At present, convolutional neural networks (CNN) have been widely used in building extraction from remote sensing imagery (RSI), but there are still some bottlenecks. On the one hand, there are so many parameters in the previous network with complex structure, which will occupy lots of memories and consume much time during training process. On the other hand, low-level features extracted by shallow layers and abstract features extracted by deep layers of artificial neural network cannot be fully fused, which leads to an inaccurate building extraction from RSI. To alleviate these disadvantages, a dense residual neural network (DR-Net) was proposed in this paper. DR-Net uses a deeplabv3+Net encoder/decoder backbone, in combination with densely connected convolution neural network (DCNN) and residual network (ResNet) structure. Compared with deeplabv3+net (containing about 41 million parameters) and BRRNet (containing about 17 million parameters), DR-Net contains about 9 million parameters; So, the number of parameters reduced a lot. The experimental results for both the WHU Building Dataset and Massachusetts Building Dataset, DR-Net show better performance in building extraction than other two state-of-the-art methods. Experiments on WHU building data set showed that Intersection over Union (IoU) increased by 2.4% and F1 score increased by 1.4%; in terms of Massachusetts Building Dataset, IoU increased by 3.8% and F1 score increased by 2.9%.
29 February 2024 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
The extraction of buildings from High-resolution Remote Sensing Imagery (HRSI) is crucial across various applications and stands as a pivotal task in the field of remote sensing. While recent methods based on convolutional neural networks exhibit superior performance in building extraction from HRSI, there are still challenges such as incomplete and missing extractions of buildings especially the building boundaries and the small buildings. To address these issues, we propose a Supervised Attention U-Net (SAU-Net), which combines a well-designed encoder and decoder. In the encoder, we incorporate a novel Residual Channel Attention Block (RCAB) and a Densely Connected Multidilated Convolutional Block (DMCB) to enhance semantic features in the channel and spatial dimensions, respectively. Additionally, in the decoder, we design a Supervised Attention Block (SAB) which reconstructs fine-grained semantic features by systematically refining features in a supervised way and efficiently integrating feature maps from both the encoding and decoding stages. The efficacy of SAU-Net is evaluated using four HRSI datasets encompassing varying scenarios. The experimental results highlight that SAU-Net exhibits superior performance in building extraction, particularly excelling in the extraction of building boundaries and small buildings.