11.2 Digital Photogrammetry, UAV/Drone Surveying, Orthophoto Generation, and LiDAR Data Processing

11.2 Digital Photogrammetry, UAV/Drone Surveying, Orthophoto Generation, and LiDAR Data Processing

Digital photogrammetry and remote sensing technology have evolved rapidly, replacing traditional analogue analytical plotters with automated computer vision algorithms, Unmanned Aerial Vehicle (UAV) survey platforms, and airborne Light Detection and Ranging (LiDAR) sensors. Modern Geodetic Engineers leverage these tools to generate high-resolution orthomosaics, 3D point clouds, and high-precision terrain models for land administration, engineering design, and environmental monitoring under Land Management Bureau (LMB) guidelines.

---

1. Digital Cameras and Ground Sample Distance (GSD)

Digital aerial cameras use solid-state Charge-Coupled Device (CCD) or Complementary Metal-Oxide-Semiconductor (CMOS) sensor matrices to record electromagnetic radiation directly as digital grid pixels.

Ground Sample Distance (GSD)

Ground Sample Distance (GSD) is the physical distance between the centers of two consecutive image pixels measured on the ground surface. GSD represents the fundamental spatial resolution limit of a digital imagery survey.

$$\text{GSD} = \frac{H' \cdot p_s}{f} = \frac{(H - h) \cdot p_s}{f}$$

Where:

• $\text{GSD} = \text{Ground Sample Distance (m/pixel or cm/pixel)}$
• $H' = H - h = \text{flying height above ground level (m)}$
• $p_s = \text{physical pixel pitch or sensor element size (m or } \mu\text{m)}$
• $f = \text{effective camera focal length (m or mm)}$

Alternatively, if sensor width ($W_{sensor}$) in pixels ($N_{pixels}$) and sensor physical width ($S_w$) are known:

$$p_s = \frac{S_w}{N_{pixels}} \implies \text{GSD} = \frac{H' \cdot S_w}{f \cdot N_{pixels}}$$

[!TIP] GSD Rule of Thumb: To detect and measure features reliably for cadastral mapping (e.g., boundary monuments of $15\text{ cm}$ diameter), the flight GSD should be selected such that the target spans at least 3 to 5 pixels (e.g., $\text{GSD} \le 2.5 - 3.0\text{ cm/pixel}$).# 2. UAV / Drone Surveying Workflows and Parameters

Unmanned Aerial Vehicle (UAV) surveying combines lightweight drone aircraft platforms with compact high-resolution RGB optical sensors or multispectral cameras.

UAV Aircraft Configurations

1. Multi-Rotor UAVs (Quadcopters, Hexacopters): Vertical Take-Off and Landing (VTOL) capability, high maneuvering flexibility, hover ability. Ideal for small, localized site surveys ($< 50\text{ hectares}$), complex structures, and corridor detail mapping.
2. Fixed-Wing UAVs: Aerodynamic wing lift allowing high flight speeds and extended endurance (60–120+ minutes). Preferred for large-area rural surveys ($> 100\text{ hectares}$), agricultural mapping, and linear pipeline/highway corridors.

Direct Georeferencing: RTK / PPK vs. Conventional GCPs

• Real-Time Kinematic (RTK) UAVs: The drone onboard GNSS receiver receives real-time differential corrections from a local GNSS base station or Network RTK service (NTRIP via cellular link), logging camera exposure center phase center coordinates to centimeter precision ($1-2\text{ cm}$).
• Post-Processed Kinematic (PPK) UAVs: Onboard raw GNSS satellite observational data and base station observations are post-processed after the flight using carrier-phase differential algorithms. PPK eliminates radio link dropouts during flight.
• GCP Requirement: Even with RTK/PPK GNSS positioning, independent Check Points (CPs) (and at least 3–5 anchor GCPs) must be surveyed on the ground to verify vertical block systematic shifts and satisfy LMB regulatory accuracy audits.

---

3. Structure from Motion (SfM) and Multi-View Stereo (MVS)

Structure from Motion (SfM) is a photogrammetric technology derived from computer vision. SfM automatically reconstructs 3D scene geometry, camera positions, and camera calibration parameters from a series of highly overlapping, unordered 2D digital photographs.

+-----------------------+     +-----------------------+     +-----------------------+
| 2D Overlapping Photos | --> | SIFT Feature Matching | --> | Bundle Adjustment     |
| (80% End, 70% Side)   |     | Keypoint Alignment    |     | (Sparse Point Cloud)  |
+-----------------------+     +-----------------------+     +-----------------------+
                                                                        |
                                                                        v
+-----------------------+     +-----------------------+     +-----------------------+
| High-Res Orthomosaic  | <-- | DTM/DSM Generation    | <-- | Multi-View Stereo     |
| & Contour Maps        |     | Surface Mesh Model    |     | (Dense Point Cloud)   |
+-----------------------+     +-----------------------+     +-----------------------+

Sequential SfM Processing Pipeline

1. Feature Detection & Description: Computer vision algorithms (e.g., Scale-Invariant Feature Transform - SIFT) identify tens of thousands of invariant keypoints across each photo frame.
2. Feature Matching & Relative Orientation: Keypoints are matched across multiple overlapping photographs using robust matching algorithms (RANSAC) to establish relative orientation vectors.
3. Bundle Block Adjustment (BBA): Simultaneous non-linear least-squares adjustment based on collinearity equations: $$x - x_0 = -f \left[ \frac{m_{11}(X - X_0) + m_{12}(Y - Y_0) + m_{13}(Z - Z_0)}{m_{31}(X - X_0) + m_{32}(Y - Y_0) + m_{33}(Z - Z_0)} \right]$$ $$y - y_0 = -f \left[ \frac{m_{21}(X - X_0) + m_{22}(Y - Y_0) + m_{23}(Z - Z_0)}{m_{31}(X - X_0) + m_{32}(Y - Y_0) + m_{33}(Z - Z_0)} \right]$$ BBA solves for interior orientation ($f, x_0, y_0, k_1, k_2, p_1, p_2$), exterior orientation ($X_0, Y_0, Z_0, \omega, \phi, \kappa$), and 3D sparse cloud coordinates ($X, Y, Z$) simultaneously.
4. Multi-View Stereo (MVS) Dense Matching: Dense depth maps are calculated across all pixel matches, generating a dense 3D point cloud containing millions of $X, Y, Z, R, G, B$ points.

---

4. Digital Elevation Models and True Orthophoto Generation

In spatial photogrammetry, elevation data models are strictly differentiated based on feature inclusion:

DEM Surface Terminology

• Digital Surface Model (DSM): Captures the top reflective elevation surface of the Earth, including natural vegetation canopies, buildings, bridges, powerlines, and elevated infrastructure.
• Digital Terrain Model (DTM): Represents the bare-earth surface stripped of all above-ground vegetation, artificial structures, and temporary objects. DTM incorporates structural breaklines (ridge lines, valley bottoms, road curbs).
• Digital Elevation Model (DEM): Generic umbrella term representing digital elevation raster arrays; in Philippine NAMRIA standards, DEM commonly denotes bare-earth DTM grids.

True Orthophoto vs. Traditional Orthophoto

Orthorectification is the process of eliminating geometric distortions caused by camera tilt and terrain relief displacement, converting a perspective photo into an orthographic projection (map-accurate image).

• Traditional Orthophoto: Rectified using a bare-earth DTM. Terrain distortion is removed, but above-ground objects (buildings, bridges) lean radially away from the nadir point, obscuring adjacent ground areas (building lean / occlusion shadows).
• True Orthophoto: Rectified using a high-density DSM. Both terrain displacement and building lean are eliminated. Buildings appear strictly vertically rectilinear, and occluded ground zones are filled using pixel data seamlessly retrieved from adjacent overlapping multi-angle photos.

---

5. Airborne LiDAR Principles and Data Processing

LiDAR (Light Detection and Ranging) is an active optical remote sensing technique that measures distances to target surfaces by emitting high-frequency directional laser pulses and recording the elapsed travel time of reflected light.

LiDAR Time-of-Flight (TOF) Principle

$$d = \frac{c \cdot t}{2}$$

Where:

• $d = \text{distance (range) from laser sensor to ground target (m)}$
• $c = \text{speed of light in atmosphere } (\approx 299,792,458\text{ m/s})$
• $t = \text{round-trip pulse travel time (seconds or nanoseconds)}$

Integrated Airborne LiDAR System Components

1. Laser Scanner Unit: Emits pulses (typically near-infrared $\lambda = 1,064\text{ nm}$ for topographic land mapping, or green $\lambda = 532\text{ nm}$ for bathymetric water penetration) at rates exceeding $100\text{ kHz} - 1.5\text{ MHz}$.
2. Geodetic GNSS Receiver: Records sensor phase position ($X, Y, Z$) at $1-10\text{ Hz}$.
3. Inertial Measurement Unit (IMU): Measures aircraft angular orientation roll ($\omega$), pitch ($\phi$), and yaw ($\kappa$) at high frequencies ($100-500\text{ Hz}$).

Multi-Return LiDAR Canopy Penetration

When a broad laser beam pulse encounters porous forest foliage, fractions of the pulse reflect off upper leaves, intermediate branches, and lower structures, while the remaining energy reaches the ground surface. The sensor records multiple Discrete Returns (First, Second, Third, Last Return).

Aircraft LiDAR Scanner
      \  |  /
       \ | /
        \|/     Pulse Laser Beam
       / | \
      /  |  \  
   +-----+-----+  <-- First Return: Top Tree Canopy
  /  Foliage    \
 +---------------+ <-- Second Return: Intermediate Branches
        |  
       ===        <-- Last Return: Bare Ground Surface
-------------------------------------------------------------

ASPRS LAS Point Cloud Classification Standard

The American Society for Photogrammetry and Remote Sensing (ASPRS) defines standardized numerical class codes stored within binary .LAS or .LAZ point cloud files:

Vertical Accuracy Assessment (NSSDA / ASPRS Standards)

Vertical precision is evaluated by comparing LiDAR-derived elevation grid values ($Z_{LiDAR}$) against independent high-precision GNSS check points ($Z_{Check}$):

$$\text{RMSE}_z = \sqrt{\frac{\sum_{i=1}^{N} (Z_{LiDAR, i} - Z_{Check, i})^2}{N}}$$

According to National Standard for Spatial Data Accuracy (NSSDA) guidelines:

$$\text{Vertical Accuracy at 95\% Confidence Level} = 1.9600 \times \text{RMSE}_z$$

---

6. Step-by-Step Worked Numerical Examples

Example 1: UAV Ground Sample Distance (GSD) Calculation

Problem: A quadcopter UAV carries a digital camera with focal length $f = 24\text{ mm}$ and a 20-Megapixel CMOS sensor having a physical sensor width of $13.2\text{ mm}$ and $5,472\text{ pixels}$ horizontally. The drone flies at an altitude $H' = 120\text{ m}$ above flat terrain. Calculate:

1. The physical pixel size ($p_s$) in micrometers ($\mu\text{m}$).
2. The Ground Sample Distance (GSD) in centimeters per pixel.

Solution:

Step 1: Calculate physical pixel size ($p_s$)

$$p_s = \frac{\text{Sensor Width}}{\text{Horizontal Pixels}} = \frac{13.2\text{ mm}}{5,472\text{ pixels}} = 0.00241228\text{ mm} = 2.412\text{ }\mu\text{m}$$ $$p_s = 2.41228 \times 10^{-6}\text{ m}$$

Step 2: Calculate GSD Convert $f = 24\text{ mm} = 0.024\text{ m}$:

$$\text{GSD} = \frac{H' \cdot p_s}{f} = \frac{120\text{ m} \times (2.41228 \times 10^{-6}\text{ m})}{0.024\text{ m}}$$ $$\text{GSD} = \frac{0.0002894736}{0.024} = 0.012061\text{ m/pixel} = 1.206\text{ cm/pixel}$$

Answer: The pixel size is $2.41\text{ }\mu\text{m}$ and the GSD is $1.21\text{ cm/pixel}$.

---

Example 2: LiDAR Time-of-Flight and Elevation Calculation

Problem: An airborne LiDAR system flying at an altitude $H = 1,500.00\text{ m}$ above MSL emits a laser pulse vertically downward. The sensor electronics record a round-trip travel time $t = 8.342\text{ microseconds } (\mu\text{s})$ for the last return laser signal. Calculate:

1. The range distance ($d$) from the aircraft to the ground target.
2. The terrain elevation ($h$) above MSL of the ground reflection point.

Solution:

Step 1: Convert time to seconds

$$t = 8.342\text{ }\mu\text{s} = 8.342 \times 10^{-6}\text{ seconds}$$

Step 2: Compute range distance ($d$) Using $c = 299,792,458\text{ m/s}$:

$$d = \frac{c \cdot t}{2} = \frac{299,792,458\text{ m/s} \times (8.342 \times 10^{-6}\text{ s})}{2}$$ $$d = \frac{2,500.8687}{2} = 1,250.434\text{ m}$$

Step 3: Compute ground elevation ($h$)

$$h = H - d = 1,500.00\text{ m} - 1,250.434\text{ m} = 249.566\text{ m}$$

Answer: The measured range is $1,250.43\text{ m}$ and the ground elevation is $249.57\text{ m}$ above MSL.

---

Example 3: LiDAR Vertical Accuracy Assessment (RMSE_z)

Problem: A Geodetic Engineer verifies a LiDAR DTM against 5 high-precision GNSS check points. The elevation residuals ($Z_{LiDAR} - Z_{GNSS}$) are recorded as follows:

• Checkpoint 1: $+0.04\text{ m}$
• Checkpoint 2: $-0.06\text{ m}$
• Checkpoint 3: $+0.03\text{ m}$
• Checkpoint 4: $-0.05\text{ m}$
• Checkpoint 5: $+0.02\text{ m}$

Calculate:

1. The Root Mean Square Error in elevation ($\text{RMSE}_z$).
2. The vertical accuracy at the 95% confidence level under NSSDA guidelines.

Solution:

Step 1: Compute squared residuals

$$\Delta Z_1^2 = (+0.04)^2 = 0.0016$$ $$\Delta Z_2^2 = (-0.06)^2 = 0.0036$$ $$\Delta Z_3^2 = (+0.03)^2 = 0.0009$$ $$\Delta Z_4^2 = (-0.05)^2 = 0.0025$$ $$\Delta Z_5^2 = (+0.02)^2 = 0.0004$$

Step 2: Sum squared residuals

$$\sum \Delta Z^2 = 0.0016 + 0.0036 + 0.0009 + 0.0025 + 0.0004 = 0.0090\text{ m}^2$$

Step 3: Calculate $\text{RMSE}_z$

$$\text{RMSE}_z = \sqrt{\frac{0.0090}{5}} = \sqrt{0.0018} = 0.042426\text{ m} = 4.24\text{ cm}$$

Step 4: Compute NSSDA 95% Vertical Accuracy

$$\text{Vertical Accuracy}_{95\%} = 1.9600 \times \text{RMSE}_z = 1.9600 \times 0.042426\text{ m} = 0.083155\text{ m} = 8.32\text{ cm}$$

Answer: The vertical $\text{RMSE}_z$ is $4.24\text{ cm}$ and the 95% vertical accuracy is $8.32\text{ cm}$.