The Categorical Distinction Between Targets and Distractors Facilitates Tracking in Multiple Identity Tracking Task

Abstract

This study investigates the tracking facilitation effect during categorical distinction between targets and distractors in the Multiple Identity Tracking task. We asked observers to track four targets in a total of eight moving objects, and manipulated categorical distinctions of targets and distractors across four experiments, with different combinations of inter-category and intra-category differences. Results show that tracking performance was significantly better when the targets and distractors were inter-category different, compared to when the targets and distractors were identical or intra-category distinctive. As the inter-category distinction between targets and distractors narrowed, tracking performance improved, but the inter-category facilitation effect decreased. These results may indicate a category-based grouping effect: the observers organized the targets within the same semantic category into one group and made the targets more easily and accurately rediscovered when lost during tracking. Furthermore, the tracking facilitation of categorical distinction diminished when all the objects were inverted. This proved that besides their visual distinctiveness, objects’ semantic category information also played an important role during tracking.

Multiple-target tracking in human and machine vision

“…The effect of semantic information. Semantic difference between target and distractor categories facilitates MIT [25]. In [47], authors attribute this facilitation to the categorical distinction of targets and distractors, supported by four processes.…”

“…Although the fields of neuroscience and computer-vision believe in the same framework for mechanisms underlying MTT, there are many differences in the approaches used in these two areas. For example, studies suggest that any visual separation between targets and distractors, which can be due to surface features or semantic information, can streamline perceptual grouping and improve the performance of subjects during tracking [25][26][27][28]; however, no computer-vision algorithm has utilized this finding as inspiration behind its methods. This could be incorporated by processing semantic and surface information of targets in an object recognition module in MTT algorithms, which may improve tracking performance.…”

“…Still more findings from MTT cognitive studies are worth noting for the design of algorithms. For example, it is reported that any surface features or semantic information that leads to visual separation between targets and distractors can improve the performance of subjects during tracking [25][26][27][28]. However, no MTT algorithm in computer vision has utilized this finding as inspiration for its methods.…”

The semantic category-based grouping in the Multiple Identity Tracking task

“…That study showed a consistent categorical-distinction advantage in tracking performance, even when the categorical distinction was narrow. Most importantly, the study demonstrated that the tracking facilitation effect was caused not only by visual distinctiveness between the targets and distractors, but also by their semantic-based categorical difference (Wei et al, 2016). However, the mechanism of this categorical-distinction advantage was still unclear.…”

“…Thus, visual object knowledge can be organized into different categories innate in reality (Caramazza & Shelton, 1998;Mahon & Caramazza, 2011). In a recent study (Wei, Zhang, Lyu, & Li, 2016), we used typical and nameable objects to examine the influence of objects' category information on tracking performance. We manipulated the categorical distinction between targets and distractors; specifically, the targets belonged to one category (e.g., animals) and the distractors to another (e.g., furniture).…”