An Efficient Building Extraction Method from High Spatial Resolution Remote Sensing Images Based on Improved Mask R-CNN

Abstract

In this paper, we consider building extraction from high spatial resolution remote sensing images. At present, most building extraction methods are based on artificial features. However, the diversity and complexity of buildings mean that building extraction methods still face great challenges, so methods based on deep learning have recently been proposed. In this paper, a building extraction framework based on a convolution neural network and edge detection algorithm is proposed. The method is called Mask R-CNN Fusion Sobel. Because of the outstanding achievement of Mask R-CNN in the field of image segmentation, this paper improves it and then applies it in remote sensing image building extraction. Our method consists of three parts. First, the convolutional neural network is used for rough location and pixel level classification, and the problem of false and missed extraction is solved by automatically discovering semantic features. Second, Sobel edge detection algorithm is used to segment building edges accurately so as to solve the problem of edge extraction and the integrity of the object of deep convolutional neural networks in semantic segmentation. Third, buildings are extracted by the fusion algorithm. We utilize the proposed framework to extract the building in high-resolution remote sensing images from Chinese satellite GF-2, and the experiments show that the average value of IOU (intersection over union) of the proposed method was 88.7% and the average value of Kappa was 87.8%, respectively. Therefore, our method can be applied to the recognition and segmentation of complex buildings and is superior to the classical method in accuracy. up objects of different sizes, then extract them by using the unique spectral information, shape and texture features of the buildings [3]. Due to different shooting angles, light and other factors, remote sensing images in different periods have considerable internal variability, and buildings have a variety of structure, texture and spectral information, therefore, the methods above cannot perform well in the extraction of complex buildings.In recent years, deep learning technology has ushered in a new wave of revival. At present, deep learning technology represented by deep convolutional neural networks has achieved excellent results in the field of computer vision [4][5][6]. Compared with the traditional method of feature extraction in artificial design, deep convolutional neural networks can obtain the structure, texture, semantics and other information of the object through multiple convolutional layers, and their performance is closer to visual interpretation in object recognition. The experiments in [7,8] had the advantages of deep convolutional neural networks in the field of object detection, but also revealed the problems of local absence and edge blur in image segmentation. This phenomenon is especially serious when the hardware equipment is insufficient or the dataset is small. Figure 1 shows this phenomenon using a mask image [9]. The main reasons for poor...

U-Net-Id, an Instance Segmentation Model for Building Extraction from Satellite Images—Case Study in the Joanópolis City, Brazil

“…The Joanópolis building dataset is also complex; for example, it can have two instances under the same roof, which represents a reality in the field but a difficult or impossible case to resolve by the algorithm. Because the most recent works using Mask R-CNN to segment building instances are applied to different datasets [11,12], unfortunately, accuracies are not compatible with our results and comparison of the different methods on the same datasets will be made in a future work.…”

“…For example, the Mask R-CNN algorithm, which is probably the most widely used method nowadays, consists of two distinct modules: the first module is Faster R-CNN [10], which attributes a label to the object and generates the bounding box to encapsulate the object, while the second module produces a mask of the object [6]. The most recent published works focusing on building instance segmentation have used Mask R-CNN and shown that it performs poorly for edge extraction and to preserve the integrity of the building instances [11,12]. The second method is based on the prediction of the the mask characteristics to further segment the instance in a posterior step.…”

Building Extraction from High Spatial Resolution Remote Sensing Images via Multiscale-Aware and Segmentation-Prior Conditional Random Fields

“…Combined with edge detection technology, convolutional neural networks can effectively deal with the recognition and segmentation of complex buildings [28].…”

Comparison of Different U-Net Models for Building Extraction from High-Resolution Aerial Imagery

“…Ji and Lu (2019) proposed a novel FCN based Siamese architecture and tested the architecture using their dataset which contains images from different sources. Zhang et al (2020) improved the building extraction efficiency of wellknown deep learning model Mask R-CNN using Sobel edge detection algorithm.…”