• Innovators@JU

The cost effective aerial monitoring system for COVID-19 suspect detection

Updated: Mar 28, 2020

The outbreak of Corona virus has left more than 2,00,000 people sickened and causing more than 9000 deaths till the last report from the World Health Organization (WHO) .The spread of the viral disease is so severe that it has driven the powerful countries like China, Italy to lockdown completely. Simultaneously the medical service system of the affected countries is undergoing immense pressure of the patients and if the spread cannot be controlled then the medical systems may breakdown very soon. As WHO has already declared, flattening the spread curve is urgent and authorities have to take necessary steps to realize it. From the guidelines of WHO, it is now known that mild fever, cough, pneumonia are common symptoms of this disease and most of the COVID-19 positive patients are found to show these symptoms. As a safety measure, the people with fever, cough are also repeatedly requested to take the COVID-19 tests or at least maintain home quarantine for 10-14 days so as to restrict the spread of the virus. But unfortunately not all people are following these instructions correctly. To tackle this situation, cops can take help of the unmanned systems.

Robots and drones are already being used across the globe to fight against the Corona outbreak. That includes thermal camera embedded drones also to monitor people and identify the ones having elevated body temperature. But in reality this approach has its inherent limitations. Thermal image sensors generate heat maps to analyse the temperature distribution in that area. But it's not easy to detect the person having elevated temperature directly from the data of the thermal sensor. Moreover, thermal cameras are very expensive indeed.

We need a cost-effective alternative to it so that mass production and deployment across the country is possible. We at the UAV and Robotics Innovators' Group, Jadavpur University in Kolkata are working on a low-cost aerial monitoring device equipped with Infra Red thermometer and normal RGB camera that will identify COVID-19 suspects who have abnormal body temperature. As soon as the system will detect any person with abnormal temperature, it will send the image of the person to the control station as a result of which police can inquire the suspected person.

Infrared or IR thermometers have the capability to detect the temperature of the target object even from a long distance. This distance of operation depends on the Distance to Spot or DTS ratio of the thermometers. Suppose an IR thermometer has DTS 12:1. This implies that it can measure the temperature of a circular region of 1 feet diameter accurately from a distance of 12 feet. It means to measure human body temperature using this thermometer we can safely fly our monitoring drone at a height of 24-30 feet. Now, we are designing our system so that every time it detects the temperature of a single person and simultaneously the camera captures his/her image. Thanks to the flexibility offered by the 3-axis gimbals, we have been able to synchronize the temperature measurement and image acquisition operation. In this way it will form a database of the form (image of the person, the body temperature of the person). This data will immediately be fed as the input to our two-stage algorithm that detects if there is any abnormality in the temperature data. If the result is YES, the corresponding person's image is immediately sent to the control station using 4G communication network. The officials at the control station are supposed to quickly take action to inquire of that particular person to check if he/she is affected or not.

Designing the two-stage algorithm is a crucial step in our case to achieve accurate system performance. The two stage prediction algorithm has the following two levels -

Level 1: Error Correction segment

Level 2: Abnormality Detection segment

The IR thermometers are very good for non-contact temperature measurement but their drawback is that these devices produce accuracy in the range of +/- 2°C which is not acceptable for human temperatures. To address this issue, we are using the Error Correction segment in our algorithm. It uses a set of training data (I.e. temperature values) and models the error function which is further used to predict the error in the test data. It was assumed that the error function is dependent on the measured temperature range. This assumption has been validated using the test data-set collected by our team. This algorithm produces accurate temperature data based on the input data.

The corrected temperature data is then fed to the Abnormality Detection segment which identifies if the measured temperature is normal or elevated. The algorithm developed so far is based on a threshold temperature value. If it detects an elevated temperature, the corresponding person's image is sent to the control station. Using this image, the police can identify and track the suspected person.

This system is unique in its operational features. It not only detects a person with abnormal temperature but also allows the police to easily identify and track him. It costs only 60,000 which is far less than the existing thermal imaging drones. The estimated time required by this system to complete the task for one person is only 10 seconds that is very time efficient indeed.

In open public areas like railway stations, bus stops, markets where ground based monitoring is not possible to implement, we urgently require this aerial monitoring system to continuously monitor the crowd and manage the people. The wide coverage and mobility offered by the drones are essential for this purpose. Deployment of this aerial monitoring system will be helpful to restrict the spread of the virus and help avoid country lockdown and in turn the breakdown of the economy.

147 views0 comments