Abstract. Artificial illumination identification within images is a useful tool for many applications. Performing such identification allows for an estimation of the illumination source spectrum, which in turn can be used for additional applications ranging from spectral detection and exploitation to statistics about nighttime light usage. Illumination identification has been performed in laboratory settings but not from an unmanned aerial vehicle (UAV) platform. Here, we test the feasibility of using a UAV and commercial off-the-shelf multispectral imaging sensor to perform such artificial illumination identification through linear discriminant analysis using nighttime UAV images. The results are very promising, showing source classification accuracies of 83.3%, 92.3%, 100%, and 100% for the incandescent, light-emitting diode, high pressure sodium, and metal halide illumination sources, respectively. We show that the information gained from the source identification can be further used to inform additional analysis, such as spectral identification. The high resolution of UAV imaging techniques combined with the knowledge of the illumination source can lead to better exploitation of such nighttime data for many applications.

中文翻译：

---

无人机人工照明识别

摘要。图像内的人工照明识别是许多应用程序的有用工具。执行这样的识别允许估计照明源光谱，这又可以用于从光谱检测和利用到关于夜间光使用的统计的其他应用。照明识别已在实验室环境中进行，但未在无人机 (UAV) 平台上进行。在这里，我们测试了使用无人机和商用现成多光谱成像传感器通过使用夜间无人机图像的线性判别分析来执行此类人工照明识别的可行性。结果非常有希望，白炽灯、发光二极管的光源分类准确率分别为 83.3%、92.3%、100% 和 100%，分别是高压钠灯和金属卤化物照明源。我们表明，从源识别中获得的信息可以进一步用于为其他分析提供信息，例如光谱识别。无人机成像技术的高分辨率与照明源的知识相结合，可以更好地利用此类夜间数据进行许多应用。