深度卷积网络支持下的遥感影像井盖部件检测

数字城市管理发展中城市部件调查是一项重要的任务，但是城市井盖部件信息获取存在人工调绘效率低、精度难以保证等缺陷，影响城市井盖部件的及时更新。因此本文利用深度卷积神经网络模型，通过小卷积核、尾部裁剪和保持输入大小等改进边缘检测网络（HED）并增加两层卷积运算提取目标，提出HED-C网络模型，实现了端到端的井盖部件目标检测。试验结果表明，利用HED-C模型井盖部件召回率可达96.58%，查准率可达97.93%，相较Faster R-CNN、YOLO和SSD网络模型，综合性能有了较大提高。

Manhole cover object detection in remote sensing imagery with deep convolutional neural networks

Urban component survey is an important task in the development of digital city management. However, the manhole cover information acquisition still has shortcomings such as low efficiency of manual surveying and high leakage rate. To address these problems, this paper proposes an effective method for detecting manhole cover objects in remote sensing images. We redesign the feature extractor by adopting VGG (visual geometry group) and HED (holistically-nested edge detection) side-output module, which can increase the variety of receptive field size. Then, the detection is performed by a multi-level convolution matching network for object detection based on fused feature maps, which combines several feature maps that enables small and densely packed manhole cover objects to produce stronger response. The results show that the proposed method is more accurate than existing methods for detecting manhole cover in remote sensing images.