Absolute IOP/EOP Estimation Models without Initial Information of Various Smart City Sensors

Area of Science:

• Geomatics Engineering
• Computer Vision
• Smart City Technology

Background:

• Smart cities extensively deploy optical sensors like CCTV and UAVs.
• Crucial sensor data (3D position, orientation, principal distance) is often missing.
• This limits the utility of collected visual data for precise mapping and analysis.

Purpose of the Study:

• To develop and evaluate methods for estimating optical sensor parameters without prior information.
• To determine the optimal model for calculating sensor position, orientation, and principal distance.
• To address the challenge of missing metadata in smart city optical sensing.

Main Methods:

• Utilized structured mobile mapping system point clouds for sensor calibration.
• Applied two Direct Linear Transformation (DLT) models and a perspective projection model.
• Calculated principal distance and estimated sensor position and orientation.
• Tested the stability and accuracy of estimation methods with real-world sensor data.

Main Results:

• The perspective projection model yielded the most accurate estimations for camera position and orientation.
• The original DLT model showed significant orientation estimation errors, possibly due to parameter correlation.
• Achieved position and orientation errors of 0.80 m and 2.55° respectively with the perspective projection model.
• Fixed-wing UAV application faced challenges due to ground control point placement issues.

Conclusions:

• The perspective projection model is superior for estimating optical sensor pose and principal distance in smart city contexts.
• Mobile mapping point clouds offer a viable solution for calibrating uncalibrated optical sensors.
• Further research is needed to overcome challenges in specific platforms like fixed-wing UAVs and ensure robust ground control.