基于高分辨率无人机影像的太阳能热水器自动检测算法研究  

Research on Automatic Solar Water Heater Inspection Algorithm Based on High Resolution UAV Images

作　　者：张志文 孙保金 徐岱 刘志东 ZHANG Zhi-wen;SUN Bao-jin;XU Dai;LIU Zhi-dong(Suzhou Building Construction Quality Inspection&Test Center Co.Ltd.,Suzhou Jiangsu 215000,China)

机构地区：[1]苏州市建设工程质量检测中心有限公司,江苏苏州215000

出　　处：《现代测绘》2024年第5期1-6,共6页Modern Surveying and Mapping

基　　金：江苏省建设系统科技项目(2022ZD089)

摘　　要：太阳能资源的利用情况对能源政策制定、实际碳减排量核算以及相关设备市场潜力评估具有重要意义。以城镇化率超过70%的苏州市城南花苑及其周边地区为研究区,利用卷积神经网络学习方法,结合高分辨率的无人机航片影像,构建了计算机深度学习识别太阳能热水器的数据集及检测模型,采用实例分割模型Mask R-CNN方法提取DOM目标地物,使用研究区的DSM-DEM高程信息建立目标筛选模型,开展城镇区域内太阳能热水器快速建立空间数据库的研究,探究该方法对于高效建立太阳能资源利用设备空间分布数据库的适用性。结果表明,目标检测数据成果召回率达到96.88%,精确率达到99.14%,目标检测效果较好。有助于快速建立大范围的太阳能利用设备的空间分布数据库,相对于传统人工清点的作业流程,具备高效率、低成本、数据全面、时效性强等优点。The utilisation of solar energy resources is important for energy policy formulation,accounting for actual carbon emission reductions and the assessment of the market potential of related equipment.In this study,the study area was taken as the study area of Ghengnan garden community of Suzhou city,which has an urbanisation rate of over 70%,and a computer deep learning dataset and detection model for identifying solar water heaters was constructed using a convolutional neural network learning method in combination with high-resolution aerial film images from UAVs.The Mask R-CNN method was used to extract DOM target features,and the DSM-DEM elevation information of the study area was used to establish a target screening model to carry out a study on the rapid establishment of a spatial database of solar water heaters in urban areas,and to explore the applicability of the method for the efficient establishment of a database of the spatial distribution of solar resource utilisation equipment.The results show that the target detection data results have a recall rate of 96.88%and an accuracy rate of 99.14%,which is a good effect of target detection.This study helps to quickly establish a spatial distribution database of solar energy utilisation equipment on a large scale,which has the advantages of being more efficient,low-cost,comprehensive data and time-sensitive compared to the traditional manual inventory operation process.

关 键 词：深度学习 无人机影像 太阳能热水器 数据优化模型 

分 类 号：P231[天文地球—摄影测量与遥感] TP273.5[天文地球—测绘科学与技术]