Spatial–Spectral Fusion of HY-1C COCTS/CZI Data for Coastal Water Remote Sensing Using Deep Belief Network

Ji, Hongren; Tian, Liqiao; Li, Jian; Tong, Ruqing; Guo, Yulong; Zeng, Qun

doi:10.1109/jstars.2020.3045516

Cited by 7 publications

(7 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The DBN method is one of the most representative deep learning models [21,22,33,34] and consists of multiple restricted Boltzmann machine (RBM) layers and a back propagation (BP) structure [11,35]. Lower layers of DBN can extract low-level features and the upper layers are used to represent more abstract characteristics of the input data [36].…”

Section: Sla Estimation Model Based On Dbn Methodsmentioning

confidence: 99%

“…Including scheme (b), the results of the four methods are compared, and the CORRs, STDs and RMSE for the six schemes are shown in Table 4. [ 2,11,12,14,16,17,18,19,20,22] Inside e [ 1,3,5,6,7,8,9,10,13,15,21] Upper left corner f [ 4,12,14,17,18,19,22 Figure 6. The distribution of in-situ tide gauges used as external control for different schemes and altimetry-derived along-track SLAs in training data.…”

Section: Step Three: Training With In-situ Tide Gaugesmentioning

confidence: 99%

“…Experimental verification and comparisons demonstrated that the DBN model exhibited better prediction and stability than the multiple linear regression, back propagation neural network and support vector machine models. Using a DBN model, Ji et al [21] fused images from the Chinese ocean color and temperature scanner and coastal zone imager onboard HaiYang-1C satellite to generate 50-m, eight-band, and three-day observations for coastal waters. The results indicated that the DBN was better than Gram-Schmidt transformation and inversion-based fusion algorithms.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Sea Level Fusion of Satellite Altimetry and Tide Gauge Data by Deep Learning in the Mediterranean Sea

et al. 2021

View full text Add to dashboard Cite

Satellite altimetry and tide gauges are the two main techniques used to measure sea level. Due to the limitations of satellite altimetry, a high-quality unified sea level model from coast to open ocean has traditionally been difficult to achieve. This study proposes a fusion approach of altimetry and tide gauge data based on a deep belief network (DBN) method. Taking the Mediterranean Sea as the case study area, a progressive three-step experiment was designed to compare the fused sea level anomalies from the DBN method with those from the inverse distance weighted (IDW) method, the kriging (KRG) method and the curvature continuous splines in tension (CCS) method for different cases. The results show that the fusion precision varies with the methods and the input measurements. The precision of the DBN method is better than that of the other three methods in most schemes and is reduced by approximately 20% when the limited altimetry along-track data and in-situ tide gauge data are used. In addition, the distribution of satellite altimetry data and tide gauge data has a large effect on the other three methods but less impact on the DBN model. Furthermore, the sea level anomalies in the Mediterranean Sea with a spatial resolution of 0.25° × 0.25° generated by the DBN model contain more spatial distribution information than others, which means the DBN can be applied as a more feasible and robust way to fuse these two kinds of sea levels.

show abstract

Section: Sla Estimation Model Based On Dbn Methodsmentioning

confidence: 99%

Section: Step Three: Training With In-situ Tide Gaugesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Sea Level Fusion of Satellite Altimetry and Tide Gauge Data by Deep Learning in the Mediterranean Sea

et al. 2021

View full text Add to dashboard Cite

show abstract

“…In a remote sensing scene classification task, many researchers explore the potential of multibranch networks by fusing features of different branches [8], [9], [36]. Teffahi and Yao [37], Ji et al [38], and Wang et al [39] achieved the feature fusion method by extracting multiple spectral and spatial features and concatenating them, which improves the accuracy of remote sensing image classification. Tan et al [8] used a multibranch lightweight network to extract image features and built a graph model based on the learned features.…”

Section: B Multibranch Networkmentioning

confidence: 99%

Embedded Self-Distillation in Compact Multibranch Ensemble Network for Remote Sensing Scene Classification

Zhao

Lyu

et al. 2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…In RS scene classification task, many researchers explore the potential of multi-branch networks by fusing features of different branches [37], [38], [39]. [40], [41], [42] achieve feature fusion method by extracting multiple spectral and spatial features and concatenate them, which improves the accuracy of RS image classification. [7] uses a multi-branch lightweight network to extract image features, and builds a graph model based on the learned features.…”

Section: B Multi-branch Networkmentioning

confidence: 99%

Embedded Self-Distillation in Compact Multi-Branch Ensemble Network for Remote Sensing Scene Classification

Zhao,

Ma,

Lyu

et al. 2021

Preprint

View full text Add to dashboard Cite

Remote sensing (RS) image scene classification task faces many challenges due to the interference from different characteristics of different geographical elements. To solve this problem, we propose a multi-branch ensemble network to enhance the feature representation ability by fusing features in final output logits and intermediate feature maps. However, simply adding branches will increase the complexity of models and decline the inference efficiency. On this issue, we embed selfdistillation (SD) method to transfer knowledge from ensemble network to main-branch in it. Through optimizing with SD, main-branch will have close performance as ensemble network. During inference, we can cut other branches to simplify the whole model. In this paper, we first design compact multibranch ensemble network, which can be trained in an endto-end manner. Then, we insert SD method on output logits and feature maps. Compared to previous methods, our proposed architecture (ESD-MBENet) performs strongly on classification accuracy with compact design. Extensive experiments are applied on three benchmark RS datasets AID, NWPU-RESISC45 and UC-Merced with three classic baseline models, VGG16, ResNet50 and DenseNet121. Results prove that our proposed ESD-MBENet can achieve better accuracy than previous state-of-the-art (SOTA) complex models. Moreover, abundant visualization analysis make our method more convincing and interpretable.

show abstract

Spatial–Spectral Fusion of HY-1C COCTS/CZI Data for Coastal Water Remote Sensing Using Deep Belief Network

Cited by 7 publications

References 57 publications

Sea Level Fusion of Satellite Altimetry and Tide Gauge Data by Deep Learning in the Mediterranean Sea

Sea Level Fusion of Satellite Altimetry and Tide Gauge Data by Deep Learning in the Mediterranean Sea

Embedded Self-Distillation in Compact Multibranch Ensemble Network for Remote Sensing Scene Classification

Embedded Self-Distillation in Compact Multi-Branch Ensemble Network for Remote Sensing Scene Classification

Contact Info

Product

Resources

About