High-Dimensional Dense Residual Convolutional Neural Network for Light Field Reconstruction

Source positioning based on energy time-inverse focus is a hot subject in the sphere of shallow underground source positioning. Due to the grave wave group aliasing and the complex, irregular geological structure typical of the shallow underground explosion, the reconstruction accuracy of the energy focus is low and thus the recognition of the focus is a difficult task, ultimately leading to a low accuracy of source positioning. To address the above problems, this paper proposes a method based on deep learning energy focus recognition, whereby the process of recognizing and positioning the energy focus in an energy field is made equivalent to the end-to-end feature extraction of the energy field-energy focus. The time-variant characteristics of explosive vibration signals are put to use in the construction of an adaptive time window. First, within the time window and by combining cross-correlation and autocorrelation operations, a 3D energy field image sequence in the time-space domain is produced by grouped energy synthesis; second, a densely connected 3DCNN network is built and, through multiple layer span layer splicing, a higher repetitive use is made of the focus point features in the energy field images; third, a spatial pyramid pooling network is used to extract multi-scale features from different focus areas, which helps achieve high-precision focus recognition. Finally, numerical simulations and field tests were conducted.The results demonstrated that compared with the quantum particle swarm optimization (QPSO)-based energy focus search method, the proposed one is more effectively in recognizing the coordinates of the focus in the energy field, thus allowing high-precision localization of shallow underground sources. This method is of some engineering application value in the field of underground source positioning.

show abstract

“…Content may change prior to final publication. [27]. With this network, the feature information in the 3D energy field image is extracted.…”

Section: ⅲ Methods Of Energy Field Focus Recognition and Positioningmentioning

confidence: 99%

The Underground Explosion Point Measurement Method Based on High-Precision Location of Energy Focus

Zhao

Wang

et al. 2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Choi et al [22] propose a view extrapolation method with large baselines using learned depth probability volumes together with an image refinement network. Meng et al [23] develop a learning framework based on a two-stage restoration with a 4-dimensional convolutional rsidual network for light field spatio-angular superresolution. Yeung et al [24] follow a two-step approach based on view synthesis network that first generates the whole set of novel views, and a view refinement network that retrieves spatial texture details.…”

Section: View Synthesis With Learned Representationsmentioning

confidence: 99%

Learning Fused Pixel and Feature-Based View Reconstructions for Light Fields

Shi

Jiang

Guillemot

2020

2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

In this paper, we present a learning-based framework for light field view synthesis from a subset of input views. Building upon a lightweight optical flow estimation network to obtain depth maps, our method employs two reconstruction modules in pixel and feature domains respectively. For the pixel-wise reconstruction, occlusions are explicitly handled by a disparity-dependent interpolation filter, whereas inpainting on disoccluded areas is learned by convolutional layers. Due to disparity inconsistencies, the pixel-based reconstruction may lead to blurriness in highly textured areas as well as on object contours. On the contrary, the featurebased reconstruction well performs on high frequencies, making the reconstruction in the two domains complementary. End-to-end learning is finally performed including a fusion module merging pixel and feature-based reconstructions. Experimental results show that our method achieves state-of-the-art performance on both synthetic and realworld datasets, moreover, it is even able to extend light fields' baseline by extrapolating high quality views without additional training.

show abstract

“…As depicted in Fig. 4, each HDRB is composed of two HDC layers [18] with the 3 × 3 angular receptive field. With such a structure, the angular receptive field of every HDRB will cover the entire 5 × 5 viewpoints, allowing the module to learn the complete spatio-angular structure and redundancy of a light field.…”

Section: Network Architecturementioning

confidence: 99%

“…Recently, deep learning has been proved to be a powerful technique in a wide range of applications [14], [15]. With the availability of the light field dataset [16], methods based on the convolutional neural networks (CNNs) have been successfully applied to light field super-resolution [17], [18]. Yoon et al [19] establish the first deep learning framework LFCNN for both spatial and angular super-resolution but do not exploit the correlation among adjacent views.…”

Section: Introductionmentioning

confidence: 99%

LightGAN: A Deep Generative Model for Light Field Reconstruction

et al. 2020

Self Cite

View full text Add to dashboard Cite

A light field image captured by a plenoptic camera can be considered a sampling of light distribution within a given space. However, with the limited pixel count of the sensor, the acquisition of a high-resolution sample often comes at the expense of losing parallax information. In this work, we present a learning-based generative framework to overcome such tradeoff by directly simulating the light field distribution. An important module of our model is the high-dimensional residual block, which fully exploits the spatio-angular information. By directly learning the distribution, our approach can generate both high-quality sub-aperture images and densely-sampled light fields. Experimental results on both real-world and synthetic datasets demonstrate that the proposed method outperforms other state-of-the-art approaches and achieves visually more realistic results. INDEX TERMS Light field reconstruction, view synthesis, 4D convolution, high-dimension residual block, generative adversarial networks, deep learning, computational imaging

show abstract

High-Dimensional Dense Residual Convolutional Neural Network for Light Field Reconstruction

Cited by 103 publications

References 43 publications

The Underground Explosion Point Measurement Method Based on High-Precision Location of Energy Focus

The Underground Explosion Point Measurement Method Based on High-Precision Location of Energy Focus

Learning Fused Pixel and Feature-Based View Reconstructions for Light Fields

LightGAN: A Deep Generative Model for Light Field Reconstruction

Contact Info

Product

Resources

About