Human Performance in Vehicle Recognition with Visual and Infrared Images from Unmanned Aerial Vehicle

Recognition of vehicles and other objects is important in military operations. Unmanned aerial vehicles (UAVs) equipped with different sensors can provide images from which a human operator can recognise vehicles. Two experiments were carried out to study human performance in recognising vehicle types in visual and thermal long-wavelength infrared images respectively. The effect of vehicle model, stimuli size, and response time limit was studied in both experiments. In the experiments, trained laypeople classified images including different vehicle models, six military and one civilian. The results show that the vehicle model had an effect on both the classification accuracy and the response time. The results for the civilian vehicle stood out in particular as it was classified more accurately and faster than the military vehicles. Using larger stimuli and not limiting the response time generally led to higher classification accuracy and longer response times. There were also interactions between the variables, where for example different vehicle models were affected differently by stimuli size and limiting the response time. The results provide an overall indication of what level of performance can be expected for human vehicle recognition and how vehicle type, stimuli size, and time pressure effect performance.