Detection, Classification and Boundary Regularization of Buildings in Satellite Imagery Using Faster Edge Region Convolutional Neural Networks

Abstract

With the development of effective deep learning algorithms, it became possible to achieve high accuracy when conducting remote sensing analyses on very high-resolution images (VHRS), especially in the context of building detection and classification. In this article, in order to improve the accuracy of building detection and classification, we propose a Faster Edge Region Convolutional Neural Networks (FER-CNN) algorithm. This proposed algorithm is trained and evaluated on different datasets. In addition, we propose a new method to improve the detection of the boundaries of detected buildings. The results of our algorithm are compared with those of other methods, such as classical Faster Region Convolution Neural Network (Faster R-CNN) with the original VGG16 and the Single-Shot Multibox Detector (SSD). The experimental results show that our methods make it possible to obtain an average detection accuracy of 97.5% with a false positive classification rate of 8.4%. An additional advantage of our method is better resistance to shadows, which is a very common issue for satellite images of urban areas. Future research will include designing and training the neural network to detect small buildings, as well as irregularly shaped buildings that are partially obscured by shadows or other occlusions.

Deep Learning Assisted Fixed Wireless Access Network Coverage Planning

“…Building boundary regularization is proposed in [42] to refine the building footprint predictions from Mask R-CNN. Faster Edge Region CNN (FER-CNN) is proposed in [43] to improve the building detection results particularly for buildings with irregular shapes. The work in [44] has presented a bounding box rotation method for Mask R-CNN to improve the precision of building extraction from Google Earth images.…”

EKSTRAKSI FITUR BANGUNAN SECARA OTOMATIS PADA CITRA SATELIT RESOLUSI TINGGI MENGGUNAKAN PENDEKATAN DEEP LEARNING BERBASIS MASK R-CNN (Studi Kasus di Kota Ambon)

“…Hal tersebut menunjukkan bahwa untuk memperoleh poligon bangunan yang sudah terpisah secara individual dalam 1 NLP lebih cepat dibandingkan melalui digitasi secara manual. Pengaturan (regularize) bentuk bangunan dilakukan untuk merapikan poligon-poligon bangunan yang terbentuk dari piksel-piksel citra hasil ekstraksi bangunan (Reda & Kedzierski, 2020;Zhao et al, 2018). Dapat dilihat bahwa baik pada AoI 1 (Gambar 4) maupun AoI 2 (Gambar 5) bentuk poligon bangunan hasil ekstraksi belum sepenuhnya rapi, dan setelah dilakukan regularize bentuk bangunan menjadi lebih tegas dan realistis menyerupai bentuk hasil digitasi skala 1:5.000.…”

Application of deep-learning techniques to very-high-resolution satellite images supporting population censuses in developing countries

“…18,19 Yet, none of these methods provide an unambiguous answer as to the precise location of buildings, let alone their properties. 16 Due to the advantages of accurate building detection in VHR satellite images, work on building detection models has accelerated considerably since the introduction of neural networks. These research efforts have grown even more dynamically since 2013, when convolutional neural networks (CNNs) were first introduced into image processing.…”