Spatial Depth Research Articles

Depth of Space and Radius of the Electron

The cosmological model of the expanding balloon in 4D-space (CM) delivers in interaction with a homogeneous vector field exactly Newton’s law of gravitation with its 1/r-shape of the gravitational funnel. So far, the depth of space, W, in the 4-th spatial dimension can only be calculated using the theoretical approach of Feynman’s radius of excess rex=a/3 with Schwarzschild-radius a. With this, the connection to the general theory of relativity (GR) is established, but the situation is unsatisfactory. In the present study, the possibilities of an experimental approach to the calculation of spatial depth, W, are explored. The only experimental approach so far is the bending of light on a central mass. We hypothesize in addition to the main effect φ = -4a/y, i.e., the angle of diffraction of a light beam on a heavy central mass in the distance y and with Schwarzschild-radius a, an additional effect close to the center of the form φC ~ -1/y4. This additional effect has on the edge of the central mass about 1/3 of the strength of the main effect. However, its influence disappears very quickly with increasing distance. For this reason the sun cannot be used as the central mass. The bright corona and the strong magnetosphere do not allow measurements close to the sun. However, ESA’s GAIA mission puts the planet Jupiter at the center of interest. This spacecraft measures with extremely high precision the positions of billions of stars. Results of first data analyses have already been published. As a side effect - the application of the CM to small particles provides an indication that the radius of the electron could be in the order of 10-23 m.

Fluorescence/photoacoustic imaging-guided nanomaterials for highly efficient cancer theragnostic agent

Imaging modalities combined with a multimodal nanocomposite contrast agent hold great potential for significant contributions in the biomedical field. Among modern imaging techniques, photoacoustic (PA) and fluorescence (FL) imaging gained much attention due to their non-invasive feature and the mutually supportive characteristic in terms of spatial resolution, penetration depth, imaging sensitivity, and speed. In this present study, we synthesized IR783 conjugated chitosan–polypyrrole nanocomposites (IR-CS–PPy NCs) as a theragnostic agent used for FL/PA dual-modal imaging. A customized FL and photoacoustic imaging system was constructed to perform required imaging experiments and create high-contrast images. The proposed nanocomposites were confirmed to have great biosafety, essentially a near-infrared (NIR) absorbance property with enhanced photostability. The in vitro photothermal results indicate the high-efficiency MDA-MB-231 breast cancer cell ablation ability of IR-CS–PPy NCs under 808 nm NIR laser irradiation. The in vivo PTT study revealed the complete destruction of the tumor tissues with IR-CS–PPy NCs without further recurrence. The in vitro and in vivo results suggest that the demonstrated nanocomposites, together with the proposed imaging systems could be an effective theragnostic agent for imaging-guided cancer treatment.

Simultaneous multielement imaging of liver tissue using laser ablation inductively coupled plasma mass spectrometry

Analysis of the spatial distribution of metals, metalloids, and non-metals in biological tissues is of significant interest in the life sciences, helping to illuminate the function and roles these elements play within various biological pathways. Chemical imaging methods are commonly employed to address biological questions and reveal individual spatial distributions of analytes of interest. Elucidation of these spatial distributions can help determine key elemental and molecular information within the respective biological specimens. However, traditionally utilized imaging methods prove challenging for certain biological tissue analysis, especially with respect to applications that require high spatial resolution or depth profiling. Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) has been shown to be effective for direct elemental analysis of solid materials with high levels of precision. In this work, chemical imaging using LA-ICP-MS has been applied as a powerful analytical methodology for the analysis of liver tissue samples. The proposed analytical methodology successfully produced both qualitative and quantitative information regarding specific elemental distributions within images of thin tissue sections with high levels of sensitivity and spatial resolution. The spatial resolution of the analytical methodology was innovatively enhanced, helping to broaden applicability of this technique to applications requiring significantly high spatial resolutions. This information can be used to further understand the role these elements play within biological systems and impacts dysregulation may have.

The influence of seasonal temperature variation on embryonic development in two species of spider crabs (Majoidea)

The aim of this study was to analyze the ecological influence of seasonal temperature variations on embryonic development in two species of Majoidea. Leurocyclus tuberculosus and Libinia spinosa. These species inhabit stable populations in the San José Gulf (42°25′41″S, 64°08′27″O, Patagonia-Argentina) in a wide depth range and seasonal temperature variations from 8 °C to 18 °C. The influence of seasonal temperatures on embryonic development was examined in ovigerous females of both species, through a bioassay in laboratory conditions at 8 °C, 14 °C, and 18 °C from gastrula to hatching. Additionally, monthly female samplings were conducted for one year, characterizing their spatial depth distribution associated with the seawater temperature in the study area and their ovigerous/non ovigerous condition. Laboratory experimental results showed that duration of incubation increased 67 % in L. tuberculosus and 55% in L. spinosa by a decline in temperature from 18 °C to 8 °C with one diapause period. The embryonic development for the 3 assayed temperatures was divided in five periods. The assayed temperatures did not affect the morphology of the embryo, and its growth during each period did not present significant differences among the different temperatures. Field analysis showed significant differences in the spatial distribution of ovigerous females which were associated to the assessed temperatures along the year. This distribution might modulate the development of embryos leading to two spawnings synchronized with the phyto- and zooplankton peaks in the San José gulf. Finally, ecological and potential impact of ocean warming in these species was also addressed.

Feature Recognition of Human Motion Behavior Based on Depth Sequence Analysis

The current research on still image recognition has been very successful, but the study of action recognition for video classes is still a challenging topic. In this work, we propose a random projection-based human action recognition algorithm to address the lack of depth information in color information (RGB video frames) that is not easily affected by environmental factors such as illumination and the lack of ability to recognize actions along the direction of view. A network structure is designed to take the obvious advantage of long- and short-term memory networks for controlling and remembering long sequences of historical information. The network structure in this paper is constituted by multiple memory units. At the same time, this paper constructs the spatial features, temporal features, and depth features of the three recognition stream outputs into a feature matrix, whose feature matrix is divided into multiple temporal segments according to the temporal dimension, then inputs them into the network layer in order, and achieves the fusion of the feature matrix in this paper according to their correlation characteristics on the temporal axis. Here, we proposed the concept of random batch projection operators. This basically uses as much sublimitation information as possible to improve projection accuracy by randomly selecting several subdependencies as projections defined during projection. A compressed sensing design of human motion acceleration data for low-power body area networks is proposed, and the basic idea and implementation process of compressed sensing theory for human motion data compression and reconstruction in wireless body area networks are introduced in detail.

Pseudo Video and Refocused Images-Based Blind Light Field Image Quality Assessment

The commercial light field camera is able to capture four-dimensional Light Field Image (LFI), which can be visualized to LFI contents on 2D displays by means of the Pseudo Video (PV) or the Refocused Images (RIs) generated with the refocusing function of LFI. However, the quality degradation of LFI will affect user’s visual experience of LFI contents. Hence, it is crucial to develop an effective LFI quality assessment method to monitor the LFI quality. Most existing subjective databases of LFI use PV and RIs visualization techniques to assess the quality of LFI. Therefore, as the way of presenting LFI on 2D display, PV and RIs are closely related to the subjective perception of LFI by human eyes. Based on these two visualization techniques, this article proposes a novel PV and RIs based blind LFI quality assessment method, in which the feature extraction is divided into two parts. In the first part, the PV’s structure, motion and disparity information are extracted with multi-scale and multi-directional Shearlet transform. In the other part, the spatial structure, depth and semantic information of the RIs are obtained. Finally, support vector regression is used to nonlinear map the perceptual features to quality score of LFI. The experimental results on four LFI databases show that the proposed method has better correlation with human visual perception, compared with the classical 2D image quality assessment methods as well as the state-of-the-art LFI quality assessment methods.

Hierarchical Object Relationship Constrained Monocular Depth Estimation.

Monocular depth estimation has been gaining growing momentum in recent years. Despite significant advances of this task, due to the inherent difficulty of reliably capturing contextual cues from RGB images, it remains challenging to accurately predict depth in scenes with complicated and cluttered spatial arrangement of objects. Instead of naively utilizing the primary features in the single RGB image, in this paper we propose a hierarchical object relationship constrained network for monocular depth estimation, which could enable accurate and smooth depth prediction from monocular RGB image. The key idea of our method is to exploit object-centric hierarchical relationship as contextual constraints to compensate for the regularity of spatial depth changing. In particular, we design a semantics-guided CNN network to encode the original image into a global context feature map and encode the objects’ relationship into a local relationship feature map simultaneously, so that we can leverage such effective and consolidated coding scheme over scenario samples to guide the depth prediction in a more accurate way. Benefiting from the local-to-global context constraints, our method can well respect the global depth changing and preserve the local depth details at the same time. In addition, our approach could make full use of the hierarchical semantic relationship across inner-object components and neighboring objects to define depth changing constraints. We conduct extensive experiments and make comprehensive evaluations on widely-used public datasets, and the experiments confirm that our method outperforms most state-of-the-art depth estimation methods in preserving the local details in depth.

ViSTRA2: Video coding using spatial resolution and effective bit depth adaptation

We present a new video compression framework (ViSTRA2) which exploits adaptation of spatial resolution and effective bit depth, down-sampling these parameters at the encoder based on perceptual criteria, and up-sampling at the decoder using a deep convolution neural network. ViSTRA2 has been integrated with the reference software of both the HEVC (HM 16.20) and VVC (VTM 4.0.1), and evaluated under the Joint Video Exploration Team Common Test Conditions using the Random Access configuration. Our results show consistent and significant compression gains against HM and VVC based on Bjønegaard Delta measurements, with average BD-rate savings of 12.6% (PSNR) and 19.5% (VMAF) over HM and 5.5% (PSNR) and 8.6% (VMAF) over VTM.

Motion Estimation Using Region-Level Segmentation and Extended Kalman Filter for Autonomous Driving

Motion estimation is crucial to predict where other traffic participants will be at a certain period of time, and accordingly plan the route of the ego-vehicle. This paper presents a novel approach to estimate the motion state by using region-level instance segmentation and extended Kalman filter (EKF). Motion estimation involves three stages of object detection, tracking and parameter estimate. We first use a region-level segmentation to accurately locate the object region for the latter two stages. The region-level segmentation combines color, temporal (optical flow), and spatial (depth) information as the basis for segmentation by using super-pixels and Conditional Random Field. The optical flow is then employed to track the feature points within the object area. In the stage of parameter estimate, we develop a relative motion model of the ego-vehicle and the object, and accordingly establish an EKF model for point tracking and parameter estimate. The EKF model integrates the ego-motion, optical flow, and disparity to generate optimized motion parameters. During tracking and parameter estimate, we apply edge point constraint and consistency constraint to eliminate outliers of tracking points so that the feature points used for tracking are ensured within the object body and the parameter estimates are refined by inner points. Experiments have been conducted on the KITTI dataset, and the results demonstrate that our method presents excellent performance and outperforms the other state-of-the-art methods either in object segmentation and parameter estimate.

Quantitative risk assessment of storm surge using GIS techniques and open data: A case study of Daya Bay Zone, China

Storm surge is a natural disaster, often causing economic damage and loss of human life in the coastal communities. In recent decades, with more people attracted to coastal areas, the potential economic losses resulted from storm surges are increasing. Therefore, it is important to make risk assessments to identify areas at risk and design risk reduction strategies. However, the quantitative risk assessment of storm surge for coastal cities in China is often difficult due to the lack of adequate data regarding the building footprint and vulnerability curves. This paper aims to provide a methodology for conducting the quantitative risk assessment of storm surge, estimating direct tangible damage, by using Geographical Information System (GIS) techniques and open data. The proposed methodology was applied to a coastal area with a high concentration of petroleum industries in the Daya Bay zone. At first, five individual typhoon scenarios with different return periods (1000, 100, 50, 20, and 10 years) were defined. Then, the Advanced Circulation model and the Simulating Waves Nearshore model were utilized to simulate storm surge. The model outputs were imported into GIS software, transformed into inundation area and inundation depth. Subsequently, the building footprint data were extracted by the use of GIS techniques, including spatial analysis and image analysis. The layer containing building footprints was superimposed on the inundation area layer to identify and quantify the exposed elements to storm surge hazard. Combining the exposed elements with their related depth–damage functions, the quantitative risk assessment translates the spatial extent and depth of storm surge into the estimation of economic losses. The quantitative risk assessment and zonation maps for sub-zones in the study area can help local decision-makers to prioritize the sub-zones that are more likely to be affected by storm surge, make risk mitigation strategies, and develop long-term urban plans.

Depth classification based on affine-invariant, weighted and kernel-based spatial depth functions

Several multivariate depth functions have been proposed in the literature, of which some satisfy all the conditions for statistical depth functions while some do not. Spatial depth is known to be invariant to spherical and shift transformations. In this paper, the possibility of using different versions of spatial depth in classification is considered. The covariance-adjusted, weighted, and kernel-based versions of spatial depth functions are presented to classify multivariate outcomes. We extend the maximal depth classification notions for the covariance-adjusted, weighted, and kernel-based spatial depth versions. The classifiers' performance is considered and compared with some existing classification methods using simulated and real datasets.

The Paramaters of Visual Art Assessment as a Measurement Instrument for Learning Products of Art Painting, Sculpture, Ceramic and Batik

Assessing learning products of visual art was unlike evaluating other learning products. Due to its relation to visual outcomes, the first assessment required proper sensitiveness and measurable references. The research was intended to be the effort to acquire parameters of visual art assessment, especially of art painting, sculpture, ceramic, and batik assessments. The findings could be used as instruments in various visual art learning activities. Data were collected by distributing a list, interviewing the informants, and having Focus Group Discussion. The respondents were visual art lectures in the Faculty of Arts and Literature in one of the state universities in norther island of Sumatera, Indonesia. They were excellent at visual art and equipped with required sensitiveness. Besides, they also had relevant education with art painting, sculpture, ceramic, and batik. The findings, in the form of assessment parameters, were first, parameters of visual art assessment were form similarity, proportion, spatial depth, technique, composition, content, ambiance, brightness, color harmony, and texture. Secondly, parameters of non-figurative sculpture assessment were technique mastery, proportion, smoothness, expression, volume, space, and idea authenticity. Thirdly, parameters of figurative sculpture assessment were visual anatomy, proportion, motion effect, and drapery. Fourthly, parameters of ceramic assessment were technique mastery/finishing, innovation creativity, novelty, smoothness, expression, harmony, and need (market need). Last, parameters of batik assessment were technique, authenticity, modernity, color harmony, innovativeness, need finishing, and affordability.

Two-phase accurate multiplexed structured image capture (2pAc-MUSIC)

Multiplexed Structural Image Capture (MUSIC) technique uses modulation of light from a scene prior to imaging to encode the temporal evolution of target into spatial frequency shifts, with each shift corresponding to a unique time and resulting in individual distinct snapshots. Computational-processing of the multiplexed single-shot image can recover the nanosecond-resolution scene evolution. One of the limitations is the loss of spatial resolution of the recovered image due to limitations of the low-pass spatial frequency filtering depth. Here we demonstrated that two images with complementary spatial phases can be used to enhance the spatial resolutions. For this application, 2 Phase Accurate (2pAc)-MUSIC is implemented that a scene is encoded with two complementary phases. The increased spatial information captured reduces the lower frequency noise in the Fourier domain found in single-phase MUSIC. A recovery algorithm combines the 180-degree spatial frequency shifted images into a single scene and applies a spatial filter in the Fourier domain that has a larger depth than previously possible to achieve a scene with increased spatial resolution. The 2pAc-MUSIC can effectively enhance the spatial resolution by removing the overlapping zero-order spatial frequencies.

Snow water equivalents exclusively from snow depths and their temporal changes: the Δsnow model

Abstract. Reliable historical manual measurements of snow depths are available for many years, sometimes decades, across the globe, and increasingly snow depth data are also available from automatic stations and remote sensing platforms. In contrast, records of snow water equivalent (SWE) are sparse, which is significant as SWE is commonly the most important snowpack feature for hydrology, climatology, agriculture, natural hazards, and other fields. Existing methods of modeling SWE either rely on detailed meteorological forcing being available or are not intended to simulate individual SWE values, such as seasonal “peak SWE”. Here we present a new semiempirical multilayer model, Δsnow, for simulating SWE and bulk snow density solely from a regular time series of snow depths. The model, which is freely available as an R package, treats snow compaction following the rules of Newtonian viscosity, considers errors in measured snow depth, and treats overburden loads due to new snow as additional unsteady compaction; if snow is melted, the water mass is stepwise distributed from top to bottom in the snowpack. Seven model parameters are subject to calibration. Snow observations of 67 winters from 14 stations, well-distributed over different altitudes and climatic regions of the Alps, are used to find an optimal parameter setting. Data from another 71 independent winters from 15 stations are used for validation. Results are very promising: median bias and root mean square error for SWE are only −3.0 and 30.8 kg m−2, and +0.3 and 36.3 kg m−2 for peak SWE, respectively. This is a major advance compared to snow models relying on empirical regressions, and even sophisticated thermodynamic snow models do not necessarily perform better. As such, the new model offers a means to derive robust SWE estimates from historical snow depth data and, with some modification, to generate distributed SWE from remotely sensed estimates of spatial snow depth distribution.

Stress-Drop Estimates for Induced Seismic Events in the Fort Worth Basin, Texas

ABSTRACTEarthquakes in the Fort Worth basin (FWB) have been induced by the disposal of recovered wastewater associated with extraction of unconventional gas since 2008. Four of the larger felt earthquakes, each on different faults, prompted deployment of local distance seismic stations and recordings from these four sequences are used to estimate the kinematic source characteristics. Source spectra and the associated source parameters, including corner frequency, seismic moment, and stress drop, are estimated using a modified generalized inversion technique (GIT). As an assessment of the validity of the modified GIT approach, corner frequencies and stress drops from the GIT are compared to estimates using the traditional empirical Green’s function (EGF) method for 14 target events. For these events, corner-frequency residuals (GIT−EGF) have a mean of −0.31 Hz, with a standard deviation of 1.30 Hz. We find consistent mean stress drops using the GIT and EGF methods, 9.56 and 11.50 MPa, respectively, for the common set of target events. The GIT mean stress drop for all 79 earthquakes is 5.33 MPa, similar to estimates for global intraplate earthquakes (1–10 MPa) as well as other estimates for induced earthquakes near the study area (1.7–9.5 MPa). Stress drops exhibit no spatial or temporal correlations or depth dependency. In addition, there are no time or space correlations between estimated FWB stress drops and modeled pore-pressure perturbations. We conclude that induced earthquakes in the FWB occurring on normal faults in the crystalline basement release pre-existing tectonic stresses and that stress drops on the four sequences targeted in this study do not directly reflect perturbations in pore-fluid pressure on the fault.

Acoustic impact of the human skull on transcranial photoacoustic imaging.

With balanced spatial resolution, imaging depth, and functional sensitivity, photoacoustic tomography (PAT) hold great promise for human brain imaging. However, the strong acoustic attenuation and aberration of the human skull (∼8 mm thick) are longstanding technical challenges for PAT of the human brain. In this work, we numerically investigated the impacts of the stratified human skull on photoacoustic wave propagation (i.e., the forward model) and PAT image formation (i.e., the inverse model). We simulated two representative transcranial PAT implementations: photoacoustic computed tomography (PACT) and photoacoustic macroscopy (PAMac). In the forward model, we simulated the detailed photoacoustic wave propagation from a point or line source through a digital human skull. The wave attenuation, refraction, mode conversation, and reverberation were thoroughly investigated. In the inverse model, we reconstructed the transcranial PACT and PAMac images of a point or line target enclosed by the human skull. Our results demonstrate that transcranial PAMac suffers mainly from wave reverberation within the skull, leading to prolonged signal duration and reduced axial resolution. Transcranial PACT is more susceptible to the skull's acoustic distortion, mode conversion, and reverberation, which collectively lead to strong image artifacts and deteriorated spatial resolutions. We also found that PACT with a ring-shaped transducer array shows more tolerance of the skull's adverse impacts and can provide more accurate image reconstruction. Our results suggest that incorporating the skull's geometry and acoustic properties can improve transcranial PAT image reconstruction. We expect that our results have provided a more comprehensive understanding of the acoustic impact of the human skull on transcranial PAT.

2D body-wave seismic interferometry as a tool for reconnaissance studies and optimization of passive reflection seismic surveys in hardrock environments

Despite the unrivalled spatial resolution and depth penetration of active-source seismic methods used for mineral exploration in hardrock environment, economic and environmental restrictions (e.g., source permitting) may preclude its full-scale application. In such a case, 2D passive reflection seismics can be considered a cost-effective way to perform reconnaissance-type survey and provide body-wave structural imaging using ambient-noise seismic interferometry (ANSI). This is, however, conditional to the presence of noise sources in the subsurface, for example produced by underground mining activity. Here, we propose a 2D ANSI workflow as an intermediate step prior to a full-scale 3D ANSI survey and an affordable tool in brownfield exploration, e.g., when trying to update current geological models beyond the drilled area. We test the applicability of this approach by analysing selected receiver lines from a 3D passive dataset acquired over the Kylylahti mine in Finland. Our methodology aims at choosing the optimal processing strategy at possibly lowest acquisition (2D geometry) and computational (small amount of data) cost. We address the fundamental questions in ANSI, i.e., (i) how much AN should one record and (ii) which SI processing approach should one choose. Therefore, we test different processing steps necessary to produce virtual shot gathers (VSG): preprocessing, selection of the ambient-noise portion, and selection of the method for retrieving the impulse responses between the receivers (crosscorrelation - CC, crosscoherence - CCh, multidimensional deconvolution - MDD). We conclude that trace energy normalization and high-pass filtering are the preferred preprocessing steps, while the best imaging is obtained when VSGs are retrieved using MDD applied in the noise-volume approach or CC in the event-driven approach. An event-driven approach may significantly reduce the acquisition time: for the Kylylahti dataset, using 10 events with energetic body-wave arrivals, extracted from one hour of data, was enough to provide results comparable to the results from the noise-volume approach using the complete one hour of noise.

Fractional snow-covered area: scale-independent peak of winter parameterization

Abstract. The spatial distribution of snow in the mountains is significantly influenced through interactions of topography with wind, precipitation, shortwave and longwave radiation, and avalanches that may relocate the accumulated snow. One of the most crucial model parameters for various applications such as weather forecasts, climate predictions and hydrological modeling is the fraction of the ground surface that is covered by snow, also called fractional snow-covered area (fSCA). While previous subgrid parameterizations for the spatial snow depth distribution and fSCA work well, performances were scale-dependent. Here, we were able to confirm a previously established empirical relationship of peak of winter parameterization for the standard deviation of snow depth σHS by evaluating it with 11 spatial snow depth data sets from 7 different geographic regions and snow climates with resolutions ranging from 0.1 to 3 m. An enhanced performance (mean percentage errors, MPE, decreased by 25 %) across all spatial scales ≥ 200 m was achieved by recalibrating and introducing a scale-dependency in the dominant scaling variables. Scale-dependent MPEs vary between −7 % and 3 % for σHS and between 0 % and 1 % for fSCA. We performed a scale- and region-dependent evaluation of the parameterizations to assess the potential performances with independent data sets. This evaluation revealed that for the majority of the regions, the MPEs mostly lie between ±10 % for σHS and between −1 % and 1.5 % for fSCA. This suggests that the new parameterizations perform similarly well in most geographical regions.

Cinematic Rendering in Cone Beam Computed Tomography: Advanced 3D Reconstructions of Dental and Maxillofacial Pathologies

Cinematic rendering (CR) is a novel 3D visualisation technique, which provides photorealistic image reconstructions with a high level of image details. Aim of this case series is to show the application of CR in Cone Beam Computed Tomography (CBCT) in dentomaxillofacial pathologies. Four exemplary CBCTs of clinical dentomaxillofacial cases were selected. 3D CR reconstructions were generated from the CBCT by using a vendor-provided standard CR software. Cases include 1) external tooth resorption, 2) ankylosed maxillary molar tooth, 3) giant cell-associated osteolytic lesion of the mandible, 4) unilateral cleft lip/palate with additional skeletofacial deformity. CBCTs of four patients showing dental and osseous pathologies were successfully reconstructed. Overall, a subjectively improved 3D understanding of the presented pathologies was observed. The CR images seem to present more plasticity, giving a better feeling for the spatial depth of the tissue. CR can be applied to CBCT images in dentomaxillofacial patients. The photorealistic appearance might improve the understanding of complex anatomy or pathology, could facilitate patient communication, and might be helpful for advanced medical education. We see potential in the use of CR for additional 3D visualization. The actual image diagnosis is done in the classic sectional planes. The significance of CR reconstruction for image diagnostics must be investigated in appropriate studies.

Downscaling Snow Depth Mapping by Fusion of Microwave and Optical Remote-Sensing Data Based on Deep Learning

Accurate high spatial resolution snow depth mapping in arid and semi-arid regions is of great importance for snow disaster assessment and hydrological modeling. However, due to the complex topography and low spatial-resolution microwave remote-sensing data, the existing snow depth datasets have large errors and uncertainty, and actual spatiotemporal heterogeneity of snow depth cannot be effectively detected. This paper proposed a deep learning approach based on downscaling snow depth retrieval by fusion of satellite remote-sensing data with multiple spatial scales and diverse characteristics. The (Fengyun-3 Microwave Radiation Imager) FY-3 MWRI data were downscaled to 500 m resolution to match Moderate-resolution Imaging Spectroradiometer (MODIS) snow cover, meteorological and geographic data. A deep neural network was constructed to capture detailed spectral and radiation signals and trained to retrieve the higher spatial resolution snow depth from the aforementioned input data and ground observation. Verified by in situ measurements, downscaled snow depth has the lowest root mean square error (RMSE) and mean absolute error (MAE) (8.16 cm, 4.73 cm respectively) among Environmental and Ecological Science Data Center for West China Snow Depth (WESTDC_SD, 9.38 cm and 5.36 cm), the Microwave Radiation Imager (MWRI) Ascend Snow Depth (MWRI_A_SD, 9.45 cm and 5.49 cm) and MWRI Descend Snow Depth (MWRI_D_SD, 10.55 cm and 6.13 cm) in the study area. Meanwhile, downscaled snow depth could provide more detailed information in spatial distribution, which has been used to analyze the decrease of retrieval accuracy by various topography factors.