Shengnan Zheng

Shengnan Zheng

Auto-generated profile, based on publicly available information

Shengnan Zheng

College of Information Science and Engineering, Hohai University, Nanjing 211100, China

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 September 2021 in Sensors
Verified
Journal Article
Open access
PDF file
LHLei Han
Lei Han
XHXiaohua Huang
Xiaohua Huang
ZSZhan Shi
Zhan Shi
SZShengnan Zheng
Shengnan Zheng
Depth estimation based on light field imaging is a new methodology that has succeeded the traditional binocular stereo matching and depth from monocular images. Significant progress has been made in light-field depth estimation. Nevertheless, the balance between computational time and the accuracy of depth estimation is still worth exploring. The geometry in light field imaging is the basis of depth estimation, and the abundant light-field data provides convenience for applying deep learning algorithms. The Epipolar Plane Image (EPI) generated from the light-field data has a line texture containing geometric information. The slope of the line is proportional to the depth of the corresponding object. Considering the light field depth estimation as a spatial density prediction task, we design a convolutional neural network (ESTNet) to estimate the accurate depth quickly. Inspired by the strong image feature extraction ability of convolutional neural networks, especially for texture images, we propose to generate EPI synthetic images from light field data as the input of ESTNet to improve the effect of feature extraction and depth estimation. The architecture of ESTNet is characterized by three input streams, encoding-decoding structure, and skipconnections. The three input streams receive horizontal EPI synthetic image (EPIh), vertical EPI synthetic image (EPIv), and central view image (CV), respectively. EPIh and EPIv contain rich texture and depth cues, while CV provides pixel position association information. ESTNet consists of two stages: encoding and decoding. The encoding stage includes several convolution modules, and correspondingly, the decoding stage embodies some transposed convolution modules. In addition to the forward propagation of the network ESTNet, some skip-connections are added between the convolution module and the corresponding transposed convolution module to fuse the shallow local and deep semantic features. ESTNet is trained on one part of a synthetic light-field dataset and then tested on another part of the synthetic light-field dataset and real light-field dataset. Ablation experiments show that our ESTNet structure is reasonable. Experiments on the synthetic light-field dataset and real light-field dataset show that our ESTNet can balance the accuracy of depth estimation and computational time.

21 March 2016 in Signal, Image and Video Processing
Verified
Journal Article
Closed access
PDF file
LHLei Han
Lei Han
CHChenrong Huang
Chenrong Huang
SZShengnan Zheng
Shengnan Zheng
ZZZhen Zhang
Zhen Zhang
LXLizhong Xu
Lizhong Xu

16 September 2013 in Journal of Computer Applications
Verified
Journal Article
Closed access
LHLei Han
Lei Han
CHChenrong Huang
Chenrong Huang
MXMengxi Xu
Mengxi Xu
SZShengnan Zheng
Shengnan Zheng

14 January 2025 in Biomimetics
Verified
Journal Article
Open access
PDF file
SZSheng Zhang
Sheng Zhang
KLKe Li
Ke Li
ZLZhonghua Luo
Zhonghua Luo
MXMengxi Xu
Mengxi Xu
SZShengnan Zheng
Shengnan Zheng
(1) Background: At present, the bio-inspired visual neural models have made significant achievements in detecting the motion direction of the translating object. Variable contrast in the figure-ground and environmental noise interference, however, have a strong influence on the existing model. The responses of the lobula plate tangential cell (LPTC) neurons of Drosophila are robust and stable in the face of variable contrast in the figure-ground and environmental noise interference, which provides an excellent paradigm for addressing these challenges. (2) Methods: To resolve these challenges, we propose a bio-inspired visual neural model, which consists of four stages. Firstly, the photoreceptors (R1–R6) are utilized to perceive the change in luminance. Secondly, the change in luminance is divided into parallel ON and OFF pathways based on the lamina monopolar cell (LMC), and the spatial denoising and the spatio-temporal lateral inhibition (LI) mechanisms can suppress environmental noise and improve motion boundaries, respectively. Thirdly, the non-linear instantaneous feedback mechanism in divisive contrast normalization is adopted to reduce local contrast sensitivity; further, the parallel ON and OFF contrast pathways are activated. Finally, the parallel motion and contrast pathways converge on the LPTC in the lobula complex. (3) Results: By comparing numerous experimental simulations with state-of-the-art (SotA) bio-inspired models, we can draw four conclusions. Firstly, the effectiveness of the contrast neural computation and the spatial denoising mechanism is verified by the ablation study. Secondly, this model can robustly detect the motion direction of the translating object against variable contrast in the figure-ground and environmental noise interference. Specifically, the average detection success rate of the proposed bio-inspired model under the pure and real-world complex noise datasets was increased by 5.38% and 5.30%. Thirdly, this model can effectively reduce the fluctuation in this model response against variable contrast in the figure-ground and environmental noise interference, which shows the stability of this model; specifically, the average inter-quartile range of the coefficient of variation in the proposed bio-inspired model under the pure and real-world complex noise datasets was reduced by 38.77% and 47.84%, respectively. The average decline ratio of the sum of the coefficient of variation in the proposed bio-inspired model under the pure and real-world complex noise datasets was 57.03% and 67.47%, respectively. Finally, the robustness and stability of this model are further verified by comparing other early visual pre-processing mechanisms and engineering denoising methods. (4) Conclusions: This model can robustly and steadily detect the motion direction of the translating object under variable contrast in the figure-ground and environmental noise interference.