[Updated: April 16] — Embedded Linux is combating the coronavirus in thermal imaging devices such as the Kogniz’ Health Cam and a Raspberry Pi based FluSense device that analyzes coughs. We also look at open source ventilator projects and products from Aaeon, Advantech, and Seeed.

AI surveillance devices that use facial recognition and other technologies to identify and track people give us the creeps and make us wonder about the future of civil liberties. But you know what else gives us the creeps? The novel coronavirus. In recent weeks we’ve heard about a variety of embedded Linux devices that are being put to work scanning people for thermal signatures that might suggest a high fever.

Below, we examine some thermal imaging technologies for detecting Covid-19 that are confirmed to run Linux, including Kogniz’ Jetson Xavier based Health Cam and a Raspberry Pi based FluSense device that incorporates a Myriad X accelerator and a ReSpeaker mic array for detecting coughs. But first we’ll take a brief tour of some other Covid-19 related projects and products that involve Linux, open source, or embedded tech in general.

Open qPCR

Opensource.com recently reported on some open source hardware projects that tackle Covid-19. The one confirmed Linux-based product is Chai’s BeagleBone-based Open qPCR system, which we covered last month. The system is being adapted to test surfaces for Covid-19 from swab samples. The story also mentioned an interesting Opentrons OT-2 open source lab automation platform that is being adapted for testing humans for Covid-19.

Several open source projects are addressing the vital need for ventilators. These devices typically don’t run Linux and can typically be constructed from 3D printed parts. [Update: But sometimes they do: Tom’s Hardware has reported that the Raspberry Pi Foundation is ramping up production of the Raspberry Pi Zero to meet rising demand for the tiny, $5 ARM11-based SBC from ventilator manufacturers.]

The Opensource.com story mentions two such projects: an OpenLung Low Resource Ventilator equipped with a bag valve mask (BVM) and a Pandemic Ventilator. Meanwhile, the Montreal General Hospital Foundation announced a Code Life Ventilator Challenge to build a low-cost, easy-to-use ventilator that can be produced locally. Prizes total $200K Canadian. Other open source ventilator projects include OS-Vent, the Open Source COVID19 Medical Supplies community, and MIT’s E-Vent.

On the software side, SUSE, the German company behind SUSE Linux, is offering free support and maintenance for medical device companies using SUSE Linux Enterprise Server (SLES) and related container technologies to address Covid-19. SLES is also being used by pharma companies like Roche and Merck to develop Covid-19 test kits and vaccines.

[Update: Yesterday, MiniNodes published a story that was later picked up by CNXSoft that says the [email protected] citizen science project currently focusing on Covid-19 vaccine research has expanded its list of computers that can donate compute cycles to Arm-based computers. You can load the BOINC remote load-sharing software used by [email protected] and other similar projects on the Raspberry Pi 4, Nvidia Jetson Nano, Rockchip RK3399-based computers like the Rock64, and other 64-bit Arm-based SBC’s that have 2GB of RAM.]

 
Kogniz Health Cam

This week, Kogniz announced a Linux-based Kogniz Health platform and Kogniz Health Cam designed to scan people entering a facility and identify anyone with an elevated temperature. Kogniz claims the solution is more accurate, affordable, and faster than posting staff with handheld thermal devices at entrances to take individual temperatures.

Kogniz Health Cam (left) andJetson AGX Xavier Development Kit
(click images to enlarge)

The camera can scan groups of people walking through an entrance at up to 16 feet away. It adjusts for ambient temperatures and focuses on the area around the eyes for the most accurate reading.

Kogniz Health is based on its general-purpose Kogniz real-time recognition platform, which is used for tasks such as surveillance, bag and weapon identification, and license plate readers. The system runs Linux on Nvidia’s high-end Jetson AGX Xavier Development Kit.

The Xavier module features 8x ARMv8.2 cores and a high-end, 512-core Nvidia Volta GPU with 64 tensor cores and 2x Nvidia Deep Learning Accelerators. There’s also a 7-way VLIW vision chip, 16GB LPDDR4, and 32GB eMMC.

The Xavier system controls a Kogniz Health Cam equipped with an infrared thermal imager with a 384 x 288 microbolometer and a 6.8mm lens. There’s also an 8-megapixel optical imager. The IP54 protected camera has a -10 to 50°C operating range and can detect temperatures between 0 and 50°C.

The Kogniz Health Cam’s Xavier module works the Linux-based Kogniz AI cloud platform to perform facial recognition to match high-temperature individuals with image databases. Results can be read immediately on an integrated screen or remotely via the Kogniz mobile app and web interface. Alerts can be sent out immediately over SMS, mobile push, and Slack.

Kogniz Health is already being used by a dozen customers, says Kogniz. The company is working on a new version with a Jetson AGX Xavier module mounted on a custom carrier board instead of the Xavier Dev Kit. Pricing for a single camera and a year’s subscription to Kogniz AI start at less than $10,000.

 
FluSense

Of the many academic research projects that have emerged to address different parts of the coronavirus problem, UMass Amherst’s FluSense captured our attention. The Raspberry Pi based device not only has a thermal camera for detecting high temperatures, but also a mic array to detect signature coughing sounds. A cough modeling system can match those with various medical conditions.

FluSense components (left) and UMass Amherst developers Forsad Al Hossain and Tauhidur Rahman with the FluSense prototype
(click images to enlarge)

As the name suggests, the portable FluSense device, which was announced on the official Raspberry Pi blog, was originally intended for detecting people with the flu. However, it can be used for Covid-19 or other conditions involving coughing and high temperatures.

FluSense uses AI analytics, but unlike Kogniz Health and other similar Advantech and Seeed solutions described farther below, it’s not linked to a facial recognition data base and not intended to catch potential super spreaders. Instead it acts as a health monitoring tool designed to be deployed in health clinic waiting room and public spaces to help determine the number of potential flu or Covid-19 victims. The aggregate data can assist health authorities in predicting and tracking outbreaks, as well as planning travel restrictions and deploying medical supplies.

FluSense prototype

FluSense has a cloud component, but most of the processing is done on the Linux-driven Raspberry Pi and an attached Intel Neural Compute Stick 2 equipped with a Movidius Myriad X AI accelerator. The system also integrates Seeed’s Rockchip RK3229 based ReSpeaker Core v2.0 microphone array and Seek Thermal’s Seek CompactPro thermal camera, which features 320 x 240 resolution and a 32 degree field of view.

By comparing the number of coughs detected by the mic array and analyzed by the AI algorithm with a count of the number of people in the room, FluSense can make an estimate of the number of victims. The thermal imaging camera augments this data by capturing the heat signature of a room minute by minute. The data is heavily encrypted and only aggregate data is saved to ensure privacy.

The device is still in undergoing testing, but will hopefully be manufactured and deployed in sufficient numbers to play a role in Covid-19 tracking. More information may be found in the FluSense academic paper.

 
A 5G Patrol Robot and a remote AI temp screener

Several major embedded tech vendors have promoted customer design wins that use their computers for identifying potential Covid-19 carriers or otherwise assist in related medical care. A story posted by China Daily and later picked up by IoT Business News and other publications describes a rather frightening “5G patrol robot” from Guangzhou Gosuncn Robot Co. that wanders public spaces in search of potential Covid-19 carriers. The device is “equipped with 5 high-resolution cameras and infrared thermometers capable of scanning the temperature of 10 people simultaneously within a radius of 5 meters,” explains IoT Business News. “If a high temperature or the absence of a mask is detected, the robots send an alert to the relevant authorities.”

5G patrol robot (source: China Daily) and Advantech’s MIC-770
(click images to enlarge)

The robot is controlled by Advantech’s MIC-770 embedded PC, which runs Linux or Windows on an Intel 8th Gen Core Coffee Lake processor. The rugged, full-featured MIC-770 offers optional FlexIO and iDoor modules for additional display interfaces and other I/O. The 5G Patrol Robot uses a GPU i-Module to accelerate AI processing.


Seeed’s Remote AI Temperature Screening System and UI
(click images to enlarge)

As reported by CNXSoft, Seeed has opened pre-orders for a $7,000 Remote AI Temperature Screening System that recognizes faces while simultaneously measuring their temperature. The system can detect temperatures at up to 3 meters.


Remote AI Temperature Screening System detail view
(click image to enlarge)

The Seeed system has an HD camera for facial recognition and a temperature sensor with a 20 to 45°C sensing range. No OS or processor is listed, but CNXSoft speculates it might be the Rockchip RK3399, which would likely run Linux.

 
Aaeon computers used in X-ray machine and telepresence robots

Aaeon recently announced three products billed as being useful in fighting the coronavirus, although some are more relevant to the pandemic than others. First, Aaeon announced that unnamed Chinese medical equipment manufacturers have ordered nearly 1,000 of its 5th Gen Broadwell based Boxer-6638U embedded PC for use in unnamed mobile X-ray machines. It’s unclear what OS is used on the robot, although the Boxer-6638U runs Linux or Windows. The mobile devices are used to perform chest X-ray scans of suspected COVID-19 patients.

Aaeon’s Boxer-6638U (left) and Boxer-8110AI
(click images to enlarge)

Aaeon also announced on Embedded Computing Design that its Nvidia Jetson TX2 powered Boxer-8110AI computer is being used for a hospitality service robot built by an unnamed vendor for providing room service in hotels. Aaeon’s somewhat tenuous pitch is that the robot can help avoid direct personal contact that might spread the virus. One problem with the coronavirus angle is that the hotel industry is projected to see a 30 percent drop in hotel occupancy over the next year due to Covid-19.

Ohmni Supercam

Aaeon also announced that its Intel Cherry Trail Atom based UP board is powering OhmniLabs’ Ohmni Supercam Telepresence Robot. As we all know, videoconferencing has received a big boost from the pandemic, and Aaeon says the robot can also be used for low-resolution medical telepresence.

Aaeon informs us that the $2,195 Ohmni Supercam runs OhmniOS, a custom operating system based on an Android fork. It’s equipped with a 10-inch screen and a 4K, 13-megapixel, wide-angle camera. The system’s UP core can be equipped with a Myriad X-based AI Core X module for adding AI processing. The foldable, 4.8-foot tall mobile robot has a 5-hour battery.