TL;DR
This paper introduces a novel, low-cost infrared thermal imaging system (MTIS) integrated with a mobile phone. By leveraging the ubiquitous nature and computing power of smartphones, the MTIS aims to make thermal imaging more accessible for civilian applications, particularly in healthcare, with a demonstrated detection distance of up to 29 meters.
What Problem Does This Paper Address?
Traditional thermal imagers, while powerful for non-destructive temperature measurement, suffer from high costs and limited portability, hindering their widespread adoption for civilian use. Existing low-cost thermal imagers often still require separate devices and data transfer methods, failing to fully utilize the capabilities of modern mobile devices. The authors address this by proposing a system that integrates a thermal imaging module with a smartphone, aiming to reduce cost, increase portability, and enhance user experience by utilizing the phone's processing and display capabilities.
Methods
The core of the MTIS is a Thermal Infrared Module (TIM) featuring a 64x62 thermopile IRFPA sensor. The TIM is designed with a simplified optical mechanism using a germanium lens with an anti-reflective coating and an f/1.0 aperture to optimize temperature sensitivity. The system architecture relies on a wireless network (Wi-Fi) connecting the TIM to mobile phone clients running a custom Android application called IRAPP.
IRAPP handles remote control, image display, and data processing. Key software components include:
- Non-Uniformity Correction (NUC): A two-point linear correction is employed, with an adaptive self-correction algorithm based on Artificial Neural Networks (ANN) for fine-tuning.
- Scene-Change Detection: This algorithm distinguishes between scenarios requiring "Coarse Correction" (two-point linear) and "Fine Correction" (ANN-based), optimizing efficiency and power consumption on the mobile device.
- Frame Averaging: A Last-In-First-Out (LIFO) method is used for frame averaging to improve image quality, especially given the detector's time constant.
Datasets
Not discussed in the paper. The experiments involved using a black body furnace and standard 4-bar patterns for calibration and testing.
Key Insights and Ideas
- Leveraging Mobile Computing Power: The paper highlights that smartphones possess sufficient computing power to handle thermal imaging processing, making them ideal platforms for integrated thermal imaging systems.
- Cost Reduction through Integration: By offloading processing and display to the mobile phone, the hardware cost of the thermal imager itself can be significantly reduced, making it more accessible.
- Pervasive Thermal Imaging: The integration of thermal imaging with a ubiquitous device like a smartphone opens up possibilities for widespread applications in daily life, particularly in areas like health monitoring.
- Adaptive NUC for Efficiency: The proposed two-tiered NUC approach (Coarse and Fine) with scene-change detection is crucial for optimizing performance and power consumption on a mobile platform.
Key Results
- Detection Distance: The MTIS demonstrated a maximum detective distance of approximately 29 meters.
- Image Quality Improvement: After applying correction algorithms, the imaging quality significantly improved, allowing for recognition of facial features and body poses.
- Recognition Performance: At distances up to 6 meters, the system showed good performance in recognizing body positions (sitting vs. crouched), with accuracy exceeding random guessing. For example, at 2 meters, sensitivity, specificity, and accuracy were all 1.000. At 6 meters, these values were 0.643, 0.500, and 0.571, respectively.
- NETD: The Noise Equivalent Temperature Difference (NETD) decreased with increased integration time, reaching 70.25 mK with 64 frames, which is comparable to commercial imagers.
- MRTD: The Minimum Resolvable Temperature Difference (MRTD) behavior was analyzed, contributing to the estimation of the maximum working distance.
- Cost Comparison: The MTIS is positioned as a low-cost alternative, with an estimated price under $500, significantly cheaper than comparable microbolometer-based imagers.
Limitations and Future Work
- Resolution: While adequate for many applications, the 64x62 pixel resolution might be a limitation for highly detailed imaging tasks.
- Future Applications: The authors suggest exploring further medical applications, particularly for infant healthcare, such as monitoring for Sudden Infant Death Syndrome (SIDS) by detecting body temperature and sleeping posture. They also plan to investigate more general daily life applications.
- Hardware Updates: The paper notes that as mobile phone technology advances, the MTIS hardware can be easily updated by simply replacing the phone.
Summary updated