德清县屋顶结构遥感识别数据

用于实现对屋顶结构位置、类型及布局的高效精准识别。自动识别遥感影像中屋顶结构的精确坐标，便于确定屋顶结构的位置与分布，为后续屋顶结构分析、光伏规划提供基础。依据算法自动对屋顶结构进行分类并验证，准确识别屋顶类型（如平屋顶、坡屋顶等）并确保识别无误，实现自动识别与智能分类，显著减少人工现场勘查与视觉判断的时间。屋顶结构遥感监测识别数据录入系统中，实现数据可视化。基于无人机航拍技术高效采集2024年9月的屋顶结构的遥感影像数据，通过YOLO算法进行目标检测，识别出屋顶结构区域。首先，将单元神经网络应用于完整的遥感影像，将图像分割成19x19的单元格，每个单元神经网络负责预测K个单元格。预测每个区域的概率，所有单元格上具有最大概率的类被选择并分配给特定的网格单元，生成区域中心坐标(x,y)，采用CGCS2000坐标系，坐标默认东经、北纬。在预测类概率后，进行NMS运算，消除不必要的锚点。抽取验证样本计算识别准确度，越接近1表示越准确，高于0.6表示识别正确，低于0.6表示识别错误。依据识别准确度，判定目标是否为屋顶。算法依据样本库自动生成类型，并匹配到当前项目识别类型：斜屋顶、平屋顶、钢结构厂房等。对识别正确的目标进行纳入操作，对识别错误的目标进行排除操作。结合镇、村行政区划边界矢量，进行空间包含关系的计算，将识别数据与行政区划边界进行比较，判断要素是否位于当前行政区划的内部，若否，则判断下一个行政区划，直到确定所有识别数据的行政区划。最后将坐标、识别类型、镇、村等数据自动上传至德清县屋顶结构识别平台，最终获得德清县屋顶结构识别数据。

This dataset is designed for efficient and accurate recognition of the position, type and layout of roof structures. It automatically identifies the precise coordinates of roof structures in remote sensing images, facilitating the determination of their positions and distributions, and providing a foundation for subsequent roof structure analysis and photovoltaic planning. It automatically classifies and verifies roof structures via algorithms, accurately recognizes roof types (e.g., flat roofs, sloped roofs, etc.) and ensures recognition accuracy, realizing automatic recognition and intelligent classification, which significantly reduces the time required for on-site manual surveys and visual judgments. It is integrated into the remote sensing monitoring and recognition data entry system for roof structures, enabling data visualization.

Remote sensing image data of roof structures were efficiently collected in September 2024 via UAV aerial photography technology, and target detection was conducted through the YOLO algorithm to identify roof structure regions. Firstly, the cell neural network is applied to the full remote sensing image, which is split into 19×19 grid cells, and each cell neural network is responsible for predicting K cells. The class probability of each region is predicted, and the class with the highest probability across all grid cells is selected and assigned to a specific grid cell, generating the regional central coordinates (x, y). The China Geodetic Coordinate System 2000 (CGCS2000) is adopted, with coordinates defaulting to east longitude and north latitude.

After predicting the class probabilities, Non-Maximum Suppression (NMS) operation is performed to eliminate redundant anchor boxes. Validation samples are extracted to calculate the recognition accuracy. The closer the value is to 1, the higher the accuracy; a value above 0.6 indicates correct recognition, while a value below 0.6 indicates incorrect recognition. Whether a target is a roof structure is determined based on the recognition accuracy. The algorithm automatically generates types based on the sample database and matches them to the recognition types for the current project: sloped roofs, flat roofs, steel-structured factory buildings, etc. Targets with correct recognition are included in the dataset, while those with incorrect recognition are excluded.

Combined with the vector boundaries of town and village administrative divisions, spatial inclusion relationship calculation is performed. The recognized data are compared with the administrative division boundaries to determine whether the elements are within the current administrative division. If not, the next administrative division is checked until the administrative division of all recognized data is determined. Finally, data such as coordinates, recognition types, towns and villages are automatically uploaded to the Deqing County Roof Structure Recognition Platform, and finally the roof structure recognition data of Deqing County are obtained.

搜集汇总

数据集介绍

特点

德清县屋顶结构遥感识别数据集包含511条数据，通过无人机航拍和YOLO算法识别屋顶结构的位置、类型及布局，用于屋顶结构分析和光伏规划。