Prediction of the thermal imaging minimum resolvable (circle) temperature difference with neural network application

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Image Processing

Background:

• Thermal imaging is crucial for defense and civilian applications, offering 24/7 operational capability.
• Outdoor thermal imaging faces challenges like atmospheric decay, fog, and distance, degrading image quality and object recognition accuracy.
• Human visual recognition excels in normal conditions but is impaired by fatigue and environmental factors, affecting thermal imaging performance metrics like MRTD and MRCTD.

Purpose of the Study:

• To explore advanced thermal imaging recognition techniques.
• To develop a method for optimal characteristic value selection and comprehensive image processing.
• To apply neural network technology for enhanced thermal imaging recognition and prediction of MRTD/MRCTD.

Main Methods:

• Utilized neural network technology for image recognition.
• Developed a method for selecting characteristic values and processing thermal images.
• Applied the method to predict minimum resolvable temperature difference (MRTD) and minimum resolvable circle temperature difference (MRCTD).

Main Results:

• Neural networks successfully recognized thermal images.
• The developed method effectively processed images and selected characteristic values.
• The system demonstrated superior recognition capabilities compared to human observers under fatigue conditions.

Conclusions:

• Neural network technology offers a robust solution for thermal imaging recognition challenges.
• The proposed method enhances the accuracy of object recognition in adverse conditions.
• This approach overcomes the limitations of human visual perception in measuring thermal imaging performance metrics.