感知特征距离：在特定决策规则下衡量动态条件下感知系统的稳定性与鲁棒性
Perception Characteristics Distance: Measuring Stability and Robustness of Perception System in Dynamic Conditions under a Certain Decision Rule
提出 PCD/aPCD 指标，将感知模型输出不确定性量化为可靠检测的最远距离，为视觉感知系统在动态天气与光照条件下的稳定性评估提供可复用度量；配套 SensorRainFall 多模态数据集（相机/雷达/激光雷达）可用于研究跨天气域泛化与不确定性感知。
原文依据 We therefore introduce the Perception Characteristics Distance (PCD), a novel metric incorporating model output uncertainty as represented by the farthest distance at which an object can be reliably detected.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论自动驾驶系统（ADS）的安全性取决于在不同距离和驾驶条件下感知的准确性。AI 感知算法的输出具有随机性，这对决策和安全性结果（包括碰撞时间估计）有重大影响。然而，当前的感知评估指标并未反映感知算法的随机性。因此，我们引入了感知特征距离（PCD），这是一种新颖的指标，将模型输出不确定性表示为物体能够被可靠检测到的最远距离。为了从可靠检测距离的角度表示系统的整体感知能力，我们使用在多个检测质量和概率阈值下平均的 PCD 值来产生平均 PCD（aPCD）。为了进行实证验证，我们提出了 SensorRainFall 数据集，该数据集在弗吉尼亚智能道路上使用配备传感器（相机、雷达和激光雷达）的车辆采集，车辆在不同天气（晴朗和雨天）和光照条件（日光、路灯和夜间）下受控行驶。该数据集包括目标物体的真实距离、边界框和分割掩码。对最优模型的实验表明，aPCD 能够捕捉到不同天气、日光和光照条件下的有意义的差异，而传统评估指标无法反映这些差异。PCD 提供了一种不确定性感知的感知性能度量，支持更安全、更稳健的 ADS 运行，而 SensorRainFall 数据集则为评估提供了有价值的基准。SensorRainFall 数据集可在 https://www.kaggle.com/datasets/datadrivenwheels/sensorrainfall 公开获取，评估代码可在 https://github.com/datadrivenwheels/PCD_Python 获取。
英文原文摘要 1801 字符
The safety of autonomous driving systems (ADS) depends on accurate perception across distance and driving conditions. The outputs of AI perception algorithms are stochastic, which has a major impact on decision making and safety outcomes, including time-to-collision estimation. However, current perception evaluation metrics do not reflect the stochastic nature of perception algorithms. We therefore introduce the Perception Characteristics Distance (PCD), a novel metric incorporating model output uncertainty as represented by the farthest distance at which an object can be reliably detected. To represent a system's overall perception capability in terms of reliable detection distance, we used the averaged PCD values across multiple detection quality and probabilistic thresholds to produce the average PCD (aPCD). For empirical validation, we present the SensorRainFall dataset, collected on the Virginia Smart Roads using a sensor-equipped vehicle (cameras, radar, and LiDAR) controlled under different weather (clear and rainy) and illumination conditions (daylight, streetlight, and night). The dataset includes ground-truth distances, bounding boxes, and segmentation masks for target objects. Experiments with state-of-the-art models show that aPCD captures meaningful differences across weather, daylight, and illumination conditions, which traditional evaluation metrics fail to reflect. PCD provides an uncertainty-aware measure of perception performance, supporting safer and more robust ADS operation, while the SensorRainFall dataset offers a valuable benchmark for evaluation. The SensorRainFall dataset is publicly available at https://www.kaggle.com/datasets/datadrivenwheels/sensorrainfall, and the evaluation code is available at https://github.com/datadrivenwheels/PCD_Python