深度感知 · Depth Perception

🧠训练原理Training Principles

深度知觉（Depth Perception）是指从二维视网膜图像中恢复三维世界结构的能力。尽管我们的两只眼睛各自只接收到一个二维投影，但大脑通过整合多种深度线索（Depth Cues），成功地重建了三维空间。Depth perception is the ability to recover three-dimensional world structure from two-dimensional retinal images. Although each of our two eyes receives only a 2D projection, the brain successfully reconstructs 3D space by integrating multiple depth cues.

深度线索可以分为两大类：Depth cues can be divided into two major categories:

单眼线索Monocular Cues

• 相对大小：同样大小的物体，视网膜上成像越小，距离越远Relative size: Objects of the same size appear smaller on the retina when farther away
• 遮挡：一个物体遮挡另一个物体，被遮挡的更远Occlusion: An object that occludes another is closer
• 纹理梯度：纹理密度随距离增加而增大Texture gradient: Texture density increases with distance
• 线性透视：平行线在远处汇聚Linear perspective: Parallel lines converge in the distance
• 光影：亮暗分布暗示物体的三维形状Shading: Brightness-darkness distribution implies 3D shape

双眼线索Binocular Cues

• 视网膜像差：两只眼睛看到的图像有微小差异Retinal disparity: Images seen by the two eyes have slight differences
• 集合角度：眼球向内转动的角度Convergence angle: The angle of inward rotation of the eyes

本训练主要利用相对大小这一单眼线索——7 个同色光点中，只有 1 个是近目标，略大。你的任务是通过比对相对大小找出异常者。This training primarily uses the monocular cue of relative size — among 7 same-colored dots, only 1 is the near target, slightly larger. Your task is to identify the outlier by comparing relative sizes.

"Stereopsis and Depth Perception.""Stereopsis and Depth Perception." — Howard IP & Rogers BJ (2012). Oxford University Press.

Howard & Rogers (2012) 在其经典著作中指出，深度知觉不是单一能力，而是多种线索整合的结果。大脑会根据每种线索的可靠性（reliability）自动分配权重——当某种线索不可靠时（如在黑暗中），大脑会更依赖其他可用的线索。Howard & Rogers (2012) noted in their classic work that depth perception is not a single ability but the result of integrating multiple cues. The brain automatically assigns weights based on each cue's reliability — when one cue is unreliable (e.g., in darkness), the brain relies more on other available cues.

🎯训练模式Training Modes

在深空背景下，7 个同色光点散布屏幕，编号 1-7。其中只有 1 个是近目标——略大。Against a deep space backdrop, 7 same-colored dots are scattered on screen, numbered 1-7. Only 1 is the near target — slightly larger.

难度递进Progressive Difficulty

答对时，差异度阶梯递减：On correct answers, the difference decreases stepwise:

• 第 1 阶：差异度 3%（近目标半径 82.4px vs 远目标 80px）Level 1: 3% difference (near target radius 82.4px vs far target 80px)
• 第 2 阶：差异度 1%（80.8 vs 80.0）Level 2: 1% difference (80.8 vs 80.0)
• 第 3 阶：差异度 0.25%（80.2 vs 80.0）Level 3: 0.25% difference (80.2 vs 80.0)
• 第 4 阶：差异度 0.2%（80.16 vs 80.0）Level 4: 0.2% difference (80.16 vs 80.0)
• 第 5 阶：极限 0.01%（80.008 vs 80.000）Level 5: Extreme 0.01% (80.008 vs 80.000)

答错时重置为滑块设定的初始值。On wrong answer, reset to the initial value set by the slider.

3D 模式3D Mode

点击 [3D] 或按 M 键切换 3D 深度模式。此时除了相对大小外，还增加了深度透视线索——近目标看起来更"突出"，远目标看起来更"凹陷"。3D 模式模拟了真实世界中的多线索整合场景。Click [3D] or press M to toggle 3D depth mode. Besides relative size, depth perspective cues are added — the near target appears more "prominent," far targets more "recessed." 3D mode simulates multi-cue integration scenarios in the real world.

⚙️关键参数Key Parameters

📅推荐训练计划Recommended Training Plan

💡关键要点Key Points

1. 比对全部 7 个光点——异常者通过相对大小显现Compare all 7 dots — the outlier reveals itself through relative size
2. 尺寸是最主要线索——3D 模式增加深度透视Size is the primary cue — 3D mode adds depth perspective
3. diff < 0.2% 时所有目标几乎一样——需要极高的判别精度When diff < 0.2%, all targets look almost identical — requires extremely high discrimination precision
4. 选错会重置差异度——不要乱猜Wrong selection resets the difference — don't guess randomly
5. 每次训练后远眺放松 30 秒Look into the distance and relax for 30 seconds after each session

⚠️注意事项Precautions

• 训练中轻微眼疲劳属正常反应；若持续不适请立即暂停Mild eye fatigue during training is normal; stop immediately if discomfort persists
• 患有确诊眼疾者请遵医嘱使用Those with diagnosed eye conditions should use as directed by their physician
• 本工具为辅助性训练程序，不可替代专业医疗建议This tool is an auxiliary training program and cannot replace professional medical advice

⌨️键盘快捷键Keyboard Shortcuts

🔬科学延伸Scientific Insights

深度知觉训练在外科手术领域有重要应用。腹腔镜手术和机器人辅助手术要求外科医生在二维显示器上操作三维手术器械——这需要极强的深度知觉能力。研究表明，经过深度知觉训练的外科住院医师在模拟器手术中的表现显著优于未训练组（Rosser et al., 2007）。Depth perception training has important applications in surgey. Laparoscopic and robot-assisted surgery require surgeons to operate 3D instruments on 2D displays — this demands strong depth perception. Studies show that surgical residents trained in depth perception performed significantly better in simulator surgeries than untrained groups (Rosser et al., 2007).

对于日常驾驶，深度知觉直接影响跟车距离的判断和超车时机的选择。深度知觉较差的司机往往在高速公路上跟车过近，或在超车时误判对面来车的距离。深度知觉训练可以提升这些关键驾驶技能的准确性。For everyday driving, depth perception directly affects judgment of following distance and overtaking timing. Drivers with poor depth perception often follow too closely on highways or misjudge oncoming vehicle distances when overtaking. Depth perception training can improve accuracy in these critical driving skills.

另一个有趣的应用是虚拟现实（VR）。VR 设备通过向每只眼睛呈现略有不同的图像来模拟立体视觉。深度知觉训练可以帮助你更好地利用 VR 提供的立体线索，获得更沉浸的体验。Another interesting application is virtual reality (VR). VR devices simulate stereoscopic vision by presenting slightly different images to each eye. Depth perception training can help you better utilize the stereoscopic cues provided by VR for a more immersive experience.

🧬训练背后的神经可塑性Neuroplasticity Behind Training

深度知觉训练利用了视觉皮层 V2 和 V3 区的深度敏感神经元的可塑性。V2 区含有大量对视网膜像差敏感的神经元——它们只在两只眼睛看到的图像存在特定差异时才会放电。Depth perception training leverages plasticity of depth-sensitive neurons in visual cortex areas V2 and V3. Area V2 contains many neurons sensitive to retinal disparity — they fire only when images seen by the two eyes have specific differences.

通过反复在极小的尺寸差异下辨别近目标和远目标，V2 和 V3 区中的深度敏感神经元逐渐提高了它们的判别阈值——它们能够检测到更细微的尺寸差异。这种提高不仅限于训练中使用的大小差异范围——它还迁移到更大的差异范围，使你在所有尺度上的深度知觉都得到改善。By repeatedly discriminating near and far targets at extremely small size differences, depth-sensitive neurons in V2 and V3 gradually raise their discrimination thresholds — they can detect subtler size differences. This improvement extends beyond the trained size difference range — it also transfers to larger difference ranges, improving your depth perception at all scales.

此外，训练还强化了顶叶皮层的空间计算能力。顶叶皮层负责将视觉深度信息转化为空间坐标和行动指令——经过训练，你的大脑学会了更精确地计算物体的三维位置和运动轨迹。Additionally, training strengthens the spatial calculation ability of the parietal cortex. The parietal cortex converts visual depth information into spatial coordinates and action commands — after training, your brain learns to calculate 3D positions and movement trajectories of objects more precisely.