Proposing Gaze-Based Interaction and Automated Screening Results for Visual Aerial Image Analysis

Visual aerial image analysis occurs in various application domains. Recently, high-resolution optical sensors became more and more available. This results in more captured image data and increased workload for the human expert image analysts. In this contribution, we address the situation of a human operator performing a visual screening task in aerial images in order to find certain vehicles. This task is common in the safety and security domain. Based on discussion with expert image analysts about typical image analysis challenges, we implemented an experimental system in order to investigate whether gaze-based interaction and automated image exploitation would provide an appropriate and efficient user interface. The system provides (1) screening results generated from an automated screening algorithm and (2) gaze-based interaction using a low-cost remote eye-tracker and a keyboard for panning/zooming the image as well as for selecting/framing of targets. The results of a pilot study (N = 12 non-expert image analysts) show, that the availability of automated screening results reduces error rates, completion time and perceived workload (NASA-TLX). Ratings of the gaze-based interaction using the ISO 9241–411 questionnaire are good to very good, except for eye strain for users with glasses.