GELE Photogrammetry & Cartography — Stereoscopy, DEM and OrthophotoSummary

Stereoscopy is the optical-perceptual process where two slightly different images of the same object, viewed simultaneously by the left and right eye through a stereoscope, fuse into a single three-dimensional image. In aerial photogrammetry, this is achieved through forward overlap (typically 60%) between successive aerial photographs. The parallax difference — the apparent shift in position of ground objects between the left and right images — provides the geometric basis for extracting terrain heights. The human brain processes these parallax differences to perceive depth, allowing photogrammetrists to measure vertical positions on the ground from aerial imagery.

The base-height ratio is defined as B/H = air-base ÷ flying height, where the air-base is the ground distance between successive camera exposure stations and flying height (H) is the altitude above mean terrain. This dimensionless ratio directly controls the geometric strength of height determination and the degree of vertical exaggeration in the stereoscopic model. A larger B/H ratio (e.g., 0.8) produces stronger height sensitivity but greater vertical exaggeration, making small terrain features appear pronounced. A smaller B/H ratio (e.g., 0.3) produces gentler vertical exaggeration but weaker height measurement capability. Typical aerial survey flights employ B/H ≈ 0.6 (60% forward overlap) to balance both factors. The ratio is inversely related to the parallax bar sensitivity required for manual height measurement.

A Digital Elevation Model is a raster or vector representation of bare-earth terrain elevations at regular grid intervals (e.g., 2 m, 5 m, 10 m posting) or as a Triangulated Irregular Network (TIN). A DEM represents only ground surface elevations, excluding buildings, vegetation, and other surface objects. When derived photogrammetrically, a DEM is produced through automated image matching algorithms that track corresponding pixels across stereo pairs and compute parallax at each grid point. DEMs can also be produced by LiDAR (Light Detection and Ranging), which directly measures ground returns via laser pulses, or by radar interferometry (InSAR). In the Philippines, national DEMs are typically referenced to PRS92 vertical datum. DEMs serve as foundational data for contour generation, slope analysis, volume calculations for earthworks, viewshed analysis, hydrological drainage modeling, and crucially, orthorectification.

A Digital Surface Model represents all above-ground surfaces, including terrain, buildings, vegetation canopies, and other features. Unlike a DEM (bare-earth only), a DSM includes the first-return reflective surface encountered by measurement sensors (photogrammetric matching or LiDAR pulses). In automated photogrammetric processing, an initial DSM is typically generated first through image correlation; this DSM is then filtered and processed to remove non-terrain points, yielding a DEM. DSMs are valuable for urban planning, building volumetric estimation, vegetation height determination, and power-line clearance analysis. However, where terrain elevation for hydrological modeling, civil engineering design, or geological analysis is required, the DEM must be used, as DSM values would be inflated by surface objects.

Relief displacement is the apparent lateral shift of image positions due to terrain elevation variations in a single photograph. Points at higher elevation appear displaced radially outward from the principal point (center of the photo), while points at lower elevation are displaced inward. Relief displacement magnitude is proportional to object height h, radial distance r from the principal point, and inversely proportional to flying height H: displacement ≈ (r × h) / H. In a single photograph, this displacement makes the scale non-uniform — ground distance cannot be measured directly because scale varies with both position (r) and elevation (h). Relief displacement is a key reason why raw aerial photographs cannot function as maps and why orthophotos are necessary for measurement applications.

Tilt displacement occurs when the camera axis is not vertical at the moment of exposure. Any tilt (deviation from nadir) causes additional scale variation across the photograph, with scale larger at the downslope direction and smaller at the upslope direction. Unlike relief displacement (which depends on terrain elevation), tilt displacement depends solely on camera attitude and focal length. In modern GPS/IMU-equipped aerial systems, tilt angles are typically less than 1–2°, minimizing tilt displacement; however, in archival photography or oblique imagery, tilt can be substantial. Combined relief and tilt displacement is collectively termed 'image distortion.' Both must be removed through orthorectification to produce a geometrically accurate orthophoto.

An orthophoto is an aerial photograph that has been geometrically corrected using a DEM (and camera orientation data) so that every point is projected to its correct planimetric (map) position at a uniform scale. The orthorectification process resamples pixel values from the original image according to the DEM elevation grid, effectively removing relief and tilt displacement. The result is an image product that combines photographic detail and tone variation with cartographic geometric accuracy — it can be measured like a map while retaining the visual information content of the original photograph. Orthophotos are typically produced at a specified pixel size (e.g., 0.5 m, 1 m) and referenced to a standard coordinate system (PPCS/UTM with PRS92 datum in the Philippines). A collection of adjacent orthophotos seamlessly mosaicked together is termed an orthomosaic or orthophoto mosaic, which functions as a visually rich base map.

Forward overlap (also called along-track overlap or longitudinal overlap) is the percentage of image area that is redundantly covered by successive photographs along a flight line. Standard practice specifies 60% forward overlap for stereophotogrammetry, meaning each point on the ground (except edges) is imaged by at least two successive photographs, enabling stereo viewing. From overlap percentage and photo format, the air-base is computed. Sidelap (cross-track overlap) refers to overlap between adjacent flight lines, typically 20–30% to ensure seamless coverage and allow relative orientation between lines. Forward overlap of 60% combined with sidelap of 25–30% creates a systematic network where every ground point is multiply imaged, enabling robust DEM generation, orthomosaic mosaicking, and aerotriangulation.

Parallax is the apparent shift in position of an object in successive images due to a change in viewpoint. In stereophotography, parallax is caused by the horizontal displacement of the camera between successive exposures. Absolute parallax p is the algebraic difference in image coordinates on left and right photographs for the same ground point: p = x_left − x_right. In the standard stereomodel, points at higher elevation have larger parallax values. The parallax-height relationship is the fundamental equation of photogrammetric height extraction: Δh = h × (Δp / p_base), where h is the flying height above the reference, Δp is the parallax difference from reference, and p_base is the parallax at the reference elevation. Manual parallax measurement uses a parallax bar (a stereoscopic measurement device with a movable dot or mark that the operator adjusts until it appears to lie on the ground feature); the parallax bar reading directly yields elevation. Modern digital photogrammetry automates parallax measurement through image matching algorithms, eliminating manual parallax bar use but applying the same parallax-to-height principle.

Vertical exaggeration is the apparent amplification of terrain relief when viewed stereoscopically, caused by the human visual system's processing of parallax. The eye is adapted to interpret horizontal parallax differences based on terrestrial viewing distances; when parallax is created by aerial photography at high altitude, the perspective is distorted. The degree of exaggeration is approximately proportional to the B/H ratio: a model with B/H = 0.6 exhibits roughly 2–3× vertical exaggeration, meaning a 100 m elevation difference appears as 200–300 m in stereo viewing. This exaggeration aids visual interpretation of terrain (subtle slopes become visually apparent) but can mislead if misinterpreted — for example, terrain slopes appear steeper than true slope angles. For rigorous engineering work, elevations must be extracted numerically (via parallax measurement), not estimated from visual stereo impression.

Photogrammetric DEMs are produced through automated image matching and parallax extraction across stereopairs. Modern workflows employ area-based matching (correlation) or feature-based matching to identify corresponding pixels in left and right images. For each grid point in the desired DEM, the matching algorithm finds the corresponding position in the stereopair, computes parallax, and converts parallax to elevation using the camera calibration, image coordinates, and exterior orientation parameters. The process typically generates an initial DSM (all surfaces); subsequent filtering removes non-terrain points (buildings, vegetation) to yield a bare-earth DEM. Point density and accuracy depend on image resolution, matching algorithm robustness, and terrain texture (areas with poor texture or heavy shadow can produce unreliable results). Modern UAV and commercial satellite stereopairs routinely produce DEMs with 0.5–5 m grid spacing and vertical accuracy of ±0.5–2 m, depending on system and ground control.

Light Detection and Ranging (LiDAR) is an active sensor that emits laser pulses and measures return times to compute distances to reflecting surfaces. Airborne LiDAR directly measures ground elevation (and first-return DSM) without requiring stereopairs or image matching. Vertical accuracy is typically ±0.1–0.5 m, and ground spacing is often denser than photogrammetric DEMs (0.5–2 m). LiDAR penetrates some vegetation, enabling bare-earth DEM extraction even under forest canopy. Synthetic Aperture Radar Interferometry (InSAR) uses phase differences between radar images from two slightly different positions to compute elevation. InSAR coverage is continental in scale and penetrates cloud cover, making it valuable for tropical regions where persistent cloud limits optical imagery. However, InSAR vertical accuracy is typically ±5–20 m (coarser than LiDAR or photogrammetry) and requires careful phase unwrapping. All three methods (photogrammetry, LiDAR, InSAR) are used in Philippine national mapping programs (e.g., the National Mapping and Resource Information Authority — NAMRIA, now Bureau of Fisheries and Aquatic Resources — BFAR, or similar agencies).

DEMs enable numerous engineering and planning analyses: (1) Contour generation — automatic extraction of iso-elevation lines at specified intervals; (2) Slope and aspect analysis — quantifying terrain steepness and orientation for stability, visibility, and solar exposure studies; (3) Volume and earthwork computation — calculating cut/fill quantities for proposed grading or dam designs; (4) Viewshed analysis — determining visibility from observer points for telecommunications, surveillance, or landscape assessment; (5) Hydrological modeling — deriving flow direction and accumulation networks for drainage design and flood forecasting; (6) Orthorectification — enabling production of orthophotos; (7) 3D visualization — generating digital terrain models for planning presentations and environmental impact assessment. In the Philippines, DEMs support disaster risk reduction (flood and landslide hazard mapping), infrastructure planning (road networks, utility corridors), and natural resource management (forest mapping, agricultural zoning) under frameworks such as CA 141 (Public Land Act) and national mapping mandates.

Orthorectification is the process of geometrically correcting an aerial photograph using a DEM and camera orientation data to remove relief and tilt displacement. The mathematical process involves: (1) Establishing exterior orientation parameters (six values: three position components X₀, Y₀, Z₀ and three rotation angles ω, φ, κ) for the raw image; (2) For each pixel in the output orthophoto grid, determining its ground coordinates; (3) Using the DEM at those ground coordinates to find the corresponding 3D position in object space; (4) Applying the inverse camera perspective transformation to find the corresponding pixel position in the original image; (5) Interpolating pixel values from the original image and placing them in the orthophoto output. The result is an image grid where pixel positions correspond to map coordinates (typically UTM/PPCS in the Philippines) at a uniform scale. Accuracy depends on DEM quality, accuracy of orientation parameters (from aerotriangulation), and image resolution.

An orthomosaic is created by seamlessly tiling multiple adjacent orthophotos into a single large image. Where adjacent orthophotos meet at edges, radiometric and geometric discrepancies can occur: (1) Radiometric differences arise from varying atmospheric conditions, lighting angle, and sensor gain during different exposures; (2) Geometric misalignment occurs if orientation or DEM errors cause orthophotos to not perfectly align. Mosaicking software typically performs edge-blending (feathering) to blend radiometric differences along seam lines, and fine-tuning of orthophoto positions to achieve geometric coherence. A well-executed orthomosaic is seamless — boundaries between individual orthophotos are visually imperceptible — and functions as a unified map product suitable for measurement, planning overlays, and public dissemination. Orthomosaics are routine deliverables in Philippine mapping projects and are often distributed through geoportals and open-data initiatives.

Stereoscopic overlap between successive photographs creates parallax (image displacement) for identical ground points. Parallax is the quantitative link between image geometry and terrain elevation. By measuring parallax differences and applying the photogrammetric equations, heights are extracted, forming the foundation for DEM generation.

A larger B/H ratio increases parallax differences for the same elevation change, making heights more sensitive and easier to measure precisely (stronger geometry) but also increasing vertical exaggeration in stereo viewing. Conversely, small B/H ratios reduce vertical exaggeration but weaken height sensitivity. Designers must balance these competing factors; typical practice uses B/H ≈ 0.6.

Orthorectification mathematically fuses three inputs to transform a distorted raw photograph into a geometrically accurate map-equivalent image. The DEM provides elevation information to correct relief displacement; orientation parameters provide camera attitude to correct tilt displacement. Without a DEM, raw photographs retain scale variation and cannot be measured as maps.

Both relief and tilt effects distort image scale, making raw photographs unsuitable for measurement. Relief displacement depends on terrain elevation, while tilt displacement depends on camera attitude. Combined, they create total image distortion that orthorectification must correct.

Flight planning begins with specifying forward overlap (typically 60%). Overlap percentage and photo format determine air-base (distance between exposures). Air-base and flying height determine B/H, which governs measurement sensitivity and project cost. This chain of relationships shows how strategic flight-planning decisions cascade through the entire photogrammetric workflow.

Orthorectification accuracy depends directly on DEM accuracy; errors in DEM elevations propagate as positional errors in the orthophoto. An orthophoto derived from a coarse or inaccurate DEM may be unsuitable for engineering measurement or detailed planning applications. DEM validation is therefore critical for orthophoto quality assurance.

Three complementary methods produce DEMs with distinct characteristics. Photogrammetry excels in areas with good image texture and reasonable cost but requires stereopairs. LiDAR penetrates vegetation and provides high accuracy but has higher cost. InSAR provides continent-scale coverage and penetrates clouds but has coarser accuracy. Project specifications determine which method is most appropriate.

A DEM is not merely an end product but a foundational input enabling diverse analyses. From a single DEM, contours are generated, slopes are computed, volumes are calculated, viewsheds are derived, and hydrological networks are extracted. This highlights the multiplier effect of investing in high-quality DEM data.

Individual orthophotos are tiled into seamless orthomosaics that function as visually rich, geometrically accurate base maps. Planners and designers overlay infrastructure projects, zoning decisions, and environmental analyses on orthomosaics, leveraging the combination of photographic detail and cartographic accuracy.

While vertical exaggeration aids visual feature detection in stereo viewing (useful for photo interpretation), it distorts height perception. Therefore, engineering decisions must be based on numerical height measurements (via parallax or DEM extraction), not on visual impressions from stereo models. This relationship emphasizes the complementary roles of qualitative interpretation and quantitative measurement.

Municipal engineers in Metro Manila or Cebu use orthophotos and DEMs to plan road networks, water distribution systems, and electrical utilities. Orthophotos provide the base map layer showing buildings and existing infrastructure; DEMs enable slope analysis for drainage design and viewshed analysis for telecommunications tower placement. The combination of photographic detail (orthophoto) and elevation data (DEM) streamlines planning workflows.

The Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) and local government units use DEMs to model flood inundation extent under various rainfall scenarios. Hydrological models depend on accurate terrain data (DEM) to route flow, accumulate runoff, and predict water surface elevations. Orthophotos overlay on DEM-derived flood maps show affected buildings, roads, and communities. These products guide evacuation planning and post-disaster assessment.

The Department of Agriculture and Bureau of Forest Development use orthophotos for crop mapping, plantation monitoring, and deforestation detection. DEMs support slope-based classification for agroforestry suitability; contour lines derived from DEMs guide terracing and conservation agriculture. DSMs derived from stereopairs or LiDAR quantify vegetation heights and biomass, informing forestry management and carbon sequestration estimates.

Under the Public Land Act (CA 141), orthophotos and DEMs support land classification into agricultural, residential, industrial, forest, and other categories. Elevation-based classification (using DEMs) distinguishes steep slopes unsuitable for agriculture from gentle slopes suitable for cultivation. Orthophotos enable feature-based delineation of water bodies, built-up areas, and open spaces. Combined products support titling and cadastral surveys.

Mining companies use orthophotos and DEMs for mine site planning, volumetric calculations, and environmental monitoring. A baseline DEM prior to mining, compared against post-mining DEMs, quantifies excavation volume and subsidence. Orthorectified aerial photographs track vegetation removal, tailings pile geometry, and rehabilitation progress over time.

Highway and railway designers use orthophotos as base maps to route corridors through existing settlements, avoiding obstacles. DEMs enable slope analysis to assess grading difficulty, cut/fill volume computation, and cost estimation. Profiles extracted from DEMs along proposed alignments guide vertical geometry design. Orthophotos and DEMs together support environmental and social impact assessments.

Orthophotos and bathymetric DEMs support coastal zone planning, aquaculture site identification, and marine protected area delineation. Aerial orthophotos show coastal features (mangroves, seagrass beds, coral reefs, built infrastructure); marine DEMs reveal bathymetry and channels. Together they enable integrated coastal management.

Utility companies use orthophotos to identify corridor options for water pipelines, electrical transmission lines, and fiber optic cables. DEMs enable visibility analysis (viewshed) for telecommunications tower site selection and support accessibility analysis (slope-based) for pipeline maintenance. Orthomosaics serve as base layers for utility asset inventory and maintenance scheduling.

Real estate developers and appraisers use orthophotos for property boundary identification, site area measurement, and market analysis. Orthophotos provide historical snapshots enabling temporal analysis of urban growth and property value trends. DEMs inform elevation-based property classification and premium assessments for properties with advantageous view or drainage.

Researchers and government agencies maintain archived orthophotos spanning decades, enabling historical analysis of landscape change. Multi-temporal orthophoto comparison quantifies urban sprawl, deforestation, infrastructure expansion, and natural disasters. Orthomosaics from different years overlaid in GIS systems reveal temporal patterns and validate environmental compliance.