Triplet-Based Semantic Relation Learning for Aerial Remote Sensing Image Change Detection

Zhang, Mengya; Xu, Guangluan; Chen, Keming; Yan, Menglong; Sun, Xian

doi:10.1109/lgrs.2018.2869608

Cited by 193 publications

(116 citation statements)

References 13 publications

Supporting

Mentioning

116

Contrasting

Order By: Relevance

“…Traditional CD methods PBCD Bruzzone et al [7], Celik [8], Deng et al [9], Wu et al [10], Huang et al [11], Benedek et al [14], and Bazi et al [18] OBCD Ma et al [20], Zhang et al [21], Gil-Yepes et al [22], Qin et al [23] Deep learning CD methods FB-DLCD Sakurada et al [26], Saha et al [27], Hou et al [28], El Amin et al [29], Zhan et al [31], Zhang et al [32], Niu et al [33], and Zhan et al [34] PB-DLCD Gong et al [36], Arabi et al [38], Ma et al [40], Zhang et al [41], Khan et al [42], Daudt et al [44], Wang et al [45], Wiratama et al [46], Zhang et al [47], Mou et al [49], and Gong et al [50] IB-DLCD Lei et al [52], Daudy et al [53], Lebedev et al [54], and Guo et al [55] To address the above-mentioned issues, we proposed a novel end-to-end method based on improved UNet++ [58], which is an effective encoder-decoder architecture for semantic segmentation. A novel loss function was designed and an effective deep supervision (DS) strategy was implemented, which are capable of capturing changes with varying sizes effectively in complex scenes.…”

Section: Category Example Studiesmentioning

confidence: 99%

“…Nevertheless, DBN, unlike CNN, has weak feature learning abilities. Thus, Siamese CNN architectures with weighted contrastive loss [31] and improved triplet loss [32] were exploited to learn discriminative deep features between changed and unchanged pixels, then DIs were generated based on the Euclidean distances of deep features, and finally CM could be obtained by a simple threshold.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++

2019

View full text Add to dashboard Cite

Change detection (CD) is essential to the accurate understanding of land surface changes using available Earth observation data. Due to the great advantages in deep feature representation and nonlinear problem modeling, deep learning is becoming increasingly popular to solve CD tasks in remote-sensing community. However, most existing deep learning-based CD methods are implemented by either generating difference images using deep features or learning change relations between pixel patches, which leads to error accumulation problems since many intermediate processing steps are needed to obtain final change maps. To address the above-mentioned issues, a novel end-to-end CD method is proposed based on an effective encoder-decoder architecture for semantic segmentation named UNet++, where change maps could be learned from scratch using available annotated datasets. Firstly, co-registered image pairs are concatenated as an input for the improved UNet++ network, where both global and fine-grained information can be utilized to generate feature maps with high spatial accuracy. Then, the fusion strategy of multiple side outputs is adopted to combine change maps from different semantic levels, thereby generating a final change map with high accuracy. The effectiveness and reliability of our proposed CD method are verified on very-high-resolution (VHR) satellite image datasets. Extensive experimental results have shown that our proposed approach outperforms the other state-of-the-art CD methods.

show abstract

Section: Category Example Studiesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++

2019

View full text Add to dashboard Cite

show abstract

“…As shown in Equation (21), the Re scores will be improved, which may lead to a decrease in the Pr scores. To sum up, the model can get better CD results when the balance between the Pr and Re is achieved, which can be made by adjusting the parameter w p as shown in Equation (12).…”

Section: Experiments On Different Hyperparameters Of the Proposed Methodsmentioning

confidence: 99%

“…Different from homogeneous CD, the pixels in heterogeneous images are in different distinct feature spaces [7], and the change map (CM) cannot be obtained by simple linear operations or some homogeneous methods, which is also the main difficulty for heterogeneous CD. Over the past several decades, much attention has been paid to homogeneous CD [8], and many excellent methods have been explored [9][10][11][12][13]. With the increase of different types of satellite sensors, however, CD based on homogeneous images is far away from the practical demands [8] especially when the homogeneous images are not available.…”

Section: Introductionmentioning

confidence: 99%

A Feature Space Constraint-Based Method for Change Detection in Heterogeneous Images

et al. 2020

Self Cite

View full text Add to dashboard Cite

With the development of remote sensing technologies, change detection in heterogeneous images becomes much more necessary and significant. The main difficulty lies in how to make input heterogeneous images comparable so that the changes can be detected. In this paper, we propose an end-to-end heterogeneous change detection method based on the feature space constraint. First, considering that the input heterogeneous images are in two distinct feature spaces, two encoders with the same structure are used to extract features, respectively. A decoder is used to obtain the change map from the extracted features. Then, the Gram matrices, which include the correlations between features, are calculated to represent different feature spaces, respectively. The squared Euclidean distance between Gram matrices, termed as feature space loss, is used to constrain the extracted features. After that, a combined loss function consisting of the binary cross entropy loss and feature space loss is designed for training the model. Finally, the change detection results between heterogeneous images can be obtained when the model is trained well. The proposed method can constrain the features of two heterogeneous images to the same feature space while keeping their unique features so that the comparability between features can be enhanced and better detection results can be achieved. Experiments on two heterogeneous image datasets consisting of optical and SAR images demonstrate the effectiveness and superiority of the proposed method.

show abstract

“…DL-based change detection methods have been applied to different targets such as urban [46][47][48][49], land use/land cover [50][51][52], and landslides [53], among others. Peng et al [54] proposed a subdivision of DL-based change detection methods that considered three units of analysis: (1) feature [55][56][57]; (2) patch [58][59][60][61]; and (3) image [62,63]. In the case of image-based DL change detection, the algorithms learn the segmentation of changes directly from bi-temporal image pairs, avoiding the negative effects caused when using pixel patches [54].…”

Section: Introductionmentioning

confidence: 99%

Change Detection of Deforestation in the Brazilian Amazon Using Landsat Data and Convolutional Neural Networks

et al. 2020

View full text Add to dashboard Cite

Mapping deforestation is an essential step in the process of managing tropical rainforests. It lets us understand and monitor both legal and illegal deforestation and its implications, which include the effect deforestation may have on climate change through greenhouse gas emissions. Given that there is ample room for improvements when it comes to mapping deforestation using satellite imagery, in this study, we aimed to test and evaluate the use of algorithms belonging to the growing field of deep learning (DL), particularly convolutional neural networks (CNNs), to this end. Although studies have been using DL algorithms for a variety of remote sensing tasks for the past few years, they are still relatively unexplored for deforestation mapping. We attempted to map the deforestation between images approximately one year apart, specifically between 2017 and 2018 and between 2018 and 2019. Three CNN architectures that are available in the literature—SharpMask, U-Net, and ResUnet—were used to classify the change between years and were then compared to two classic machine learning (ML) algorithms—random forest (RF) and multilayer perceptron (MLP)—as points of reference. After validation, we found that the DL models were better in most performance metrics including the Kappa index, F1 score, and mean intersection over union (mIoU) measure, while the ResUnet model achieved the best overall results with a value of 0.94 in all three measures in both time sequences. Visually, the DL models also provided classifications with better defined deforestation patches and did not need any sort of post-processing to remove noise, unlike the ML models, which needed some noise removal to improve results.

show abstract

Triplet-Based Semantic Relation Learning for Aerial Remote Sensing Image Change Detection

Cited by 193 publications

References 13 publications

End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++

End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++

A Feature Space Constraint-Based Method for Change Detection in Heterogeneous Images

Change Detection of Deforestation in the Brazilian Amazon Using Landsat Data and Convolutional Neural Networks

Contact Info

Product

Resources

About