基于Mask R-CNN的高分遥感影像建筑物目标检测

Building Object Detection in High-Resolution Remote Sensing Image Based on Mask R-CNN

深度学习方法在目标检测和语义分割领域得到了广泛应用,但在遥感影像中,由于建筑物呈聚集型分布且目标之间间隔紧密,建筑物目标检测暂未取得较好的效果。针对上述问题,提出一种基于Mask R-CNN的高分辨率遥感影像建筑物目标检测方法,将边界框识别与像素级语义分割结合起来,较好地解决了聚集分布且间隔紧密的建筑物目标检测问题。实验结果表明,该方法具有较高的检测精度。

Deep learning methods have been widely used in the field of object detection and semantic segmentation.However,due to the clustered distribution of buildings and close intervals between targets in remote sensing images,building object detection has not yet achieved good results.To solve the above problems,we propose a detection method for building objects in high-resolution remote sensing images based on Mask R-CNN,which combines bounding box recog⁃nition with pixel-level semantic segmentation to solve the problem of detecting building objects that are clustered and closely spaced.The experimental results show that the pro⁃posed method has high detection accuracy.