A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sensing images

Zhang, Chenxiao; Yue, Peng; Tapete, Deodato; Jiang, Liangcun; Shangguan, Boyi; Huang, Li; Liu, Guangchao

doi:10.1016/j.isprsjprs.2020.06.003

Cited by 683 publications

(419 citation statements)

References 28 publications

Supporting

Mentioning

419

Contrasting

Order By: Relevance

“…It has excellent performance on LEBEDEV, but for the boundary accuracy and multiple small objects, it still can be improved. IFN [ 1 ] utilizes a feature extraction network with shared parameters to encode the original images, which is similar to the FC-Siam-conc. In the decoding process, it implements discrimination learning on the difference between the features of each layer in the former streams.…”

Section: Experiments and Resultsmentioning

confidence: 99%

“…of the objects on the land surface. Recently, research on change detection has been an active topic with the rapid development of computer vision techniques [ 1 , 2 , 3 , 4 , 5 ].…”

Section: Introductionmentioning

confidence: 99%

“…Based on these baselines, the PGA-SiamNet [ 4 ] introduces a co-attention guide onto the bridge before decoding, to emphasize the importance of correlation among the feature maps pairs. In Zhang et al [ 1 ], a deep supervised fusion strategy is introduced to the FC-Siam-conc like model, which is completed by using attention modules after on concatenated encoding feature maps and reconstructed maps, to suppress the irrelevant samples and spatial transformed refine the maps. By doing that, it improves the change maps boundary completeness and internal compactness.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Coarse-to-Fine Satellite Images Change Detection Framework via Boundary-Aware Attentive Network

Zhang

2020

Sensors

View full text Add to dashboard Cite

Timely and accurate change detection on satellite images by using computer vision techniques has been attracting lots of research efforts in recent years. Existing approaches based on deep learning frameworks have achieved good performance for the task of change detection on satellite images. However, under the scenario of disjoint changed areas in various shapes on land surface, existing methods still have shortcomings in detecting all changed areas correctly and representing the changed areas boundary. To deal with these problems, we design a coarse-to-fine detection framework via a boundary-aware attentive network with a hybrid loss to detect the change in high resolution satellite images. Specifically, we first perform an attention guided encoder-decoder subnet to obtain the coarse change map of the bi-temporal image pairs, and then apply residual learning to obtain the refined change map. We also propose a hybrid loss to provide the supervision from pixel, patch, and map levels. Comprehensive experiments are conducted on two benchmark datasets: LEBEDEV and SZTAKI to verify the effectiveness of the proposed method and the experimental results show that our model achieves state-of-the-art performance.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

“…of the objects on the land surface. Recently, research on change detection has been an active topic with the rapid development of computer vision techniques [ 1 , 2 , 3 , 4 , 5 ].…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Coarse-to-Fine Satellite Images Change Detection Framework via Boundary-Aware Attentive Network

Zhang

2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Despite the increasing number of raw RS images, the timeconsuming and labor-intensive work of manual interpretation still hinders the development of CD methods considering the data-hungry nature of DNNs. Currently, only a few open available labelled datasets can be used for model training and evaluation, such as LEVIR-CD [6], Season-Varying [11], DSIFN [8], HRSCD [4]. Unfortunately, all of them have less data than the ImageNet and COCO datasets.…”

Section: Introductionmentioning

confidence: 99%

3M-CDNet: A Universal Powerful Benchmark for Remote Sensing Image Change Detection

Song¹,

Jiang²

2021

Preprint

View full text Add to dashboard Cite

Abstract—While deep learning-based methods have gained popularity and have made remarkable progress in remote sensing (RS) image change detection (CD), the limited amount of available data hinders the performance of most supervised methods. The CD networks transferred or derived from other fields can be fronted with a weak generalization capability. Developing a universal benchmark for performance evaluations based on the available datasets is urgent. To address these problems, we proposed a lightweight network, termed 3M-CDNet, which only requires about 3.12 M parameters. The lighter the network, the easier it is to train and alleviate overfitting the limited amount of data, resulting in a better generalization capability. 3M-CDNet has a flexible modular design that achieves performance improvements by incorporating plug-and-play modules. 3M-CDNet gains accuracy improvements in two ways: (1) the application of deformable convolutions (DConv) in the backbone network to gain a good geometric transformation modeling capacity for CD and (2) the application of an effective two-level feature fusion strategy to enhance the feature representation capacity. 3M-CDNet gains a good generalization capacity by incorporating effective “tricks” to alleviate overfitting, in which online data augmentation (Online DA) is applied to increase the diversity of the training samples, and Dropout regularization is applied in the classifier. Extensive ablation studies have proved the effectiveness of the core components. Experiment results suggest that 3M-CDNet outperforms state-of-the-art methods on several optical RS datasets and serves as a new universal benchmark. Specifically, 3M-CDNet achieves the best F1-score, i.e., LEVIR-CD (0.9161), Season-Varying (0.9473), and DSIFN (0.7031).

show abstract

“…The main methods for detecting land cover change are the post-classification comparison method and the direct comparison method [9]. Subject to current automatic image change detection techniques, post-classification comparison is still the dominant technical approach, and supervised classification methods are mostly used [10]. Land cover evolution models are used to describe, interpret, predict, optimize, and support decision making on land cover change.…”

Section: Introductionmentioning

confidence: 99%

Analysis of land cover evolution within the built-up areas of provincial capital cities in northeastern China based on nighttime light data and Landsat data

et al. 2020

View full text Add to dashboard Cite

Mastering the evolution of urban land cover is important for urban management and planning. In this paper, a method for analyzing land cover evolution within urban built-up areas based on nighttime light data and Landsat data is proposed. The method solves the problem of inaccurate descriptions of urban built-up area boundaries from the use of single-source diurnal or nocturnal remote sensing data and was able to achieve an effective analysis of land cover evolution within built-up areas. Four main procedures are involved: (1) The neighborhood extremum method and maximum likelihood method are used to extract nighttime light data and the urban built-up area boundaries from the Landsat data, respectively; (2) multisource urban boundaries are obtained using boundary pixel fusion of the nighttime light data and Landsat urban built-up area boundaries; (3) the maximum likelihood method is used to classify Landsat data within multisource urban boundaries into land cover classes, such as impervious surface, vegetation and water, and to calculate landscape indexes, such as overall landscape trends, degree of fragmentation and degree of aggregation; (4) the changes in the multisource urban boundaries and landscape indexes were obtained using the abovementioned methods, which were supported by multitemporal nighttime light data and Landsat data, to model the urban land cover evolution. Using the cities of Shenyang, Changchun and Harbin in northeastern China as experimental areas, the multitemporal landscape index showed that the integration and aggregation of land cover in the urban areas had an increasing trend, the natural environment of Shenyang and Harbin was improving, while Changchun laid more emphasis on the construction of artificial facilities. At the same time, the method proposed in this paper to extract built-up areas from multi-source city data showed that the user accuracy, production accuracy, overall accuracy and Kappa coefficient are at least 3%, 1%, 1% and 0.04 higher than the single-source data method.

show abstract

A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sensing images

Cited by 683 publications

References 28 publications

Coarse-to-Fine Satellite Images Change Detection Framework via Boundary-Aware Attentive Network

Coarse-to-Fine Satellite Images Change Detection Framework via Boundary-Aware Attentive Network

3M-CDNet: A Universal Powerful Benchmark for Remote Sensing Image Change Detection

Analysis of land cover evolution within the built-up areas of provincial capital cities in northeastern China based on nighttime light data and Landsat data

Contact Info

Product

Resources

About