Thermal Infrared Arrays for Human Presence Perception Systems in Highly Autonomous Agricultural Machinery.

Highlights: Thermal infrared array sensors can detect simulated human presence at distances up to 3m. Simulated human models for testing thermal infrared arrays need to be heated. Testing models should be designed to incorporate clothing, PPE, and features that change heat distribution. It is critical to match the pixel observable area and the expected size of heated regions at critical distances.

Abstract: This project consists of two sets of experiments using low-cost thermal infrared arrays-the TPAM 166 L3.9 array and the AMG8833, which were operated as 16-pixel arrays. These sensors were tested to determine how well they detected a simulated human model. The TPAM 166 L3.9 was tested with a heated, water-filled, clothed model representing a standing adult human. The AMG8833 was tested with a heated and non-heated version of the ISO 18497 simulated human model, which represented a seated human. These sensors were able to respond to heated models at distances up to 3 m from the sensor. The unheated human model did not produce a response. For heated models, the strength of the detection increases at shorter distances if the warmest parts of the model were visible to the sensor. As the warm parts of the model approached the sensors, the number of pixels impacted and the temperatures that they detected increased. However, detection decreased as the simulated human model approached at distances less than 1.5 m if the sensor was focused on a clothed section of the model. For deployment, the pixel observable area should match the expected size of heated regions considering PPE, clothing, and operational considerations. It is critical to consider pixel observable area and pixel response levels rather than the sensor level characteristics, as detection will occur at the pixel level.