Geometric Corrections in Remote Sensing Summary
Geometric Corrections in Remote Sensing
Geometric Correction Concept
- Definition: Geometric correction involves modifying the position of pixels in an image to improve accuracy.
- Importance: Enables accurate extraction of distance, area, and direction from images.
- Transformation: Changes coordinate positions (c', l') based on original position (c, l) or map coordinates (x, y).
Types of Geometric Error
- Internal Geometric Error: Caused by the remote sensing system itself.
- Sources: Earth rotation, scanning system variations, relief displacement, tangential scale distortion.
- Skew: Introduced by Earth rotation effects.
- Ground Resolution Cell Size Variation: Affected by system altitude and scan angles.
- Example: 705 km above ground level for Landsat 7 minimizes distortion.
- One-Dimensional Relief Displacement: Affects objects above local terrain, creating displacement in off-nadir views.
- Tangential Scale Distortion: Differences in scanning distance at nadir vs. the edge, altering object shape.
External Geometric Error
- Causes: Random aircraft movements during data collection including altitude and attitude changes (roll, pitch, yaw).
- Impacts: Variable imagery scale depending on altitude; gyro stabilization equipment can mitigate roll/pitch errors.
Ground Control Points (GCPs)
- Definition: Known locations on Earth (e.g., intersections) used to calibrate image data.
- Requirements: Two sets of coordinates for each GCP: image (i, j) and map (x, y).
- Application: Paired coordinates allow derivation of transformation coefficients for geometric rectification.
- Registration: Aligning spatial data images to an Earth-based coordinate system.
- Ensures spatial consistency among multiple layers.
- Misalignment Correction: Uses GCPs; requires accuracy analysis for effectiveness.
RMS Error and GCP Criteria
- RMS Error: A metric for assessing the accuracy of geometric rectification.
- Aim: Establish GCPs with high accuracy, widespread distribution, and sufficient quantity for reliable transformation.
- First-Order Affine Transformation: Deals with moderate distortions.
- Parameters: Six transformations (translation, scale, skew, rotation).
- Input-to-Output (Forward) Mapping: Processes pixel values from the input grid to a rectified output image; can sometimes lead to pixel value loss.
- Output-to-Input (Inverse) Mapping: Preferred for rectification, ensuring every pixel location receives a value.
Spatial Interpolation Logic
- Objective: Populate a standard map projection with values from a distorted input image.
- Root Mean Squared Error provides logistical feedback on the coefficients used in the geometric adjustment.
Intensity Interpolation Methods
- Nearest Neighbor Resampling: Assigns the closest brightness value from the input to the output pixel location.
- Bilinear Interpolation: Averages brightness from four nearest pixels based on weighted distances to the desired position.
- Cubic Convolution: Similar to bilinear but incorporates 16 surrounding pixels for increased accuracy.