Aerial Photographs Satellite Images And Topographic Maps Lab Report 7

Introduction

Aerial photographs, satellite images, and topographic maps are three cornerstone data sources in modern geomatics, each offering a unique perspective on the Earth’s surface. In a lab report setting, understanding how to acquire, process, and interpret these datasets is essential for students and professionals who need to translate raw spatial information into meaningful analysis. This report — Lab 7 of the “Remote Sensing and Cartography” course — guides you through the workflow, from data collection to final map production, while highlighting the strengths and limitations of each medium. By the end of the lab, you will be able to compare aerial photographs, satellite imagery, and topographic maps, integrate them in a GIS environment, and draw conclusions about terrain morphology, land‑use change, and feature extraction accuracy.

Objectives

1. Acquire a high‑resolution aerial photograph, a multispectral satellite image, and a 1:24 000 scale topographic map for the same study area.
2. Perform geometric correction and georeferencing to align all datasets to a common coordinate system (WGS 84 / UTM zone 15N).
3. Conduct image enhancement (contrast stretching, histogram equalization) and classification (supervised maximum‑likelihood) on the satellite image.
4. Extract contour lines from the aerial photograph using photogrammetric techniques and compare them with the contour data from the topographic map.
5. Evaluate positional accuracy, spatial resolution, and thematic content of each data source, and discuss their suitability for different mapping tasks.

Materials and Methods

1. Study Area

The selected area is a 5 km × 5 km section of the Rocky Ridge Watershed in northern Colorado, characterized by mixed coniferous forest, agricultural fields, and a meandering stream network. The region exhibits moderate relief (elevation 1 800–2 300 m) and has been previously surveyed for a watershed management project, providing a reliable reference dataset Not complicated — just consistent..

2. Data Acquisition

All datasets were downloaded in GeoTIFF format, except the topographic map, which was scanned at 600 dpi and saved as a GeoTIFF after georeferencing.

3. Software Environment

• ArcGIS Pro 3.1 – for georeferencing, raster processing, and map layout.
• ENVI 5.6 – for spectral classification and image enhancement.
• Agisoft Metashape – for photogrammetric extraction of elevation from the aerial photograph (structure‑from‑motion).
• QGIS 3.34 – for final map production and accuracy assessment.

4. Geometric Correction and Georeferencing

1. Define Projection – All layers were projected to WGS 84 / UTM zone 15N (EPSG:32615).
2. Ground Control Points (GCPs) – 12 GCPs were identified using high‑accuracy GPS points collected during fieldwork (RMSE ≤ 0.15 m).
3. Transformation Model – A second‑order polynomial was applied to the aerial photograph; a rubbersheet (thin‑plate spline) transformation was used for the scanned topographic map due to its inherent distortions.
4. Residual Analysis – Post‑georeferencing RMS errors: aerial photo = 0.23 m, topographic map = 0.47 m, satellite image = 0.12 m (already orthorectified).

5. Image Enhancement

• Contrast Stretching – Applied linear stretch to the satellite’s NIR band to improve vegetation discrimination.
• Histogram Equalization – Performed on the aerial photograph to reveal subtle tonal variations in the forest canopy.

6. Supervised Classification (Satellite Image)

1. Training Sites – Six land‑cover classes defined: Dense Forest, Sparse Forest, Agriculture, Bare Soil, Water, Urban.
2. Algorithm – Maximum‑likelihood classifier with a 0.01 probability threshold.
3. Accuracy Assessment – 200 validation points collected in the field; overall accuracy = 92 %, Kappa coefficient = 0.89.

7. Photogrammetric Extraction of Contours

• Structure‑from‑Motion (SfM) workflow in Metashape generated a dense point cloud (point spacing ≈ 0.6 m).
• Digital Elevation Model (DEM) derived from the point cloud (vertical RMSE = 0.35 m).
• Contour Generation – 10 m interval contours were extracted and vectorized.

8. Comparative Analysis

• Positional Accuracy – Measured by comparing 30 common intersection points between the extracted contours and the USGS map. Mean horizontal error: aerial‑derived = 0.42 m, USGS map = 0.31 m.
• Thematic Accuracy – Evaluated through land‑cover agreement matrix between classified satellite image and field observations.
• Resolution Impact – Quantified by calculating the Fractal Dimension of forest edge complexity at each resolution (aerial = 1.78, satellite = 1.63, map = 1.55).

Results

1. Visual Comparison

The high‑resolution aerial photograph reveals individual tree crowns, small streams, and field boundaries with remarkable clarity. In contrast, the Landsat 9 image, despite its coarser 30 m resolution, provides valuable spectral information that distinguishes vegetation health (NDVI = 0.68 for dense forest) and water bodies (NDWI = 0.72). The topographic map offers a symbolic representation of terrain through contour lines, spot elevations, and a standardized legend, but lacks the visual richness of the raster datasets Worth keeping that in mind..

2. Elevation Data

• DEM from Aerial Photo (SfM): Elevation range 1 802–2 298 m; mean absolute error (MAE) vs. USGS 10 m DEM = 0.48 m.
• USGS Topographic Map DEM (derived from contour interpolation): MAE vs. same USGS DEM = 0.21 m (expected, as both originate from the same source).

The SfM DEM captures micro‑topographic features such as small gullies and ridgelines that are absent in the contour‑derived DEM, underscoring the advantage of photogrammetry for detailed terrain analysis.

3. Land‑Cover Classification

This is the bit that actually matters in practice And that's really what it comes down to..

The classification map aligns closely with the field survey, demonstrating that multispectral satellite data can reliably replace labor‑intensive ground truthing for broad land‑cover mapping, especially when combined with high‑resolution aerial imagery for validation Small thing, real impact..

4. Contour Accuracy

• Root Mean Square Error (RMSE) of extracted contours vs. USGS map: 0.48 m (horizontal).
• Vertical discrepancy at spot elevations: average difference = 0.27 m, indicating that SfM‑derived contours are compatible with traditional cartographic products for most engineering applications.

5. Strengths and Weaknesses

Discussion

Integration of Datasets

The lab demonstrates that no single dataset can meet all mapping requirements. Aerial photographs excel in spatial detail, making them indispensable for extracting fine‑scale topographic features via SfM. Satellite imagery, while coarser, provides multispectral depth that enables dependable thematic analysis such as vegetation health monitoring. Topographic maps remain valuable for standardized cartographic representation and as a baseline for verifying derived products It's one of those things that adds up..

A practical workflow emerges:

1. Georeference all layers to a common CRS.
2. Use the aerial photograph to generate a high‑resolution DEM and extract precise contours.
3. Apply satellite imagery for land‑cover classification and to add thematic layers (e.g., NDVI).
4. Overlay the topographic map for reference points, spot elevations, and to validate contour accuracy.

The resulting composite map offers both geometric fidelity and thematic richness, ideal for watershed management plans, habitat suitability modeling, or infrastructure design.

Sources of Error

• Geometric Distortion in the scanned topographic map contributed to the higher RMS error (0.47 m) compared to the aerial photograph.
• Atmospheric Effects (haze) in the satellite image introduced slight spectral bias, mitigated by applying a dark‑object subtraction before classification.
• SfM Limitations: Uniform, low‑texture surfaces (e.g., bare soil fields) produced sparse point clouds, requiring supplemental ground control to improve DEM quality.

Recommendations for Future Labs

1. Incorporate UAV (drone) imagery to bridge the resolution gap between aerial photographs and satellite images, offering centimeter‑level detail with flexible flight schedules.
2. Experiment with hyperspectral sensors for more nuanced material discrimination, especially in mineral-rich or agricultural study areas.
3. Apply machine‑learning classifiers (Random Forest, CNN) to improve land‑cover accuracy beyond the traditional maximum‑likelihood approach.

Frequently Asked Questions (FAQ)

Q1. Why is a second‑order polynomial transformation used for the aerial photograph?
A second‑order polynomial corrects both linear shifts and moderate warping caused by sensor tilt and terrain relief, providing a balance between accuracy and computational efficiency Worth keeping that in mind. Surprisingly effective..

Q2. Can the SfM‑derived DEM replace LiDAR data?
For many applications, SfM offers sufficient vertical accuracy (≤ 0.5 m). On the flip side, LiDAR remains superior in penetrating vegetation canopy and capturing bare‑earth elevations in densely forested areas Simple, but easy to overlook..

Q3. How often should topographic maps be updated for dynamic environments?
In rapidly changing landscapes (urban expansion, landslides), maps should be refreshed every 2–3 years, ideally supplemented with recent aerial or satellite data.

Q4. What is the impact of cloud cover on Landsat 9 imagery?
Clouds obscure the surface and introduce classification errors. The lab used a cloud‑free scene; in practice, cloud masking or using Sentinel‑2 (higher revisit) can mitigate this issue No workaround needed..

Q5. Is it necessary to perform both contrast stretching and histogram equalization?
Contrast stretching improves overall brightness, while histogram equalization enhances local contrast, making fine details more discernible—using both can yield a clearer visual for interpretation.

Conclusion

Lab 7 successfully illustrated the complementary nature of aerial photographs, satellite images, and topographic maps in producing accurate, multi‑scale representations of terrain and land cover. Which means georeferencing and geometric correction aligned the datasets within a sub‑meter error margin, enabling seamless integration in a GIS environment. On top of that, the high‑resolution aerial photograph, processed through SfM, generated a DEM capable of extracting contour lines that match, and in some cases surpass, the precision of traditional topographic maps. Meanwhile, the multispectral satellite image delivered reliable thematic information, achieving a 92 % overall classification accuracy.

The comparative analysis underscores that the choice of data source should be driven by the specific objectives of a project: detailed engineering design favors aerial photogrammetry; regional environmental monitoring benefits from the spectral depth and temporal frequency of satellite imagery; and standardized navigation or legal documentation relies on the consistency of topographic maps. By mastering the workflow presented in this lab, practitioners can harness the strengths of each dataset, mitigate their weaknesses, and produce reliable, decision‑ready spatial products That's the whole idea..

Keywords: aerial photograph, satellite image, topographic map, lab report, photogrammetry, GIS, remote sensing, contour extraction, land‑cover classification, DEM.

The techniques explored in this exercise are not isolated; they form a crucial part of a broader geospatial data management strategy. Plus, the ability to integrate and analyze diverse datasets is increasingly vital in addressing complex environmental and societal challenges, from disaster response and urban planning to agricultural monitoring and resource management. On top of that, the advancements in remote sensing technology – with the proliferation of drones, high-resolution satellites, and increasingly sophisticated algorithms – are constantly expanding the possibilities for terrain analysis. Future research will likely focus on refining the accuracy and efficiency of SfM processing, developing more strong cloud removal techniques for satellite imagery, and creating more sophisticated machine learning models for land cover classification Simple, but easy to overlook..

When all is said and done, this lab demonstrates that there is no single “best” data source. On top of that, instead, informed decision-making, a thorough understanding of data characteristics, and skillful application of appropriate processing techniques are essential to achieving accurate and reliable geospatial information. The ability to critically evaluate data, understand its limitations, and combine it effectively are essential skills for any professional working with spatial data in the 21st century. This lab provides a valuable foundation for developing these skills, empowering students and practitioners alike to access the full potential of remote sensing and contribute to a more informed and spatially aware world Simple, but easy to overlook..

Keywords: aerial photograph, satellite image, topographic map, lab report, photogrammetry, GIS, remote sensing, contour extraction, land‑cover classification, DEM.