OptimizeAT - Control Point Constrained Aerotriangulation Optimization

Overview

The OptimizeAT interface is used to optimize aerotriangulation results using control points. This interface is called after completing tie point measurement, improving the absolute accuracy of aerotriangulation by introducing ground control points (GCP).

Applicable Scenarios

Surveying projects requiring high absolute accuracy
Have ground control point data
Need to verify and improve aerotriangulation accuracy
Engineering surveying and professional mapping

Workflow

Interface Invocation

Command Line Invocation

reconstruct_full_engine(.exe) -reconstruct_type 3 -task_json optimize_config.json

Parameter Description

reconstruct_type: Fixed as 3 (represents OptimizeAT)
task_json: Configuration file path

Configuration Parameters

OptimizeAT uses the same parameters as ReconstructAT, with additional control point parameters:

Required Parameters

All required parameters of ReconstructAT, plus:

Parameter Name	Type	Description
`control_point`	JSON	Control point group information

Control Point Data Structure

ControlPointGroup

{
  "coordinate_system": {      // Control point coordinate system
    "type": 3,               // Usually projected coordinate system
    "epsg_code": 32650       // Such as UTM Zone 50N
  },
  "points": [                // Control point list
    // ControlPoint object array
  ]
}

ControlPoint

{
  "id": "GCP001",           // Control point name
  "coordinate": [x, y, z],  // Control point coordinates
  "usage": 0,               // 0=Control point, 1=Check point, 2=Disabled
  "observations": [         // Image observations
    {
      "id": 1,             // Image ID
      "uv": [1234.5, 2345.6]  // Pixel coordinates
    }
  ]
}

Complete Configuration Examples

Basic Control Point Optimization

{
  "working_dir": "C:/Projects/AT_Optimize",
  "gdal_folder": "C:/MipMap/SDK/data",
  "coordinate_system": {
    "type": 2,
    "epsg_code": 4326
  },
  "camera_meta_data": [...],  // Same as ReconstructAT
  "image_meta_data": [...],   // Same as ReconstructAT
  "control_point": {
    "coordinate_system": {
      "type": 3,
      "epsg_code": 32650      // UTM Zone 50N
    },
    "points": [
      {
        "id": "GCP001",
        "coordinate": [500123.456, 2500123.456, 123.456],
        "usage": 0,
        "observations": [
          {
            "id": 1,
            "uv": [2736.5, 1824.3]
          },
          {
            "id": 5,
            "uv": [1892.7, 2104.8]
          }
        ]
      },
      {
        "id": "GCP002",
        "coordinate": [500223.456, 2500223.456, 125.678],
        "usage": 0,
        "observations": [
          {
            "id": 3,
            "uv": [3104.2, 1567.9]
          },
          {
            "id": 7,
            "uv": [2345.6, 1890.2]
          }
        ]
      },
      {
        "id": "CHECK001",
        "coordinate": [500323.456, 2500323.456, 127.890],
        "usage": 1,           // Check point
        "observations": [
          {
            "id": 2,
            "uv": [1567.8, 2345.6]
          }
        ]
      }
    ]
  }
}

High-Precision Surveying Project Configuration

{
  "working_dir": "C:/Projects/HighPrecision_AT",
  "gdal_folder": "C:/MipMap/SDK/data",
  "coordinate_system": {
    "type": 2,
    "epsg_code": 4326
  },
  "camera_meta_data": [...],
  "image_meta_data": [
    {
      "id": 1,
      "path": "IMG_0001.JPG",
      "meta_data": {
        "pos": [114.123, 22.123, 150.0],
        "pos_sigma": [0.05, 0.05, 0.10],  // RTK high precision
        "position_constant": false          // Allow optimization
      }
    }
  ],
  "control_point": {
    "coordinate_system": {
      "type": 3,
      "epsg_code": 4978       // ECEF for high precision
    },
    "points": [
      // Multiple evenly distributed control points
      {
        "id": "GCP_NW",
        "coordinate": [...],
        "usage": 0,
        "observations": [...]
      },
      {
        "id": "GCP_NE",
        "coordinate": [...],
        "usage": 0,
        "observations": [...]
      },
      {
        "id": "GCP_SW",
        "coordinate": [...],
        "usage": 0,
        "observations": [...]
      },
      {
        "id": "GCP_SE",
        "coordinate": [...],
        "usage": 0,
        "observations": [...]
      },
      {
        "id": "GCP_CENTER",
        "coordinate": [...],
        "usage": 0,
        "observations": [...]
      }
    ]
  }
}

Control Point Layout Principles

1. Quantity Requirements

Minimum: 3 control points (required for solution)
Recommended: 5-8 control points + 2-3 check points
Large area: At least 1 control point per 50-100 images

2. Distribution Requirements

Ideal control point distribution:

+-----+-----+-----+
| GCP | GCP | GCP |
+-----+-----+-----+
| GCP | CHK | GCP |  GCP: Control point
+-----+-----+-----+  CHK: Check point
| GCP | GCP | GCP |
+-----+-----+-----+

3. Elevation Distribution

Place control points at different elevation levels
Avoid all control points on the same plane
Pay special attention to elevation changes in mountain projects

Tie Point Accuracy Requirements

Pixel Accuracy

Ideal: < 1 pixel
Acceptable: < 2 pixels
Needs improvement: > 3 pixels

Multi-View Observations

Each control point should be visible in at least 2 images
Ideally 3-5 images
Observation angle difference > 15°

Coordinate System Considerations

1. Coordinate System Consistency

{
  // Image positions are usually WGS84
  "coordinate_system": {
    "type": 2,
    "epsg_code": 4326
  },
  
  // Control points are usually local projected coordinate system
  "control_point": {
    "coordinate_system": {
      "type": 3,
      "epsg_code": 32650  // Select according to project location
    }
  }
}

2. Common Coordinate Systems

Region	EPSG	Description
China	4490	CGCS2000 geographic coordinates
China	4547-4554	CGCS2000 Gauss projection
Global	32601-32660	UTM Northern Hemisphere
Global	32701-32760	UTM Southern Hemisphere

Quality Assessment

1. Optimization Report Interpretation

After optimization is complete, check the accuracy report in the logs:

[INFO] Control Point Residuals:
  GCP001: 0.023m (X), 0.015m (Y), 0.041m (Z)
  GCP002: 0.019m (X), 0.022m (Y), 0.038m (Z)
  
[INFO] Check Point Errors:
  CHECK001: 0.045m (X), 0.052m (Y), 0.068m (Z)
  
[INFO] RMS Error: 0.048m

2. Accuracy Standards

Project Type	Planar Accuracy	Elevation Accuracy
Topographic surveying	< 0.05m	< 0.10m
Engineering surveying	< 0.10m	< 0.15m
General applications	< 0.30m	< 0.50m

Best Practices

1. Control Point Collection

Use high-precision equipment such as RTK/total station
Choose stable, easily identifiable features
Record detailed point descriptions and photos

2. Tie Point Process

Load images in aerotriangulation result viewer
Find features corresponding to control points
Precisely mark pixel positions
Check multi-view consistency

3. Iterative Optimization

# First optimization
reconstruct_full_engine.exe -reconstruct_type 3 -task_json optimize_v1.json

# Check results, adjust outliers

# Second optimization
reconstruct_full_engine.exe -reconstruct_type 3 -task_json optimize_v2.json

4. Common Issue Handling

Excessive Residuals

Check tie point accuracy
Verify control point coordinates
Consider setting the point as a check point

Systematic Bias

Check coordinate system settings
Verify control point coordinate system
Consider camera calibration issues

Output Results

Optimized aerotriangulation results are saved in the same location:

milestones/mvs_optimized.xml - Optimized internal format
products/AT/block_exchange_optimized.xml - Optimized exchange format
log/optimization_report.txt - Optimization report

Example: Complete Workflow

import subprocess
import json

# 1. Execute initial aerotriangulation
at_config = {
    "working_dir": "C:/Project",
    # ... Other parameters
}
with open("at_config.json", "w") as f:
    json.dump(at_config, f)
    
subprocess.run(["reconstruct_full_engine.exe", "-reconstruct_type", "1", "-task_json", "at_config.json"])

# 2. Perform control point measurement (usually done in external software)

# 3. Prepare optimization configuration
optimize_config = at_config.copy()
optimize_config["control_point"] = {
    "coordinate_system": {"type": 3, "epsg_code": 32650},
    "points": [
        # Control point data
    ]
}
with open("optimize_config.json", "w") as f:
    json.dump(optimize_config, f)

# 4. Execute optimization
subprocess.run(["reconstruct_full_engine.exe", "-reconstruct_type", "3", "-task_json", "optimize_config.json"])

# 5. Use optimized results for 3D reconstruction
# ...

Next Steps

Use optimized aerotriangulation results for Reconstruct3D
Learn about large-scale data tiling processing
Check basic concepts to learn about control point usage

Tip: Control points are key to ensuring surveying accuracy. Reasonable control point layout and precise tie points are the foundation for obtaining high-precision results.

Overview​

Workflow​

Interface Invocation​

Command Line Invocation​

Parameter Description​

Configuration Parameters​

Required Parameters​

Control Point Data Structure​

ControlPointGroup​

ControlPoint​

Complete Configuration Examples​

Basic Control Point Optimization​

High-Precision Surveying Project Configuration​

Control Point Layout Principles​

1. Quantity Requirements​

2. Distribution Requirements​

3. Elevation Distribution​

Tie Point Accuracy Requirements​

Pixel Accuracy​

Multi-View Observations​

Coordinate System Considerations​

1. Coordinate System Consistency​

2. Common Coordinate Systems​

Quality Assessment​

1. Optimization Report Interpretation​

2. Accuracy Standards​

Best Practices​

1. Control Point Collection​

2. Tie Point Process​

3. Iterative Optimization​

4. Common Issue Handling​

Excessive Residuals​

Systematic Bias​

Output Results​

Example: Complete Workflow​

Next Steps​