A Multispectral Image Creating Method for a New Airborne Four-Camera System with Different Bandpass Filters

Area of Science:

• Remote Sensing
• Photogrammetry
• Optical Engineering

Background:

• High-resolution multispectral imaging is vital for detailed environmental and geological analysis.
• Accurate band-to-band registration is a significant challenge in multispectral data fusion, especially between visible and near-infrared spectra.

Purpose of the Study:

• To develop and validate an automatic method for composing multispectral data from an airborne four-camera system.
• To address the registration challenges between different spectral bands, particularly in natural environments lacking distinct man-made features.

Main Methods:

• Utilized a four-camera monochrome system with interchangeable bandpass filters.
• Implemented a homography registration model employing Scale-Invariant Feature Transform (SIFT) and Random Sample Consensus (RANSAC).
• Developed a novel registration approach based on system structural characteristics for challenging visible-to-near-infrared image registration.

Main Results:

• Achieved automatic composition of multispectral data.
• Demonstrated effective registration between visible and near-infrared bands even without man-made objects.
• Attained a band-to-band alignment error of less than 2.5 pixels for the composed images.

Conclusions:

• The proposed automatic data composing method enables the acquisition of high-quality multispectral images from the airborne system.
• The novel registration technique successfully overcomes limitations in fusing images from different spectral ranges.
• The system and method offer a robust solution for precise multispectral data fusion in remote sensing.