Thermal imaging has gone mainstream. In 2014, FLIR launched a thermal infrared camera so tiny, low-power and inexpensive that it has been integrated into a cellphone—the CAT S60. The cost point is achieved by using high-volume manufacturing processes and materials developed for mobile phone cameras, including packaging, calibration and assembly. Based on the smaller sizes and lower costs, micro-miniature, thermal cameras can now be used in applications that were unthinkable even a few years ago: non-contact thermometers integrated in smartphones, new gestural and thermal touch user interfaces, wearable spectrometers, IoT climate control and care devices, and people counting.
FLIR Lepton Micro-Thermal Camera
The Lepton is a micro-thermal imaging camera that produces a stream of longwave infrared (LWIR) images that are 80x60 pixels in size at a frame rate of <9 Hz. Realizing a thermal camera that is small enough to be integrated into a mobile-phone is the next step in a rapid evolution in infrared camera technology that has taken place over the last 15 years. FLIR has led to a dramatic reduction in the size, weight, power and cost of thermal imaging cameras, making them more accessible to manufacturers to integrate into devices. This trend is illustrated in Figure 1, which shows the reduction in the size of FLIR thermal camera cores that has taken place from 2010 to 2014.
Fig. 1. Thermal camera miniaturization.
Lepton is innovative in every aspect and required a significant investment in technology development. These are some of the key technology advances used in the design: very low cost lenses built in a novel high-volume wafer-level process, wafer-level packaging of sensors to reduce camera size and number of manufacturing steps, high-speed automated camera assembly and calibration with minimal touch labor, and single-chip camera electronics to reduce size, power requirements and number of interconnects.
"For IoT thermal devices, cloud computing offers the possibility of maintaining low local power dissipation, small installed footprint and low cost"
Thermal Sensors Everywhere
Because of the recent reductions in size, power and cost of thermal sensors, many never-before-possible high-volume applications are emerging. The Internet of Things (IoT) can be defined as the aggregation of all the sensing modules which are linked to the Cloud. Yole (Lyon, France) expects the total IoT market sensor volume to be 11.8 billion units by 2024.
Compact, low-power thermal sensors will be embedded in many IoT applications including smart appliances, thermostats and other distributed environmental control systems as well as presence detection and elder care monitoring. Each IoT application will use cloud computing and most will enable remote/mobile notifications. Low-cost thermal imagers can be used to detect fire and or floods and sensor arrays will become key enablers of increased energy efficiency by allowing building and outdoor lighting to be adjusted based on the presence of people and further by allowing optimized and localized climate control. Energy will be allocated to lighting and heating dynamically on an as- needed basis. In elder care applications, compact thermal imagers allow for activity, presence and fall detection in critical and hazardous areas such as bathrooms while preserving privacy with their low spatial resolution.
Fig. 2. Applications for thermal imagers in a smart-home.
In addition to smart homes and buildings, thermal imaging sensors can be used in industrial IoT applications such as people counting. Thermal imaging is one of the leading methods of counting people in challenging lighting conditions and outdoor settings. People counting and tracking is a relatively new capability offered to retail locations to determine how many people are shopping and what their habits are while shopping. People counting is also used for managing lines and wait times and making assessments of the total number of people and their movements in public areas like airports and large shopping malls. Using thermal imaging as the people counting method offers two advantages over visible-light imaging systems. Additionally, low-resolution thermal infrared does not generate personally identifiable video information and so allows for anonymous collection of statistics on people movement and presence detection.
Current people counting systems use visible-light two dimensional or three dimensional (two cameras per sensor) cameras or ultra-low resolution thermal sensors. These sensors are very inexpensive compared to the traditional thermal imaging modules that often have resolutions of 320x240 pixels, but can cost more than $1000.
The new generation of very small, low-cost, low-power microbolometer cameras offer cost advantages over traditional thermal imagers, while also delivering system and performance advantages over the ultra-low resolution systems deployed today.
Lepton’s 80x60 image size provides almost 19 times the number of pixels for detection compared to a 16x16 array. Depending on lens configuration and location of the people counting device, this provides the potential to more accurately count people over an area that is substantially larger than the current system without sacrificing resolution compared to current systems. In a situation where multiple people-counting systems might be required, now one system could do the job for a substantially lower net cost. Finally, a higher resolution device like Lepton could be configured to perform certain types of discrimination functions that a 16x16 sensor could not do. These discrimination tasks could include determining if a person is carrying an object, assessing whether objects are in a shopping cart and if they are at room temperature, or hot or cold.
Fig. 3. Image produced by a typical 16x16 pixel thermal imaging sensor used in people counting contrasted with a 80x60 Lepton image
Edge and Cloud Computing Optimization with Thermal Imaging
Thermal imaging is uniquely suited to assist in optimizing the tradeoffs between the need for low power, edge detection and computation and data transmission to the cloud. Next, initial edge processing can be performed to classify the event as either a false alarm or a significant event which triggers more sensors such as visible cameras in the edge device to power on and the transmission of video data to the cloud for further processing. For IoT thermal devices, cloud computing offers the possibility of maintaining low local power dissipation, small installed footprint and low cost. The device does not require the cost of an expensive processor while still accessing the functionality of high-end algorithms.
While it is impossible to predict the “killer IoT app” and ultimate volume drivers in advance, the rich array of thermal application possibilities show that the future of thermal sensors and imagers is very bright indeed.