Distributed Solar Photovoltaic Array Location and Extent Data Set for Remote Sensing Object Identification

Earth-observing remote sensing data, including aerial photography and satellite imagery, offer a snapshot of the world from which we can learn about the state of our environment, anthropogenic systems, and natural resources. The components of energy systems that are visible from above may be assessed with these remote sensing data when combined with machine learning methods. Here we focus on the information gap in distributed solar photovoltaic (PV) arrays, of which there is limited data on solar PV deployments at small geographic scales. We created a machine learning dataset to develop the process of automatically identifying solar PV locations through the use of remote sensing imagery.<br>This dataset contains the geospatial coordinates and border vertices for 19,433 solar panels across 601 high resolution images from four cities in California. Dataset applications include training object detection and other machine learning algorithms that use remote sensing imagery, developing specific algorithms for predictive detection of distributed PV systems, and analysis of the socioeconomic correlates of PV deployment.<br>Links to the aerial photographs from Fresno, Stockton, Oxnard, and Modesto can be found in the references.<br><i>Notes: this version of the dataset has been improved to increase the accuracy of polygon georeferencing so that the data can be more easily integrated with imagery other than than the original imagery from which the annotations were based. Additionally, a small number of polygons were found to be erroneous annotations and either corrected or removed.</i><i><br></i><i>Update July 30, 2020: </i><i>Panel area in pixels and area in square meters were labeled incorrectly in the original version; those labels were reversed. Updated files include:</i><i> 'polygonDataExceptVertices.csv', 'SolarArrayPolygons.geojson', 'SolarArrayPolygons.json'.</i>地球观测遥感数据（含航空摄影与卫星影像）可提供全球实景快照，助力我们洞悉环境、人为系统与自然资源的现状。结合机器学习方法，可利用此类遥感数据对从上空可见的能源系统组成部分开展评估。本研究聚焦分布式太阳能光伏（PV, solar photovoltaic）阵列领域的信息缺口：当前小地理尺度下的太阳能光伏部署相关数据极为有限。我们构建了本机器学习数据集，以研发通过遥感影像自动识别太阳能光伏阵列位置的方法流程。<br>本数据集涵盖美国加利福尼亚州4座城市的601张高分辨率影像中的19433块太阳能光伏板的地理空间坐标与边界顶点信息。该数据集的应用场景包括：训练基于遥感影像的目标检测及其他机器学习算法、研发用于分布式光伏系统预测性检测的专用算法，以及分析太阳能光伏部署与社会经济因素的相关性。<br>弗雷斯诺（Fresno）、斯托克顿（Stockton）、奥克斯纳德（Oxnard）以及莫德斯托（Modesto）的航空影像链接可参见参考文献。<br><i>备注：本数据集版本已进行优化，提升了多边形地理配准的精度，使得数据可更便捷地与标注所基于的原始影像之外的其他影像进行集成。此外，本次更新修正或移除了少量存在标注错误的多边形。</i><i><br></i><i>2020年7月30日更新：</i><i>原始版本中，光伏板像素面积与平方米面积的标注出现颠倒错误。已更新的文件包括：</i><i> 'polygonDataExceptVertices.csv', 'SolarArrayPolygons.geojson', 'SolarArrayPolygons.json'.</i>