Change detection of urban objects using 3D point clouds: A review

ABSTRACT Data fusion has shown potential to improve the accuracy of land cover mapping, and selection of the optimal fusion technique remains a challenge. This study investigated the performance of fusing Sentinel-1 (S-1) and Sentinel-2 (S-2) data, using layer-stacking method at the pixel level and Dempster-Shafer (D-S) theory-based approach at the decision level, for mapping six land cover classes in Thu Dau Mot City, Vietnam. At the pixel level, S-1 and S-2 bands and their extracted textures and indices were stacked into the different single-sensor and multi-sensor datasets (i.e. fused datasets). The datasets were categorized into two groups. One group included the datasets containing only spectral and backscattering bands, and the other group included the datasets consisting of these bands and their extracted features. The random forest (RF) classifier was then applied to the datasets within each group. At the decision level, the RF classification outputs of the single-sensor datasets within each group were fused together based on D-S theory. Finally, the accuracy of the mapping results at both levels within each group was compared. The results showed that fusion at the decision level provided the best mapping accuracy compared to the results from other products within each group. The highest overall accuracy (OA) and Kappa coefficient of the map using D-S theory were 92.67% and 0.91, respectively. The decision-level fusion helped increase the OA of the map by 0.75% to 2.07% compared to that of corresponding S-2 products in the groups. Meanwhile, the data fusion at the pixel level delivered the mapping results, which yielded an OA of 4.88% to 6.58% lower than that of corresponding S-2 products in the groups.

https://www.tandfonline.com/doi/pdf/10.1080/10095020.2022.2035656?needAccess=true

Comparison of Layer-stacking and Dempster-Shafer Theory-based Methods Using Sentinel-1 and Sentinel-2 Data Fusion in Urban Land Cover Mapping

Building change detection using remote sensing images is significant to urban planning and city monitoring. The height information extracted from very high resolution (VHR) satellite stereo images provides valuable information for the detection of 3D changes in urban buildings. However, most existing 3D change detection algorithms are based on the independent segmentation of two-temporal images and the feature fusion of spectral change and height change. These methods do not consider 3D change information and spatial context information simultaneously. In this paper, we propose a novel building change detection algorithm based on 3D Co-segmentation, which makes full use of the 3D change information contained in the stereoscope data. An energy function containing spectral change information, height change information, and spatial context information is constructed. Image change feature is extracted using morphological building index (MBI), and height change feature is obtained by robust normalized digital surface models (nDSM) difference. 3D Co-segmentation divides the two-temporal images into the changed foreground and unchanged background through the graph-cut-based energy minimization method. The object-to-object detection results are obtained through overlay analysis, and the quantitative height change values are calculated according to this correspondence. The superiority of the proposed algorithm is that it can obtain the changes of buildings in planar and vertical simultaneously. The performance of the algorithm is evaluated in detail using six groups of satellite datasets. The experimental results prove the effectiveness of the proposed building change detection algorithm.

/pdf/building-change-detection-based-on-3d-co-segmentation-using-1hlpqfns.pdf

Building Change Detection Based on 3D Co-Segmentation Using Satellite Stereo Imagery

High-resolution synthetic aperture radar (SAR) images are mostly used in the current field of ship classification, but in practical applications, moderate-resolution SAR images that can offer wider swath are more suitable for maritime surveillance. The ship targets in moderate-resolution SAR images occupy only a few pixels, and some of them show the shape of bright spots, which brings great difficulty for ship classification. To fully explore the deep-level feature representations of moderate-resolution SAR images and avoid the “dimension disaster”, we innovatively proposed a feature fusion framework based on the classification ability of individual features and the efficiency of overall information representation, called maximum-information-minimum-redundancy (MIMR). First, we applied the Filter method and Kernel Principal Component Analysis (KPCA) method to form two feature subsets representing the best classification ability and the highest information representation efficiency in linear space and nonlinear space. Second, the MIMR feature fusion method is adopted to assign different weights to feature vectors with different physical properties and discriminability. Comprehensive experiments on the open dataset OpenSARShip show that compared with traditional and emerging deep learning methods, the proposed method can effectively fuse non-redundant complementary feature subsets to improve the performance of ship classification in moderate-resolution SAR images.

https://www.mdpi.com/1424-8220/21/2/519/pdf

A Maximum-Information-Minimum-Redundancy-Based Feature Fusion Framework for Ship Classification in Moderate-Resolution SAR Image.


 This paper intends to propose an automated deep learning-based land cover classification model of remote sensing images. The model includes (i) pre-processing, (ii) feature extraction and (iii) classification. The captured synthetic-aperture radar (SAR) and Landsat-8 images are initially pre-processed using the Gabor filtering model. Subsequently, from SAR images the gray-level-co-occurrence matrix-based texture characteristics are extracted, and temperature vegetation index-based characteristics, normalized vegetation index-based features, normalized difference index-based features and coloration index features are extracted from Landsat-8 images. Finally, the extracted features are subjected to an optimized deep belief network (DBN), where the weight is fine-tuned by the optimization logic. For this, a new Sunflower adopted Red Deer (SARD) algorithm is introduced in this work that hybrids the concept of Red Deer algorithm and Sunflower optimization. The performance of the proposed classification model is compared over other conventional models concerning different measures. Especially, the accuracy of the presented work (SARD+DBN) for Testcase3 is 5, 7, 6 and 30% better than existing DA + DBN, JA + DBN, SLnO+DBN and LA + DBN methods, respectively.

An Optimized Deep Belief Network for Land Cover Classification Using Synthetic-Aperture Radar Images and Landsat Images

Each of the urban land cover types has unique thermal pattern. Therefore, thermal remote sensing can be used over urban areas for indicating temperature differences and comparing the relationships between urban surface temperatures and land cover types. On the other hand, synthetic-aperture radar (SAR) sensors are playing an increasingly important role in land cover classification due to their ability to operate day and night through cloud cover, and capturing the structure and dielectric properties of the earth surface materials. In this research, a feature-level fusion of SAR image and all bands (optical and thermal) of Landsat 8 data is proposed in order to modify the accuracy of urban land cover classification. In the proposed object-based image analysis algorithm, segmented regions of both Landsat 8 and SAR images are utilized for performing knowledge-based classification based on the land surface temperatures, spectral relationships between thermal and optical bands, and SAR texture features measured in the gray-level co-occurrence matrix space. The evaluated results showed the improvements of about 2.48 and 0.06 for overall accuracy and kappa after performing feature-level fusion on Landsat 8 and SAR data.

Feature-Level Fusion of Landsat 8 Data and SAR Texture Images for Urban Land Cover Classification

Increasing human desire for urbanization and, as a result, urban growth and development has caused the replacement of natural objects such as vegetation and mountains with building structures and roads. This exacerbates the phenomenon of urban heat islands (UHI), one of the main threats to the quality of human life. In this research, a post-classification change detection algorithm is proposed based on UHI classification maps are generated using multi-temporal Landsat images from a rapidly developing urban area, in Iran in 17-years’ time interval. The main research objective is to investigate the effects of constructing new built-up areas on urban heat islands considering the land surface temperature, vegetation coverage, and surface object types. The production of UHI classification maps is based on the probabilistic reasoning rules for integrating the LST, Albedo, and NDVI as lines of evidence considering land use/land cover (LU/LC) object types. The analysis of the UHI change detection results in the rapidly developing urban area in this research indicates that the area percentage of built-up object class has increased by 12.1% in 17 years. This amount of construction led to a 23.7% increase in the high UHI class, which is equivalent to 2931 hectares. Moreover, the regions with medium and low UHI have decreased by − 16.09 and − 9.12%, respectively, which indicates an increase in the intensities of urban heat islands during this period.

Post-Classification Urban Heat Island Change Detection Based on Multi-Temporal Satellite Images

Natural disasters such as earthquakes and floods together with the urban sprawl conducted by increasing the population make multi-temporal changes in building areas. Destruction, buildings’ renovation, and constructing new buildings are the main changes of the urban areas that should be detected to update three-dimensional city models. The results of performing three-dimensional changes detecting of high altitude objects such as buildings are more close to reality than the two-dimensional methods. In this study, a three-dimensional changes detection method is proposed based on digital elevation models (DEMs). In the first step of this proposed method, the normalized digital surface model (nDSM) is generated for timely datasets. Then, object-based image analysis is utilized by performing segmentation followed by the structural classification of DEMs. Differencing and comparing the multi-temporal classification maps as the third step of the proposed algorithm led to analyzing the occurred changes. The obtained results are evaluated in an urban area in Tehran, Iran, in a 9-year time interval. These results represent −9.7% decreasing rate in low-rise buildings and also −1.37% decreases in the ground. Moreover, the class of high-rise buildings increased for +16.4% which conforms to making new constructions in addition to the renovation of low-rise buildings. According to the area analyzing of the changes, 4.8% of the investigated study area has new constructions, 3.05% has buildings’ renovation, and 3.89% has destruction in that 9-year period.

Fatemeh Tabib Mahmoudi

Papers

Feature-Level Fusion of Landsat 8 Data and SAR Texture Images for Urban Land Cover Classification

Post-Classification Urban Heat Island Change Detection Based on Multi-Temporal Satellite Images

Three-dimensional building change detection using object-based image analysis (case study: Tehran)