Finding Simple and Camouflaged Objects in Noise and Natural Backgrounds

项目总结/摘要 在上一个项目期间，我们开发并测试了识别 在噪声和自然背景中遮挡目标和深度边界（先前的目标1和2）。 目前的目标是为了扩展这些目标，我们以前在视觉搜索方面的工作， 研究隐蔽和公开的视觉搜索添加剂和遮挡目标。初步 研究，我们制定了一个原则性的理论，隐蔽和公开的搜索，然后测试的理论 在简单噪声背景下的隐蔽搜索任务中，我们发现， 搜索性能接近最佳，考虑到整个视野的可检测性（d'图） 的人类观察者。然而，令人惊讶的是，可检测性在 中央凹区域，但不是在周边。换句话说，存在实质性的“中央凹忽视”， 当秘密搜查被指向一个扩展区域时发生。我们建议将 测试这种现象及其在隐蔽搜索中的作用的原则性假设， 自然背景在初步研究中，我们还设计了刺激，分离的作用， 可用于识别遮挡目标的边界线索。在这些刺激中，目标和 背景具有统计上相同的表面纹理。我们将开发和测试最佳的 和次优模型的识别这些最大的目标，无论是噪音 和自然纹理。最后，我们计划测量和建模公开的搜索精度和眼睛， 运动的添加剂和最大限度地扩大闭塞目标。

Project Summary/Abstract In the previous project period we developed and tested principled theories for identification of occluding targets and depth boundaries in noise and natural backgrounds (previous Aims 1 & 2). The current aims are directed at extending these aims, and our previous work on visual search, to the study of covert and overt visual search for additive and occluding targets. In preliminary studies, we developed a principled theory of covert and overt search, and then tested the theory in a covert search task with additive targets in simple noise backgrounds. We found that overall search performance was near optimal, given the detectability across the visual field (the d' map) of the human observers. However, surprisingly, detectability was substantially reduced in the foveal region, but not in the periphery. In other words, there is substantial “foveal neglect” that occurs when covert search is directed over an extended area. We propose to characterize and test principled hypotheses for this phenomenon and its role in covert search both in simple and natural backgrounds. In preliminary studies we also designed stimuli that isolate the role of the boundary cues available for identification of occluding targets. In these stimuli, the target and background have surface textures that are statistically the same. We will develop and test optimal and suboptimal models of identification of these maximally-camouflaged targets, both for noise and natural textures. Finally, we plan to measure and model overt search accuracy and eye movements for additive and maximally-camouflaged occluding targets.