R08.03.01.02 · 任务小类

遥感影像分析

Remote sensing image analysis

Process satellite imagery into classified or index products.

20 条本项目编写的提案 · 分类 v1.1 · 新增基准运行 0 次。下方验证方式与部分得分规则是设计，不是执行结果。

典型应用: SNAP · ENVI · QGIS
小类 GUI 需求标签: HY · 混合

任务提案 · 20

展开查看完整原文与建议检查方式；两种界面均保留提案英文原文。

R08.03.01.02-T01 Compute offset-corrected Sentinel-2 NDVI for an AOI in QGIS

原始指令

input/ has Sentinel-2 L2A bands B04 and B08 (processing baseline 04.00, so BOA_ADD_OFFSET = −1000 applies) and aoi.gpkg. In QGIS compute NDVI with the Raster Calculator using correctly offset reflectance, clip to the AOI and save a Float32 GeoTIFF output/ndvi.tif with nodata −9999.

平台: mac · macOS win · Windows
应用: QGIS
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 2 / 5

建议验证方式

ndvi.tif must match the reference (max abs error 1e-4) over valid pixels, with the input CRS, AOI extent and nodata −9999. Ignoring the −1000 offset produces a detectable bias (>0.01 on vegetation) and fails the value milestone.

建议部分得分

40% offset applied correctly; 30% NDVI values match; 20% clip/extent; 10% data type and nodata.

R08.03.01.02-T02 Build a Sentinel-1 GRD preprocessing graph in SNAP and terrain-correct to 10 m

原始指令

In SNAP open input/S1A_IW_GRDH.zip and, with Graph Builder, chain Apply-Orbit-File (precise orbit in input/aux), Thermal Noise Removal, Calibration to Sigma0 VV/VH, Lee 5×5 speckle filter and Range-Doppler Terrain Correction using input/dem.tif at 10 m, UTM 33N. Save output/graph.xml and write output/s1_tc.tif in dB.

平台: mac · macOS win · Windows
应用: SNAP
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 drag · 拖放 type · 文字与表单
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 3 / 5

建议验证方式

Parse graph.xml: the five operators in order plus a LinearToFromdB step with stated parameters (filter Lee 5×5, external DEM, pixel spacing 10, EPSG:32633). s1_tc.tif: two bands in dB; mean σ0 over 20 test polygons within 0.2 dB of reference.

建议部分得分

30% operator chain; 20% parameters; 30% output radiometry matches; 20% geocoding and pixel spacing.

R08.03.01.02-T03 Train a six-class land-cover classification from hand-drawn ROIs with the SCP plugin

原始指令

In QGIS with the Semi-Automatic Classification Plugin, load input/s2_subset.tif (10 bands). Draw training polygons (≥5 per class) for water, forest, cropland, grassland, urban and bare soil by inspecting the imagery, run Random Forest and save output/classification.tif and output/training.scpx.

平台: mac · macOS win · Windows
应用: QGIS · Semi-Automatic Classification Plugin
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: path · 连续路径 click · 指向与点击 vis · 视觉判断
工作产物: WP10 · 标注
条件: 未声明
提案难度: 4 / 5

建议验证方式

Overall accuracy against 300 hidden validation points ≥0.80 and kappa ≥0.75. training.scpx must hold ≥5 hand-drawn polygons per class (vertex count >4, not auto-generated grids). Validation points are not in the workspace. The PyQGIS console, qgis_process, SCP's Batch tool and the remotior_sensus Python API are disabled.

建议部分得分

25% training ROIs for all classes; 15% classifier run; 40% OA ≥0.80 (scaled from 0.6); 20% kappa ≥0.75.

R08.03.01.02-T04 Map wildfire burn severity from Landsat dNBR in ENVI and report class areas

原始指令

input/pre and input/post hold Landsat 8 surface-reflectance scenes. In ENVI compute NBR for both dates, dNBR = NBR_pre − NBR_post, mask cloud and water from QA_PIXEL, and classify with thresholds 0.10/0.27/0.44/0.66. Save output/severity.tif and class areas (ha) in output/severity.csv.

平台: win · Windows mac · macOS
应用: ENVI
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP04 · 三维 / 空间模型
条件: licensed · 授权专业软件
提案难度: 3 / 5

建议验证方式

severity.tif must agree with the reference class map on ≥98% of valid pixels; class areas within 3%. Masked pixels must be nodata. Reversing the dNBR sign or skipping scale factors fails the agreement threshold.

建议部分得分

25% NBR/dNBR correct; 20% masking; 35% class map agreement; 20% area table.

R08.03.01.02-T05 Map lake water extent from Sentinel-2 NDWI in SNAP and report area

原始指令

Open input/S2B_MSIL2A.zip in SNAP. Resample to 10 m, subset to input/lake_aoi.wkt, compute NDWI = (B3 − B8)/(B3 + B8) with Band Maths, threshold at 0 to a water mask, and export output/water.tif. Write water area in km² to output/answer.json.

平台: mac · macOS win · Windows
应用: SNAP
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 3 / 5

建议验证方式

water.tif must have 10 m pixels over the AOI and IoU ≥0.95 with the reference mask; area within 2% of reference. Band Maths expression recorded in the product history is checked.

建议部分得分

20% resample/subset; 30% NDWI expression; 30% mask IoU; 20% area.

R08.03.01.02-T06 Georeference a 1958 aerial photo against a modern orthophoto with QGIS

原始指令

input/aerial_1958.tif is an unreferenced scan. In the QGIS Georeferencer place at least 6 well-spread GCPs on stable features (road junctions, building corners) against input/ortho_2020.tif, use a first-order polynomial to EPSG:25832, and write output/aerial_1958_geo.tif and output/gcps.points.

平台: mac · macOS win · Windows
应用: QGIS
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 geom · 精确几何 vis · 视觉判断 nav · 平移 / 缩放 / 滚动
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 4 / 5

建议验证方式

Mean error on 10 hidden checkpoints ≤15 m and max ≤30 m; gcps.points must hold ≥6 GCPs spanning ≥60% of image width and height; output CRS EPSG:25832. Feature-matching libraries (OpenCV, SuperGlue) are absent, and the PyQGIS console and GDAL command-line tools are disabled.

建议部分得分

20% ≥6 well-spread GCPs; 20% transform and CRS; 60% checkpoint error (full at ≤15 m, zero at ≥45 m).

R08.03.01.02-T07 Produce a Sentinel-1 InSAR displacement map in SNAP with SNAPHU unwrapping

原始指令

input/ has two S1 IW SLC products. In SNAP run TOPS Split (IW2, bursts 3–5), orbit correction, back-geocoding with SRTM, ESD, interferogram with coherence, deburst, topographic phase removal and Goldstein filtering; export to SNAPHU, unwrap from the terminal, import, convert to LOS displacement and terrain-correct. Save output/los.tif and report displacement at input/point.csv.

平台: lin · Linux mac · macOS
应用: SNAP · SNAPHU
交互表面: desk · 桌面 GUI term · 终端
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单 code · 代码与命令
工作产物: WP07 · 仿真结果
条件: long · 长流程 multiapp · 多应用
提案难度: 5 / 5

建议验证方式

los.tif must correlate ≥0.9 with the reference displacement over coherent pixels (coh >0.4); point value within 5 mm. Mean coherence within 0.02 of reference. SNAP product history must show the full operator chain.

建议部分得分

25% coregistration; 25% interferogram/coherence; 25% unwrapping; 25% displacement map and point value.

R08.03.01.02-T08 Decode Landsat QA bits into a clear-sky mask and report clear fraction in QGIS

原始指令

input/LC09_QA_PIXEL.tif is a Landsat 9 Collection 2 QA band. In QGIS build a mask flagging dilated cloud, cirrus, cloud and cloud shadow (bits 1–4), save output/clear_mask.tif (1 = clear), and report the clear-pixel percentage inside input/aoi.gpkg in output/answer.json.

平台: mac · macOS win · Windows
应用: QGIS
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 3 / 5

建议验证方式

clear_mask.tif must agree with the reference on ≥99.9% of pixels (wrong bit order fails); fill pixels must be excluded from both numerator and denominator. Clear percentage within 0.1.

建议部分得分

50% correct bit decoding; 20% fill handling; 30% percentage correct.

R08.03.01.02-T09 Run object-based segmentation and classification with Orfeo Toolbox in QGIS

原始指令

In QGIS with the OTB provider, segment input/pleiades_subset.tif with Large-Scale Mean-Shift (spatial radius 5, range radius 15, min size 50), compute per-segment band means, train a Random Forest from input/training_points.gpkg, and produce output/objects_class.gpkg with a 'class' field.

平台: mac · macOS win · Windows lin · Linux
应用: QGIS · Orfeo Toolbox
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP10 · 标注
条件: 未声明
提案难度: 3 / 5

建议验证方式

Segment count within 10% of the reference run at the same parameters; overall accuracy against hidden validation points ≥0.80. The QGIS processing history (user-profile history database) must record the OTB parameters.

建议部分得分

30% segmentation with stated parameters; 20% per-segment features; 50% classification accuracy (scaled from 0.6).

R08.03.01.02-T10 Build a cloud-masked summer Sentinel-2 composite for a reservoir in Earth Engine

原始指令

In the Earth Engine Code Editor (sandbox project), draw a polygon tightly around the reservoir described in input/brief.txt, build a Sentinel-2 SR Harmonized median composite for Jun–Aug 2024 with SCL-based cloud and shadow masking, and export bands B2–B8 at 10 m in EPSG:32633 to the project asset 'reservoir_2024'. Save the script path to output/script.txt.

平台: web · 任意系统（网页）
应用: Google Earth Engine
交互表面: brow · 浏览器
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: path · 连续路径 code · 代码与命令 vis · 视觉判断
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 3 / 5

建议验证方式

Asset band means over the reservoir within 1% of the reference composite (no SCL mask → >3% bias). The exported footprint must contain the reservoir polygon and be ≤1.5× its area. Script must be saved in the sandbox repo.

建议部分得分

25% AOI drawn correctly; 30% cloud masking; 30% composite values; 15% export parameters.

R08.03.01.02-T11 Pan-sharpen a WorldView-2 scene and map materials with Spectral Angle Mapper in ENVI

原始指令

In ENVI pan-sharpen input/wv2_ms.tif with input/wv2_pan.tif using Gram-Schmidt, then classify the sharpened image with Spectral Angle Mapper using the five endmembers in input/endmembers.sli and a maximum angle of 0.10 rad. Save output/sam_class.tif and the rule image output/sam_rule.tif.

平台: win · Windows mac · macOS
应用: ENVI
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP10 · 标注
条件: licensed · 授权专业软件
提案难度: 3 / 5

建议验证方式

sam_class.tif must be at pan resolution with overall accuracy ≥0.85 against hidden validation points; unclassified pixels where angle >0.10. Rule image must hold five angle bands. ENVI task log records Gram-Schmidt and SAM parameters.

建议部分得分

30% pan-sharpening; 20% SAM parameters; 40% class accuracy; 10% rule image.

R08.03.01.02-T12 Retrieve lake chlorophyll from Sentinel-3 OLCI with the C2RCC processor in SNAP

原始指令

Open input/S3A_OL_1_EFR.zip in SNAP. Run the C2RCC-OLCI processor with salinity 0.1 PSU and temperature 15 °C, subset to input/lake.wkt, and report mean and median conc_chl over valid pixels (default valid-pixel expression) in output/answer.json. Save the product as output/c2rcc.dim.

平台: mac · macOS win · Windows
应用: SNAP
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 3 / 5

建议验证方式

c2rcc.dim metadata must record salinity 0.1 and temperature 15; mean and median chlorophyll within 5% of the reference run with the same parameters and mask.

建议部分得分

30% processor parameters; 20% subset; 50% statistics correct.

R08.03.01.02-T13 Digitize eight field boundaries from a 0.5 m orthophoto in QGIS

原始指令

In QGIS, using input/ortho_farm.tif, digitize the eight agricultural fields marked by points in input/field_points.gpkg as polygons in output/fields.gpkg, with attribute field_id from the points. Follow visible boundaries (tracks, hedges), enable snapping, and leave no overlaps between fields.

平台: mac · macOS win · Windows
应用: QGIS
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: path · 连续路径 click · 指向与点击 vis · 视觉判断 nav · 平移 / 缩放 / 滚动
工作产物: WP10 · 标注
条件: 未声明
提案难度: 4 / 5

建议验证方式

Each polygon must have IoU ≥0.85 with the gold boundary and the correct field_id; pairwise overlaps <1 m². Segmentation models (SAM, segment-geospatial) are absent; the PyQGIS console is disabled and edges shared by adjacent fields must be snapped (identical vertices).

建议部分得分

80% scaled by fields with IoU ≥0.85 (10 pts each); 10% ids correct; 10% topology clean.

R08.03.01.02-T14 Compute per-parcel NDVI zonal statistics and export a table in QGIS

原始指令

In QGIS run Zonal Statistics of input/ndvi.tif over the 120 parcels in input/parcels.gpkg, computing mean, standard deviation and pixel count, and export the result with parcel_id to output/parcels_ndvi.csv.

平台: mac · macOS win · Windows
应用: QGIS
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 2 / 5

建议验证方式

CSV must hold 120 rows keyed by parcel_id with mean and std within 1e-4 and count exact versus the reference computation (pixel-centre rule).

建议部分得分

40% means; 30% std; 20% counts; 10% CSV format.

R08.03.01.02-T15 Map flood extent from pre/post Sentinel-1 backscatter in SNAP

原始指令

input/ holds preprocessed Sentinel-1 σ0 VV (dB) for before and after a flood, and JRC water occurrence. In SNAP choose a water threshold from the post-event histogram, derive flooded = water_post AND NOT water_pre, exclude permanent water (occurrence >80%), and export output/flood.tif. Report flooded area in km².

平台: mac · macOS win · Windows
应用: SNAP
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: click · 指向与点击 type · 文字与表单 vis · 视觉判断
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 4 / 5

建议验证方式

flood.tif F1 ≥0.80 against the reference flood mask (from optical imagery); permanent water excluded; area within 10%. Threshold must lie between the histogram modes (checked from product history).

建议部分得分

25% threshold choice; 20% change logic; 15% permanent water exclusion; 40% F1 (scaled from 0.5).

R08.03.01.02-T16 Extract mineral endmembers from AVIRIS hyperspectral data and map them in ENVI

原始指令

Open input/cuprite_avirisng.dat in ENVI. Run MNF, Pixel Purity Index and the n-D Visualizer to pick endmembers, identify at least alunite, kaolinite, buddingtonite and calcite against the USGS library, and map them with SAM. Save output/minerals.tif, output/endmembers.sli and output/ids.csv (class, mineral).

平台: win · Windows mac · macOS
应用: ENVI
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 drag · 拖放 vis · 视觉判断
工作产物: WP10 · 标注
条件: licensed · 授权专业软件 long · 长流程
提案难度: 5 / 5

建议验证方式

Each of the four minerals must be identified with an endmember whose spectral angle to the library spectrum is <0.1 rad; per-mineral map agreement with the USGS reference map ≥0.70. n-D Visualizer session file must exist (interactive selection). The IDL command line and ENVI's IDL/Python APIs are disabled.

建议部分得分

15% MNF/PPI; 40% four minerals identified (10 pts each); 35% map agreement; 10% files.

R08.03.01.02-T17 Control: compute the offset-corrected NDVI with GDAL on the command line

原始指令

Control pair of T01. Using gdal_calc.py and gdalwarp from the terminal, compute NDVI from the B04/B08 bands in input/ with BOA_ADD_OFFSET −1000, clip to input/aoi.gpkg, and write Float32 output/ndvi.tif with nodata −9999.

平台: mac · macOS lin · Linux
应用: GDAL
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP04 · 三维 / 空间模型
条件: 未声明
提案难度: 2 / 5

建议验证方式

Same checker as T01 (values within 1e-4, CRS, extent, nodata). Scored as the CLI control of T01.

建议部分得分

40% offset applied; 30% values; 20% clip; 10% type and nodata.

R08.03.01.02-T18 Classify NAIP imagery with the ArcGIS Pro wizard and run an accuracy assessment

原始指令

In ArcGIS Pro open input/naip_4band.tif and run the Image Classification Wizard: object-based, Random Trees, with training samples you draw in the Training Samples Manager for 5 classes. Create 200 stratified random accuracy points, label them from the imagery, and save output/class.tif and output/confusion.dbf.

平台: win · Windows
应用: ArcGIS Pro
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: path · 连续路径 click · 指向与点击 vis · 视觉判断
工作产物: WP10 · 标注
条件: licensed · 授权专业软件
提案难度: 4 / 5

建议验证方式

Overall accuracy against hidden 300-point reference ≥0.85; confusion.dbf must have 200 points with labels agreeing ≥90% with gold. Training sample feature class must hold drawn polygons for all 5 classes. arcpy use is detected from geoprocessing history.

建议部分得分

20% training samples; 20% segmentation and classifier; 40% accuracy; 20% accuracy assessment table.

R08.03.01.02-T19 Compute a 35-year harmonized Landsat NDVI trend map in Earth Engine

原始指令

In the Earth Engine Code Editor (sandbox project), build annual maximum NDVI for 1990–2024 over input/region.geojson from Landsat 5, 7, 8 and 9 Collection 2 SR with cloud masking, harmonize TM/ETM+ to OLI with the Roy et al. (2016) coefficients, compute per-pixel Sen's slope and export 'ndvi_trend' at 30 m.

平台: web · 任意系统（网页）
应用: Google Earth Engine
交互表面: brow · 浏览器
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: code · 代码与命令 click · 指向与点击
工作产物: WP04 · 三维 / 空间模型
条件: long · 长流程
提案难度: 5 / 5

建议验证方式

Exported slope raster RMSE vs reference <0.0005 NDVI/yr over valid pixels; omitting harmonization or cloud masking exceeds the bound. Script must contain 35 annual composites (checked via exported per-year count band).

建议部分得分

25% cloud-masked collections; 20% harmonization; 20% annual max series; 35% Sen's slope export matches.

R08.03.01.02-T20 Find Sentinel-2 scenes whose scene classification labels cloud shadow as water

原始指令

input/ holds six Sentinel-2 L2A scenes. In SNAP compare each true-colour composite with its SCL band and list in output/answer.json the scene IDs where cloud shadows are classified as water (SCL 6) over land.

平台: mac · macOS win · Windows
应用: SNAP
交互表面: desk · 桌面 GUI
GUI 需求 · 任务设定: G2 · GUI 必要
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: click · 指向与点击 vis · 视觉判断 nav · 平移 / 缩放 / 滚动
工作产物: WP12 · 答案 / 决策
条件: 未声明
提案难度: 3 / 5

建议验证方式

Exact set match with gold (2 scenes). Distractor scenes contain real ponds and terrain shadow; product metadata carries no QC flags for the defect. SNAP's gpt and esa_snappy, GDAL and Python raster libraries are not installed on the VM, so scenes must be inspected in SNAP.

遥感影像分析

任务提案 · 20

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

已有研究映射

同工作流族的小类