The perceptual processing of fused multi-spectral imagery

Abstract

Multi-spectral imagery can enhance decision-making by supplying multiple complementary sources of information. However, overloading an observer with information can deter decision-making. Hence, it is critical to assess multi-spectral image displays using human performance. Accuracy and response times (RTs) are fundamental for assessment, although without sophisticated empirical designs, they offer little information about why performance is better or worse. Systems factorial technology (SFT) is a framework for study design and analysis that examines observers’ processing mechanisms, not just overall performance. In the current work, we use SFT to compare a display with two sensor images alongside each another with a display in which there is a single composite image. In our first experiment, the SFT results indicated that both display approaches suffered from limited workload capacity and more so for the composite imagery. In the second experiment, we examined the change in observer performance over the course of multiple days of practice. Participants’ accuracy and RTs improved with training, but their capacity limitations were unaffected. Using SFT, we found that the capacity limitation was not due to an inefficient serial examination of the imagery by the participants. There are two clear implications of these results: Observers are less efficient with multi-spectral images than single images and the side-by-side display of source images is a viable alternative to composite imagery. SFT was necessary for these conclusions because it provided an appropriate mechanism for comparing single-source images to multi-spectral images and because it ruled out serial processing as the source of the capacity limitation.

Characterizing the time course of decision-making in change detection.

“…shown to work well in model recovery tests using this statistic (Fox & Houpt, 2016). Finally, we also constructed bootstrapped confidence intervals for the SIC to assess the uncertainty in the SIC.…”

“…We, therefore, adopted a less conservative cut-off of α = 0.33. This value has been used in previous research (Blunden et al, 2020) and has shown to work well in model recovery tests using this statistic (Fox & Houpt, 2016). Finally, we also constructed bootstrapped confidence intervals for the SIC to assess the uncertainty in the SIC.…”

Switch It Up! How Context Influences the Efficiency of Redundancy Gains in a Peripheral Task

“…We leveraged existing work to assume independent and parallel processing (i.e., the race model) of basic redundant visual stimuli (e.g. Eidels et al, 2015;Fox & Houpt, 2016) in our peripheral task, and estimated the capacity coefficient using a novel time-varying parametric model. We found that, contrary to earlier reports (Morey et al, 2018b), capacity to detect peripheral redundant targets was impaired by higher primary task demands, although this was modulated by the type of task load manipulation (i.e., the addition of, or the increased difficulty of, a primary tracking task).…”

Testing selective influence directly using trackball movement tasks

“…D + is the most positive point of SIC and D − is the most negative point of SIC. We chose alpha = .33 not the conventional critical value .05 here (Fox & Houpt, 2016) as the null hypothesis is SIC = 0 for all values of RT and MIC = 0 and hence conservative alpha levels bias the tests toward indicating a serial OR signature (flat SIC and zero MIC). Figure 4: Survival functions of RT and SIC for S1, S7, S10, S13, and S14 in Experiment 1.…”

“…In the shape reproduction task, there was no apparent correspondence between the trackball move and the change of the shape: Each trackball move was transformed to the change of the amplitudes, defined by function (9). The horizontal and vertical coordinates of each shape was defined by the transformation from amplitudes (8). The transformation functions were designed for reasons.…”