Recently, crowdsourced geographic data have provided a cost-effective approach to learn deep convolutional neural networks (DCNNs) for road extraction from remote-sensing images. However, crowdsourced datasets often suffer from label noise and include error labels that can affect the performance of DCNN-based road extraction methods. Thus, we propose a novel sequence deep learning (SDL) framework that introduces label probability sequence to correct error labels and robustly learn DCNNs for road extraction. The label probability sequence is obtained by front-end developed DCNNs, which can provide valuable true label information to alleviate the adverse effect of label noise. Specifically, to pursue high effectiveness of the label correction process, an adaptive label correction method is proposed to exploit the distance between the true label distribution hidden in label probability sequence and the possible true label distribution. In addition, based on the uncertainty of label probability sequence, a noise correction loss function is introduced to robustly learn DCNNs on the improved dataset produced by the adaptive label correction. By repeating these two steps in each epoch, we can learn efficiently DCNN-based road extraction methods under noisy datasets. Experimental studies on two real-world noisy road datasets show that the proposed method can help to overcome label noise issue for DCNNs and achieve excellent performance in road extraction.

Abstract Road extraction from high‐resolution remote sensing images (HRSIs) has great importance in various practical applications. However, most existing road extraction methods have considerable limitation in capturing long‐range shape feature of road, and thus, they are ineffective in extracting road region under complex scenes. To address this issue, a novel model called long‐range context‐aware road extraction neural network (LR‐RoadNet) is proposed. LR‐RoadNet takes advantage of strip pooling to capture long‐range context from horizontal and vertical directions, aiming to improve continuity and completeness of road extraction results. Specifically, the LR‐RoadNet consists of two parts: strip residual module (SRM) and strip pyramid pooling module (SPPM). The SRM is built based on residual unit, in which the strip pooling is employed to learn general and long‐range road feature from input image. Then, the SPPM is used to obtain long‐range feature from multiple scales by multiple parallel strip pooling operations. More importantly, a structural similarity (SS) loss function is introduced to further explore road structure for optimizing LR‐RoadNet. The experimental results show that the proposed method achieves great improvement than other state‐of‐the‐art methods on three challenging datasets, Cheng‐Roads, Zimbabwe‐Roads and Mass‐Roads.