Registration of a Dynamic Multimodal Target Image Test Set for the Evaluation of Image Fusion Techniques
Abstract
In Phase 1 (2011-12) of this study we successfully collected both static and dynamic multimodal nighttime imagery of military relevant scenarios in the field. In addition, we registered and fused this imagery using our newly developed color image fusion schemes. In Phase 2 (2012-13) of this study we used the image sets previously collected in Phase 1 to assess the operational effectiveness of different sensor modalities and image fusion schemes for the enhancement of observer SA in complex (urban) scenarios. This knowledge is essential for the effective operational deployment of night vision sensors in civilian and military surveillance scenarios, and for the further development of new image fusion and representation schemes. The capability of the different imaging modalities to enhance a user s SA was investigated in three different experimental paradigms: Experiment 1: Testing the ability of observers to quickly grasp the gist of the scenes; Experiment 2: Investigating eye movements of observers inspecting the images; and Experiment 3: Estimating the perceived depth in the scenes. In Experiment 1 we determined how much information observers can pick up from brief image presentations (in a glance). Observers briefly (for 500ms) viewed either daytime color photographs, or LWIR, II, and color fused nighttime imagery of a given set of military relevant scenarios. After seeing each image they reported all details they had perceived in the scene. We quantified observer performance through the precision and recall measures (precision is the fraction of detections that are true positives, while recall is the fraction of true positives that are detected).The results of the first experiment show that both precision and recall are higher for daytime photographs and fused colorized nighttime imagery, compared to LWIR or II imagery.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 17, 2013
- Accession Number
- ADA598370
Entities
People
- Alexander Toet