A Practical Method for Blind Pixel Detection for the Push-Broom Thermal-Infrared Hyperspectral Imager

Area of Science:

• Remote Sensing
• Hyperspectral Imaging
• Infrared Technology

Background:

• Thermal infrared hyperspectral imagers are crucial for land/ocean temperature inversion and environmental monitoring.
• Infrared focal plane arrays often contain blind pixels, distorting or invalidating data.
• Existing blind pixel detection methods are unsuitable for push-broom imagers due to their unique scanning mechanism.

Purpose of the Study:

• To develop a practical blind pixel detection and repair method for push-broom thermal infrared hyperspectral imagers.
• To address limitations of current methods that are not compatible with push-broom imaging principles.
• To improve the quality and applicability of thermal infrared hyperspectral data.

Main Methods:

• A two-stage detection and repair method for four common blind pixel types: dead, dark current, blinking, and noise pixels.
• Stage 1: Spectral filter detection (dead/dark current pixels), spatial noise detection (noise pixels), and response slope detection (dark current pixels).
• Stage 2: Line-by-line spectral filter detection to address randomly appearing blinking pixels.

Main Results:

• The proposed method accurately detects and repairs dead, dark current, blinking, and noise pixels.
• Spectral anomalies caused by blind pixels were effectively eliminated.
• Image quality was significantly improved, as validated by flight tests using the Airborne Thermal-Infrared Hyperspectral Imaging System (ATHIS).

Conclusions:

• The developed method is effective for blind pixel detection and repair in push-broom thermal infrared hyperspectral data.
• This advancement overcomes limitations of existing methods for push-broom imagers.
• The improved data quality enables more reliable applications in environmental monitoring and temperature inversion.