Owing to the limitation of spatial resolution and spectral resolution, deep learning methods are rarely used for the classification of multispectral remote sensing images based on the real spectral dataset from multispectral remote sensing images. This study explores the application of a deep learning model to the spectral classification of multispectral remote sensing images. To address the problem of the large workload with respect to selecting training samples during classification by deep learning, first, linear spectral mixture analysis and the spectral index method were applied to extract the pixels of impervious surfaces, soil, vegetation, and water. Second, through the Euclidean distance threshold method, a spectral dataset of multispectral image pixels was established. Third, a deep learning classification model, ResNet-18, was constructed to classify Landsat 8 OLI images based on pixels’ real spectral information. According to the accuracy assessment, the results show that the overall accuracy of the classification results can reach 0.9436, and the kappa coefficient can reach 0.8808. This study proposes a method that allows for the more optimized establishment of the actual spectral dataset of ground objects, addresses the limitations of difficult sample selection in deep learning classification and of spectral similarity in traditional classification methods, and applies the deep learning method to the classification of multispectral remote sensing images based on a real spectral dataset.

/pdf/deep-learning-classification-by-resnet-18-based-on-the-real-qzy86lc5.pdf

Deep Learning Classification by ResNet-18 Based on the Real Spectral Dataset from Multispectral Remote Sensing Images

Algorithmic Urban Planning for Smart and Sustainable Development: Systematic Review of the Literature

Semantic segmentation for accurate visual perception is a critical task in computer vision. In principle, the automatic classification of dynamic visual scenes using predefined object classes remains unresolved. The challenging problems of learning deep convolution neural networks, specifically ResNet-based DeepLabV3+ (the most recent version), are threefold. The problems arise due to (1) biased centric exploitations of filter masks, (2) lower representational power of residual networks due to identity shortcuts, and (3) a loss of spatial relationship by using per-pixel primitives. To solve these problems, we present a proficient approach based on DeepLabV3+, along with an added evaluation metric, namely, Unified DeepLabV3+ and S3core, respectively. The presented unified version reduced the effect of biased exploitations via additional dilated convolution layers with customized dilation rates. We further tackled the problem of representational power by introducing non-linear group normalization shortcuts to solve the focused problem of semi-dark images. Meanwhile, to keep track of the spatial relationships in terms of the global and local contexts, geometrically bunched pixel cues were used. We accumulated all the proposed variants of DeepLabV3+ to propose Unified DeepLabV3+ for accurate visual decisions. Finally, the proposed S3core evaluation metric was based on the weighted combination of three different accuracy measures, i.e., the pixel accuracy, IoU (intersection over union), and Mean BFScore, as robust identification criteria. Extensive experimental analysis performed over a CamVid dataset confirmed the applicability of the proposed solution for autonomous vehicles and robotics for outdoor settings. The experimental analysis showed that the proposed Unified DeepLabV3+ outperformed DeepLabV3+ by a margin of 3% in terms of the class-wise pixel accuracy, along with a higher S3core, depicting the effectiveness of the proposed approach.

/pdf/unified-deeplabv3-for-semi-dark-image-semantic-segmentation-1gcm82v9.pdf

Unified DeepLabV3+ for Semi-Dark Image Semantic Segmentation

The uncertainty from the under-estimation and over-estimation of forest aboveground biomass (AGB) is an urgent problem in optical remote sensing estimation. In order to more accurately estimate the AGB of Pinus densata forests in Shangri-La City, we mainly discuss three non-parametric models—the artificial neural network (ANN), random forests (RFs), and the quantile regression neural network (QRNN) based on 146 sample plots and Sentinel-2 images in Shangri-La City, China. Moreover, we selected the corresponding optical quartile models with the lowest mean error at each AGB segment to combine as the best QRNN (QRNNb). The results showed that: (1) for the whole biomass segment, the QRNNb has the best fitting performance compared with the ANN and RFs, the ANN has the lowest R2 (0.602) and the highest RMSE (48.180 Mg/ha), and the difference between the QRNNb and RFs is not apparent. (2) For the different biomass segments, the QRNNb has a better performance. Especially when AGB is lower than 40 Mg/ha, the QRNNb has the highest R2 of 0.961 and the lowest RMSE of 1.733 (Mg/ha). Meanwhile, when AGB is larger than 160 Mg/ha, the QRNNb has the highest R2 of 0.867 and the lowest RMSE of 18.203 Mg/ha. This indicates that the QRNNb is more robust and can improve the over-estimation and under-estimation in AGB estimation. This means that the QRNNb combined with the optimal quantile model of each biomass segment provides a method with more potential for reducing the uncertainties in AGB estimation using optical remote sensing images.

/pdf/reduction-in-uncertainty-in-forest-aboveground-biomass-9hrrmj3p.pdf

Reduction in Uncertainty in Forest Aboveground Biomass Estimation Using Sentinel-2 Images: A Case Study of Pinus densata Forests in Shangri-La City, China

Accurate estimation of small-scale forest biomass is a prerequisite and basis for trading forest carbon sinks and optimizing the allocation of forestry resources. This study aims to develop a plot-scale methodology for estimating aboveground biomass (AGB) that combines a biomass horizontal distribution model (HDM) and sampling techniques to improve efficiency, reduce costs and provide the reliability of estimation for biomass. Simao pine (Pinus kesiya var. langbianensis) from Pu’er City, Yunnan Province, was used as the research subject in this study. A canopy profile model (CPM) was constructed based on data from branch analysis and transformed into a canopy biomass HDM. The horizontal distribution of AGB within the sample plots was simulated using the HDM based on the data from the per-wood survey and compared with the results from the location distribution model (LDM) simulation. AGB sampling estimations were carried out separately by combining different sampling methods with the AGB distribution of sample plot simulated by different biomass distribution models. The sampling effectiveness of all sampling schemes was compared and analyzed, and the best plan for the sampling estimation of AGB in plot-scale forests was optimized. The results are as follows: the power function model is the best model for constructing the CPM of the Simao pine in this study; with visual comparison and the analysis of the coefficient of variation, the AGB simulated by HDM has a larger and more continuous distribution than that simulated by LDM, which is closer to the actual distribution; HDM-based sampling plans have smaller sample sizes and sampling ratios than LDM-based ones; and lastly, the stratified sampling method (STS)-HDM-6 plan has the best sampling efficiency with a minimum sample size of 10 and a minimum sampling ratio of 15%. The result illustrates the potential of the method for estimating plot-scale forest AGB by combining HDM with sampling techniques to reduce costs and increase estimation efficiency effectively.

/pdf/a-method-for-estimating-forest-aboveground-biomass-at-the-1z4frf4e.pdf

A Method for Estimating Forest Aboveground Biomass at the Plot Scale Combining the Horizontal Distribution Model of Biomass and Sampling Technique

The classification of urban land-use information has become the underlying database for a variety of applications including urban planning and administration. The lack of datasets and changeable semantics of land-use make deep learning methods suffer from low precision, which prevent improvements in the effectiveness of using AI methods for applications. In this paper, we first used GIS data to produce a well-tagged and high-resolution urban land-use image dataset. Then, we proposed a combined convolutional neural network named DUA-Net for complex and diverse urban land-use classification. The DUA-Net combined U-Net and Densely connected Atrous Spatial Pyramid Pooling (DenseASPP) to extract Remote Sensing Imagers (RSIs) features in parallel. Then, channel attention was used to efficiently fuse the multi-source semantic information from the output of the double-layer network to learn the association between different land-use types. Finally, land-use classification of high-resolution urban RSIs was achieved. Experiments were performed on the dataset of this paper, the publicly available Vaihingen dataset and Potsdam dataset with overall accuracy levels reaching 75.90%, 89.71% and 89.91%, respectively. The results indicated that the complex land-use types with heterogeneous features were more difficult to extract than the single-feature land-cover types. The proposed DUA-Net method proved suitable for high-precision urban land-use classification, which will be of great value for urban planning and national land resource surveying.

/pdf/a-combined-convolutional-neural-network-for-urban-land-use-3pzz1uvd.pdf

A Combined Convolutional Neural Network for Urban Land-Use Classification with GIS Data

Forest biomass plays an essential role in forest carbon reservoir studies, biodiversity protection, forest management, and climate change mitigation actions. Synthetic Aperture Radar (SAR), especially the polarimetric SAR with the capability of identifying different aspects of forest structure, shows great potential in the accurate estimation of total and component forest above-ground biomass (AGB), including stem, bark, branch, and leaf biomass. This study aims to fully explore the potential of polarimetric parameters at the C- and L-bands to achieve high estimation accuracy and improve the estimation of AGB saturation levels. In this study, the backscattering coefficients at different polarimetric channels and polarimetric parameters extracted from Freeman2, Yamaguchi3, H-A-Alpha, and Target Scattering Vector Model (TSVM) decomposition methods were optimized by a random forest algorithm, first, and then inputted into linear regression models to estimate the total forest AGB and biomass components of two test sites in China. The results showed that polarimetric observations had great potential in total and component AGB estimation in the two test sites; the best performances were for leaves at test site I, with R2 = 0.637 and RMSE = 1.27 t/hm2. The estimation of biomass components at both test sites showed obvious saturation phenomenon estimation according to their scatter plots. The results obtained at both test sites demonstrated the potential of polarimetric parameters in total and component biomass estimation.

/pdf/forest-total-and-component-above-ground-biomass-agb-2fpshe0f.pdf

Forest Total and Component Above-Ground Biomass (AGB) Estimation through C- and L-band Polarimetric SAR Data

Forest total and component biomass retrieval via GA-SVR algorithm and quad-polarimetric SAR data

Forest biomass plays an essential role in forest carbon reservoirs studies, biodiversity protection, forest manage-ment, and climate change mitigation actions. Parameters extracted from Light Detection and Ranging (LiDAR) and X-band Synthetic Aperture Radar (SAR) data were used in separately and in combination to estimate total forest above-ground biomass (AGB), but rarely used in components AGB estimation. In this paper, we extracted intensity, density, and height parameters from LiDAR data, coherence coefficients from Interferometric SAR (InSAR) data, backscatter coeffi-cients and polarimetric decomposition parameters from Po-larimetric SAR (PoISAR) to estimate forest total and components AGB. The results showed that PolSAR parameters have a unique advantage to estimate leaf biomass, with the highest R2 of 0.773. And for total, bark and branch AGB, LiDAR, InSAR and PolSAR parameter combination have better accuracy, with R2 of 0.818, 0.834, and 0.842, respec-tively. The study revealed that LiDAR and SAR used in combination can effectively estimation the forest total and components AGB.

Component Forest Above Ground Biomass Estimation Using Lidar and Sardata

Remote sensing scene classification aims to automatically assign proper labels to remote sensing images. Most of the existing deep learning based methods usually consider the interclass and intraclass relationships of the image content for classification. However, these methods rarely consider the hierarchical information of scene labels, as a scene label may belong to hierarchically multi-grained levels. For example, multi-grained level labels may indicate that a remote sensing scene image may belong to the coarse-grained label “transportation land” while also belonging to the fine-grained label “airport”. In this paper, to exploit hierarchical label information, we propose an attention-embedding multi-task multi-grained network (AEMMN) for remote sensing scene classification. In the proposed AEMMN, we add a coarse-grained classifier as the first level and a fine-grained classifier as the second level to perform multi-task learning tasks. Additionally, a gradient control module is utilized to control the gradient propagation of two classifiers to suppress the negative transfer caused by the irrelevant features between tasks. In the feature extraction portion, the model uses an ECA module embedding Resnet50 to extract effective features with cross-channel interaction information. Furthermore, an external attention module is exploited to improve the discrimination of fine-grained and coarse-grained features. Experiments were conducted on the NWPU-RESISC45 and the Aerial Image Data Set (AID), and the overall accuracy of the proposed AEMMN is 92.07% on the NWPU-RESISC45 dataset and reached 94.96% on the AID. The results indicate that hierarchical label information can effectively improve the performance of scene classification tasks when categorizing remote sensing imagery.

/pdf/exploiting-hierarchical-label-information-in-an-attention-2i2p6dmn.pdf

Exploiting Hierarchical Label Information in an Attention-Embedding, Multi-Task, Multi-Grained, Network for Scene Classification of Remote Sensing Imagery

Peng Zeng

Papers

A Combined Convolutional Neural Network for Urban Land-Use Classification with GIS Data

Forest Total and Component Above-Ground Biomass (AGB) Estimation through C- and L-band Polarimetric SAR Data

Forest total and component biomass retrieval via GA-SVR algorithm and quad-polarimetric SAR data

Component Forest Above Ground Biomass Estimation Using Lidar and Sardata

Exploiting Hierarchical Label Information in an Attention-Embedding, Multi-Task, Multi-Grained, Network for Scene Classification of Remote Sensing Imagery