240703T171837_4c406dd3/wyvern_dragonette-001_20240703T171837_4c406dd3_preview.png",
"pixel_quality_mask":"https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/wyvern_dragonette-001_20240703T171837_4c406dd3/wyvern_dragonette-001_20240703T171837_4c406dd3_pixel_quality_mask.tiff",
"thumbnail_image":"https://wyvern-prod-public-open-data-program.s3.ca-central-1.amazonaws.com/wyvern_dragonette-001_20240703T171837_4c406dd3/wyvern_dragonette-001_20240703T171837_4c406dd3_thumbnail.png"
},
"bbox":[
-117.14246656709959,
48.26920917388883,
-116.64844506075083,
48.6222705607206
],
"collection_id":"2024",
"geom":"POLYGON ((-117.09090885825756 48.570741224333794, -117.09090885825756 48.57083115336763, -117.09070614118478 48.570876117884545, -117.08948983874814 48.57110094046913, -117.08847625338427 48.5712807985368, -117.07421848593255 48.57370888245031, -117.04800041118719 48.57816036962509, -117.03975658356107 48.57955426964951, -117.02671511854598 48.58175753097844, -117.01522781775549 48.58369100520587, -117.00319993810427 48.58571440846713, -116.98792858528867 48.5882773859314, -116.96988676581184 48.59129000856483, -116.9558317154329 48.593628163444514, -116.9458310065094 48.59529185057044, -116.93900619839269 48.596415963493364, -116.89765191554693 48.60320560554782, -116.89461115945532 48.60370021523391, -116.8650144668304 48.60851141854402, -116.85089184409387 48.61080460890678, -116.84812137743262 48.61125425407595, -116.84589148963212 48.61161397021129, -116.81298375148523 48.616919783207486, -116.81237560026692 48.616919783207486, -116.81217288319415 48.6164251735214, -116.81183502140618 48.61552588318306, -116.81122687018787 48.61390716057405, -116.80784825230832 48.60486929267375, -116.78994157754666 48.556667330538794, -116.77967057919281 48.52901415263488, -116.77095374506355 48.50554267480424, -116.76811570604472 48.49789870692836, -116.70905746551009 48.33845452994091, -116.70371924926037 48.324020920010575, -116.7027732362541 48.32145794254631, -116.70250294682374 48.32069354575872, -116.70250294682374 48.32042375865722, -116.70257051918134 48.320378794140304, -116.7035841045452 48.320198936072636, -116.70412468340592 48.3201090070388, -116.97833331051073 48.275054561088034, -116.98002261945051 48.27478477398653, -116.9801577641657 48.27478477398653, -116.98022533652329 48.27482973850345, -116.98036048123846 48.275054561088034, -116.98049562595365 48.27541427722337, -116.98069834302642 48.27595385142637, -116.99819958364253 48.32267198450307, -117.03381021609304 48.417906831333134, -117.0373239786878 48.42730441536877, -117.0754347883692 48.529238975219464, -117.07881340624874 48.53827684311977, -117.08861139809946 48.56449115648234, -117.09003041760887 48.56831314042028, -117.09084128589997 48.57051640174921, -117.09090885825756 48.570741224333794))",
"id":"wyvern_dragonette-001_20240703T171837_4c406dd3",
"mapbox":{
"_geometry":{
"coordinates":[
-116.89724,
48.44575
],
"type":"Point"
},
"fieldnames":[
"accuracy",
"address",
"bbox",
"city",
"confidence",
"country",
"housenumber",
"lat",
"lng",
"ok",
"postal",
"quality",
"raw",
"state",
"status",
"street"
],
"json":{
"address":"38 Miller Gulch, Priest River, Idaho 83821, United States",
"city":"Priest River",
"country":"United States",
"housenumber":"38",
"lat":48.44575,
"lng":-116.89724,
"ok":true,
"postal":"83821",
"quality":1,
"raw":{
"address":"38",
"center":[
-116.89724,
48.44575
],
"context":[
{
"id":"postcode.8128753314501880",
"text":"83821"
},
{
"id":"place.268052716",
"mapbox_id":"dXJuOm1ieHBsYzpEL29vN0E",
"text":"Priest River",
"wikidata":"Q1517705"
},
{
"id":"district.1885932",
"mapbox_id":"dXJuOm1ieHBsYzpITWJz",
"text":"Bonner County",
"wikidata":"Q483932"
},
{
"id":"region.58604",
"mapbox_id":"dXJuOm1ieHBsYzo1T3c",
"short_code":"US-ID",
"text":"Idaho",
"wikidata":"Q1221"
},
{
"id":"country.8940",
"mapbox_id":"dXJuOm1ieHBsYzpJdXc",
"short_code":"us",
"text":"United States",
"wikidata":"Q30"
}
],
"country":"United States",
"district":"Bonner County",
"geometry":{
"coordinates":[
-116.89724,
48.44575
],
"type":"Point"
},
"id":"address.8128753314501880",
"place":"Priest River",
"place_name":"38 Miller Gulch, Priest River, Idaho 83821, United States",
"place_type":[
"address"
],
"postcode":"83821",
"properties":{
"accuracy":"point",
"mapbox_id":"dXJuOm1ieGFkcjplYWRmMWVhNi03MDg3LTQzNjUtOWNkNi05NGFjMWNlZTA5NDg"
},
"region":"Idaho",
"relevance":1,
"text":"Miller Gulch",
"type":"Feature"
},
"state":"Idaho",
"status":"OK"
},
"northeast":[],
"northwest":[],
"raw":{
"address":"38",
"center":[
-116.89724,
48.44575
],
"context":[
{
"id":"postcode.8128753314501880",
"text":"83821"
},
{
"id":"place.268052716",
"mapbox_id":"dXJuOm1ieHBsYzpEL29vN0E",
"text":"Priest River",
"wikidata":"Q1517705"
},
{
"id":"district.1885932",
"mapbox_id":"dXJuOm1ieHBsYzpITWJz",
"text":"Bonner County",
"wikidata":"Q483932"
},
{
"id":"region.58604",
"mapbox_id":"dXJuOm1ieHBsYzo1T3c",
"short_code":"US-ID",
"text":"Idaho",
"wikidata":"Q1221"
},
{
"id":"country.8940",
"mapbox_id":"dXJuOm1ieHBsYzpJdXc",
"short_code":"us",
"text":"United States",
"wikidata":"Q30"
}
],
"country":"United States",
"district":"Bonner County",
"geometry":{
"coordinates":[
-116.89724,
48.44575
],
"type":"Point"
},
"id":"address.8128753314501880",
"place":"Priest River",
"place_name":"38 Miller Gulch, Priest River, Idaho 83821, United States",
"place_type":[
"address"
],
"postcode":"83821",
"properties":{
"accuracy":"point",
"mapbox_id":"dXJuOm1ieGFkcjplYWRmMWVhNi03MDg3LTQzNjUtOWNkNi05NGFjMWNlZTA5NDg"
},
"region":"Idaho",
"relevance":1,
"text":"Miller Gulch",
"type":"Feature"
},
"southeast":[],
"southwest":[]
},
"properties":{
"constellation":"Dragonette",
"created":"2024-11-01T23:03:41Z",
"datetime":"2024-07-03T17:18:40.010328Z",
"end_datetime":"2024-07-03T17:18:42.438451Z",
"eo:cloud_cover":5.63,
"gsd":5.26,
"instruments":[
"VNIR Hyperspectral Imaging Sensor"
],
"license":"other",
"platform":"Dragonette-001",
"processing:facility":"Wyvern",
"processing:level":"L1B",
"processing:version":"1.3",
"product_type":"hyperspectral",
"proj:code":"EPSG:4326",
"proj:shape":[
7311,
7852
],
"providers":[
{
"name":"Wyvern Inc.",
"roles":[
"licensor",
"producer",
"processor"
],
"url":"https://www.wyvern.space/"
}
],
"sat:platform_international_designator":"2023-054AZ",
"sensor_mode":"strip",
"sensor_type":"optical",
"start_datetime":"2024-07-03T17:18:37.582204Z",
"updated":"2024-11-01T23:03:41Z",
"view:azimuth":65.36472836367665,
"view:incidence_angle":2.106011989129783,
"view:off_nadir":2.6882806109241932,
"view:sun_azimuth":116.34019361959561,
"view:sun_elevation":50.52691114712514,
"wyvern:radiometric_resolution":"12"
}
}

数据分析

我将他们的图像元数据加载到 DuckDB 中