Location/Site
Encinitas Community Park NE ball field bathrooms
(33°01’53.25″N 117°16’47.57″W)
Date
Jun 29th, 2026
Time
07:57 AM
Executive Summary
Mission Objective
Map the bathroom building and surrounding area using a 90º crosshatch pattern in both the SW-NE & SE-NW axes, capturing nadir imagery. This mission will compare the additional perpendicular capture to determine whether it improves the geometry of the orthomosaic and 3D model.
Mission success will be defined as a complete orthomosaic with minimal visible distortion and dimensional measurements within ±5% of field measurements.
Intended Applications
- Site documentation
- Relative dimensional verification
- Pavement, vegetation, and asset inventory
- Existing conditions documentation
- Baseline comparison for future site monitoring
- Capture strategy evaluation
Mission Planning
Airspace
- Class G, Uncontrolled
- No TFRs
Weather Conditions
- Temperature: 64.4º F
- Wind: 6.8 mph, 214° SW
- Cloud Cover: OVC, 1000’ AGL
- Lighting Conditions: Overcast
Mission Success Criteria
- Complete image alignment
- Continuous orthomosaic coverage
- Minimal visible block deformation
- Relative measurements within ±5% of field measurements
- Clearly documented dataset limitations
- Safe and complete mission without any deviations or conflicts
Data Capture Strategy
Flight Parameters
| Parameter | Value |
|---|---|
| Flight Pattern | Automated 90º Crosshatch Grid and Orbitals (4 flights, manually created) |
| Altitude | 120’ AGL (Nadir), 100’ (High Orbit) & 50’ (Low Orbit) |
| Flight Duration | 19m 47s |
| Camera Angle | -90º (Nadir), -45° (High Orbit), -35º (Low Orbit) |
| Front Overlap | 80% |
| Side Overlap | 65% |
| Speed | 2.68 mph |
| Photo Interval | 5 sec |
| ISO | 100 |
| Shutter | 1/640 sec |
| Aspect Ratio | 4:3 |
| Image Capture Count | 149 JPEG (8064 x 6048) |
| Lighting | Uniform Overcast |
Processing Summary
| Metric | Result |
|---|---|
| Images Reconstructed | 149/149 shots (100.0%) |
| Average Reprojection Error | 1.38 px |
| Average GSD | 0.7 cm/pixel |
| Sparse Points | 114505 over 118709 points (96.5%) |
| Dense Points | 10.81 million |
| GPS Source/ CRS | Consumer GPS/WGS84 (No RTK/PPK) |
| GPS Error RMS | 2.16m |
| Outputs | Orthomosaic, DSM, DTM, Point Cloud, Mesh |
Interpretation
Orthomosaic
Produced complete visual coverage of the project area. Stitching and distortion are more pronounced in the Mission 1C model, most evident in parking stall lines, curbs, and roofing. My hypothesis is that this is due to WebODM trying to align perpendicular images with a wide field of view.
Suitable for documentation, visual inspection, asset inventory, and approximate measurements.
Point Cloud
Successfully reconstructed terrain, pavement, vegetation, and major structures. The sparse point clouds were well constructed with 99% of reconstruction points. As expected, building facades, vehicle sides, and tall vegetation were incompletely represented due to the nadir acquisition.
Textured Mesh
Produced a coherent three-dimensional representation of the site. Quality decreased beneath mature trees and along vertical surfaces where image geometry was insufficient. Pavement melted into the vertical surfaces of vehicles.
DSM
Successfully represented overall site topography, including the gradual elevation increase across the parking lot.
Quality Assessment
Strengths
- Complete image reconstruction
- Consistent relative measurement accuracy
- Uniform lighting provided clear visual detail
- No observable large-scale block deformation
- Repeatable acquisition workflow
Limitations
- Absolute positional accuracy is limited by the lack of GNSS corrections
- Missing vertical surfaces & tree occlusions
- Distortion resulting from 90º image stitching
- Vehicle movement introduced localized artifacts
Quality Validation
Relative Measurement Validation
Seven independent field measurements were compared against the reconstructed model.
Results demonstrated relative dimensional agreement generally within 2–4%, with one outlier associated with a table abutting a tall fence, resulting in incomplete capture and measurement inaccuracy.
| Location/object | Lat/Long Coordinates | Actual (inches) | Model (inches) | Delta (Inches) | Signed Error (%) | Model Used |
|---|---|---|---|---|---|---|
| Midway landing on the north stairs (depth) | 33°01'53.67"N 117°16'47.57"W | 48.75 | 47.64 | -1.11 | -2.28% | Ortho |
| Building pillar on NW | 33°01'53.41"N 117°16'47.78"W | 31.75 | 32.28 | 0.53 | 1.67% | Point Cloud |
| Front face men’s bathroom | 33°01'53.09"N 117°16'47.61"W | 120.375 | 119.4 | -0.975 | -0.81% | Point Cloud |
| Manhole cover (parking lot) | 33°01'53.74"N 117°16'47.04"W | 28.875 | 29.28 | 0.405 | 1.40% | Ortho |
| Asphalt patch parking lot | 33°01'52.96"N 117°16'46.88"W | 48.625 | 48.48 | -0.145 | -0.30% | Ortho |
| Top of stairs across pavers (West side, not including white exp joint) | 33°01'53.0"N 117°16'48.1"W | 139.063 | 140.04 | 0.977 | 0.70% | Ortho |
| Tactile strip near HDCP parking (south) | 33°01'52.1"N 117°16'46.8"W | 94.5 | 96.12 | 1.62 | 1.71% | Ortho |
| Steps to pavers (south) | 33°01'52.4"N 117°16'47.5"W | 34.375 | 33 | -1.375 | -4.00% | Ortho |
| Concrete slab near south ballfield (southwest, include white exp joint) | 33°01'52.7"N 117°16'48.3"W | 94 | 92.16 | -1.84 | -1.96% | Ortho |
| Concrete slab near north ballfield (northwest, include white exp joint) | 33°01'53.6"N 117°16'48.5"W | 116.5 | 117.48 | 0.98 | 0.84% | Ortho |
| Dual manhole covers, lower level (northwest) | 33°01'53.5"N 117°16'47.8"W | 43.375 | 43.8 | 0.425 | 0.98% | Ortho |
| Error Calculation | Value |
|---|---|
| Mean Signed Error (Bias) | -0.18% |
| Mean Absolute Error (MAE) | 0.94 |
| Mean Absolute Percent Error (MAPE) | 1.51% |
| Root Mean Square Error (RMSE) | 1.07 |
This dataset is suitable for educational use and approximate measurements but lacks the positional accuracy required for engineering, surveying, or legal-grade geospatial analysis.
Conclusion
Mission 2C was a significant improvement over Missions 2A and 2B. The combination of crosshatch nadir capture and two oblique orbital captures provided more comprehensive coverage of the scene.
The 2D orthomosaic provides adequate visual coverage of the site and exhibits improved relative dimensional consistency (RMSE 1.07 vs 2.42 on Mission 2B), with field-verified measurements within 4% of the observed values. Unfortunately, there were still significant stitching artifacts on the building’s pitched metal roof and on horizontal surfaces adjacent to vertical surfaces. The narrow palm trees did not render well with occlusion of the trunks and branches.
Again, the point cloud was improved but still did not reconstruct well. The oblique capture provided more detail of the building’s broad walls, but there was still a lot of noise and distortion in areas under the eaves and in recesses in the facade. Many of the 90º corners, like curbs, stair steps, and the bases of walls, were very rounded. The building’s corrugated metal roof was occluded, likely due to poor image stitching. There was considerable noise throughout the model, especially in the spectator area outside the backstop fencing, which contributed to the rounding of corners in the textured model. The vertical backstop fencing probably contributed to the increased noise in this area.
Overall, this mission, which combined crosshatch nadir imagery with two orbital oblique captures, produced a better reconstruction than the nadir-only missions, but still has plenty of room for improvement.
This dataset would be suitable for an asset inventory and progress status, rough horizontal and vertical measurements, and relative elevation measurements.