Single-shot high-precision 3D reconstruction with color fringe projection profilometry based BP neural network

Zhang, Bingwei; Lin, Shaoning; Lin, Jun; Jiang, Kaiyong

doi:10.1016/j.optcom.2022.128323

Cited by 18 publications

(8 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhang B et al proposed a single-pass high-precision BP model with color stripe projection profilometry to lessen the impact of nonlinear factors. The performance of the model was verified through experiments [10]. Carlos et al proposed a particle swarm optimization to study the characteristic variables that determine the resonant frequency of the transducer and proved that the algorithm had strong robustness and high efficiency through experiments [11].…”

Section: Related Wordsmentioning

confidence: 94%

See 1 more Smart Citation

The evaluation of marketing competitiveness of B2B E-commerce enterprises based on optimized deep learning networks

Zhang

2023

Int. J. Simul. Multidisci. Des. Optim.

View full text Add to dashboard Cite

In the context of e-commerce trade between enterprises, there are problems such as price wars, false advertising, and unreasonable operations aimed at seizing market share. However, traditional estimation methods cannot provide a reasonable evaluation of the marketing competitiveness of e-commerce enterprises from an overall perspective. To help enterprises better clarify their own position and quantify their marketing competitiveness, a competitiveness evaluation model based on optimized deep learning networks is proposed. By combining subjective and objective evaluation methods, the indicators that affect the marketing competitiveness of enterprises are assigned to obtain the final competitiveness value. The research outcomes expressed that the max absolute error of the model constructed by the research institute was 0.0006, the max relative error was 0.0045, and the model accuracy was 99.85%. In the secondary indicator experiment of marketing competition, the model determined that the turnover rate of fixed assets had the greatest impact, with a weight value of 0.1263. Nine companies were randomly selected for market value estimation, and the average relative error of the model was 14.90%, which was lower than the mean relative error of the relative valuation, cash flow and absolute valuation methods, with numerical differences of 8.03%, 2.94%, and 0.12%, respectively. The research findings illustrated that the model constructed by the research institute had good performance and certain reference values for evaluating the marketing competitiveness of e-commerce enterprises.

show abstract

Section: Related Wordsmentioning

confidence: 94%

“…In formula (9), x k (p) indicates the weighted input of neuron k. e k (p) refers to the error of neuron k. If the neuron k is located in the input layer, its error calculation is formula (10).…”

Section: Model Construction Of Competitiveness Evaluation Indicators ...mentioning

confidence: 99%

The evaluation of marketing competitiveness of B2B E-commerce enterprises based on optimized deep learning networks

Zhang

2023

Int. J. Simul. Multidisci. Des. Optim.

View full text Add to dashboard Cite

show abstract

“…Yao et al [12] employed dense connections in neural networks to achieve high accuracy in learning the wrapped phase from a single low-resolution fringe pattern. Zhang et al [13] adapted the concept of the back propagation (BP) neural network to the field of fringe analysis, and successfully verified the capability to learn the wrapped phase from a single color fringe pattern. Qian et al [14] implemented the concept of stereo phase unwrapping with the utilization of a neural network, and enabling direct learning of both the wrapped phase and phase order from the fringe pattern.…”

Section: Introductionmentioning

confidence: 99%

Depth acquisition from dual-frequency fringes based on end-to-end learning

Wu,

Wang,

Liu

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

The end-to-end networks have been successfully applied in fringe projection profilometry in recent years for their high flexibility and fast speed. Most of them can predict the depth map from a single fringe. But the depth map inherits the fringe fluctuation and loses the local details of the measured object. To address this issue, an end-to-end network based on double spatially frequency fringes (dual-frequency based depth acquisition network) is proposed. To release the periodic error of the predicted depth map, a dual-branch structure is designed to learn the global contour and local details of the measured object from dual-frequency patterns. To fully exploit the contextual information of the fringe patterns, five novel modules are proposed to accomplish feature extraction, down-sampling/up-sampling, and information feeding. Ablation experiments verify the effectiveness of the presented modules. Competitive experiments demonstrate that the proposed lightweight network presents higher accuracy compared to the existing end-to-end learning algorithms. Noise immunity test and physical validation demonstrate the generalization of the network.

show abstract

“…The composite fringe pattern provides additional information that helps alleviate ambiguity, although it may lead to reduced robustness when dealing with surfaces exhibiting texture variations. Similarly, a single-shot color-composited fringe pattern was employed for absolute 3D shape acquisition [22,23] Single-shot phase extraction methods, such as Fourier transform profilometry (FTP) [24][25][26], windowed Fourier transform [27], continuous wavelet transform [28,29], Hilbert transformation [30], and empirical mode decomposition [31][32][33], have gained popularity, particularly for dynamic scenes. These methods are favored because they avoid the introduction of additional errors.…”

Section: Introductionmentioning

confidence: 99%

Robust structured light 3D imaging with two fringe patterns using recurrent classification neural network

Yang,

Liu,

Tang

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

Robust and accurate 3D reconstruction using a limited number of fringe patterns has posed a challenge in the field of structured light 3D imaging. Unlike traditional approaches that rely on multiple fringe patterns, using only one or two patterns makes phase recovery and unwrapping difficult. To address this issue, a recurrent classification neural network (RCNN) has been developed, transforming the phase recovery and unwrapping tasks into a unified phase classification task. First, a training dataset consisting of 1,200 groups of data was collected to generate a total of 38,400 training samples, enabling the RCNN to learn the mapping between the input fringe patterns and the corresponding label maps. Then, based on the well-trained network, a label map is generated based on the input two fringe patterns using the output classification results. Finally, 3D reconstruction data could be obtained by combining the inferred label map with the vision system's parameters. A series of comprehensive experiments have been conducted to validate the performance of the proposed method.

show abstract

Single-shot high-precision 3D reconstruction with color fringe projection profilometry based BP neural network

Cited by 18 publications

References 24 publications

The evaluation of marketing competitiveness of B2B E-commerce enterprises based on optimized deep learning networks

The evaluation of marketing competitiveness of B2B E-commerce enterprises based on optimized deep learning networks

Depth acquisition from dual-frequency fringes based on end-to-end learning

Robust structured light 3D imaging with two fringe patterns using recurrent classification neural network

Contact Info

Product

Resources

About