The coastal zone is one of the most important features on the earth’s surface; therefore, it is imperative to extract the coastline, a representative coastal zone feature with high quality. Previously, related methods mainly focus on edge and small-scale information, when processing large scale images, misclassification can occur because it’s difficult to determine whether a local area belongs to land or sea. To solve this problem, in this study, a deep learning-based multiscale coastal line extraction algorithm is proposed, whose core is a scene classification-based multiscale coastal zone classifier to identify the coastal zones from low to high levels using quadtree decomposition. Compared to the conventional method, the proposed method can obtain information from a large receptive field so as to identify land and sea precisely in high resolution imagery. The results indicate that the proposed method can effectively eliminate confusing features, and is of high calculation speed.