IR image databases generation under target intrinsic thermal variability constraints

Jerome Gilles,Stephane Landeau,Tristan Dagobert,Philippe Chevalier,Christian Bolut

2024-11-12

Abstract:This paper deals with the problem of infrared image database generation for ATR assessment purposes. Huge databases are required to have quantitative and objective performance evaluations. We propose a method which superimpose targets and occultants on background under image quality metrics constraints to generate realistic images. We also propose a method to generate target signatures with intrinsic thermal variability based on 3D models plated with real infrared textures.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to generate an infrared image database for Automatic Target Recognition (ATR) evaluation. Specifically, the authors propose a method of superimposing targets and occluders in the background to generate realistic infrared images, and introduce the inherent thermal variability of targets to enhance the authenticity and diversity of the database.

### Main problems:
1. **Requirement for large - scale high - quality infrared image databases**: In order to conduct quantitative and objective performance evaluations of ATR algorithms, a large number of high - quality infrared image databases are required.
2. **Limitations of existing methods**:
- **High cost and time - consuming in actual collection**: Obtaining such a database is costly and time - consuming.
- **Limitations of scene simulators**: Although scene simulators can be used to generate images, the calculation time is long and it is difficult to comprehensively cover various operational scenarios.

### Solutions:
1. **Mixed - scene generation method**:
- Generate realistic infrared images by superimposing real targets and occluders on a real background.
- Use image quality metrics (such as Local Contrast RSS, Quantity of Detectability QD, Signal - to - Clutter Ratio SCR, etc.) to control the quality of the output images.

2. **Consideration of the inherent thermal variability of targets**:
- Propose a method based on 3D models and real infrared textures to simulate the thermal characteristics of targets in different operating states.
- Generate intermediate thermal states from states between the ambient temperature (\(T_A\)) and the operating temperature (\(T_F\)) by interpolation:
\[
T_{IR}=(1 - \lambda_R)T_{AR}+\lambda_{TF R}
\]
where \(\lambda\in[0, 1]\) represents the thermal variability rate.

### Method overview:
- **Mixed - scene generation**:
- Locate occluders and targets.
- Apply the calculated gain and offset to the histograms of each region.
- Apply sensor effects (such as Modulation Transfer Function MTF and noise).

- **Thermal variability processing**:
- Divide the vehicle surface into regions with uniform thermal behavior (such as the engine, body, exhaust pipe, window, tire, etc.).
- Select different thermal configurations according to the operating mode and generate textures in intermediate thermal states.
- Generate 2D views from different perspectives by projection to increase the number of signatures in the database.

### Results and conclusions:
- Infrared images of the same vehicle with multiple thermal configurations are generated, demonstrating the effectiveness of the method.
- Add the newly generated signatures to the target database for mixed - scene generation.
- This method can generate realistic infrared scene images containing the inherent thermal variability of targets, thereby better evaluating and optimizing ATR algorithms.

### Summary:
This paper proposes a new method that can consider the inherent thermal variability of targets when generating an infrared image database, thereby improving the authenticity and diversity of the database and providing a more effective tool for the evaluation of ATR algorithms.