美国油井正样本数据表数据集(2015–2022年)

孤井是指其运营方未知或已破产的油井。仅在美国，就有数十万口此类油井位置仍然不明。为了应对相关环境问题，亟需具备成本效益的定位技术。本文介绍了一个数据集，其中包含 120,948 幅近期记录的孤井航拍影像。每幅 512 × 512 像素的图像都配有分割掩膜，用于标注井的存在与否。这些图像来自 国家农业影像计划（NAIP），覆盖美国本土，空间分辨率在 30 厘米至 1 米之间。此外，我们还通过在美国范围内均匀选点，加入了负样本。数据集还附带元数据，包括原始影像的 ID 和空间分辨率（可通过 美国地质调查局 USGS 免费获取），以及这些影像中已记录孤井的像素坐标。本数据集旨在支持深度学习模型的开发，以便通过航拍影像定位未记录的孤井，从而减轻其造成的环境损害。Orphaned wells are oil wells whose operators are unknown or have gone bankrupt. Across the United States alone, hundreds of thousands of such wells remain with unknown locations. To address associated environmental issues, cost-effective localization technologies are urgently needed. This paper presents a dataset consisting of 120,948 recently captured aerial images of orphaned wells. Each 512 × 512 pixel image is paired with a segmentation mask that annotates the presence or absence of wells. These images are sourced from the National Agricultural Imagery Program (NAIP), covering the contiguous United States, with spatial resolutions ranging from 30 centimeters to 1 meter. Additionally, we added negative samples by uniformly selecting points across the United States. The dataset also includes metadata, including the original image ID and spatial resolution (freely accessible via the United States Geological Survey, USGS), as well as the pixel coordinates of documented orphaned wells in these images. This dataset aims to support the development of deep learning models for localizing unrecorded orphaned wells using aerial imagery, thereby mitigating the environmental damage caused by such wells.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于识别美国'孤儿井'（废弃油井）的航空图像数据集，覆盖2015-2022年，包含120,948张512×512像素的正负样本图像，空间分辨率为0.3-1.0米，图像来自美国国家农业影像计划（NAIP），并配有分割掩码标注油井位置。其目的是支持深度学习模型开发，以高效定位未记录的废弃油井，从而帮助解决相关环境问题。