Binocular depth estimation method based on multi-layer guiding feature fusion

Yunxuan Liu; Kai Yang; Jinlong Li; Zijian Bai

doi:10.1117/12.3005955

24 November 2023 Binocular depth estimation method based on multi-layer guiding feature fusion

Yunxuan Liu, Kai Yang, Jinlong Li, Zijian Bai

Proceedings Volume 12935, Fourteenth International Conference on Information Optics and Photonics (CIOP 2023); 129351D (2023) https://doi.org/10.1117/12.3005955
Event: Fourteenth International Conference on Information Optics and Photonics (CIOP 2023), 2023, Xi’an, China

Abstract

3D convolution based stereo matching network has a wide range of research prospects at present, such as 3D measurement, unmanned driving, etc., but there is still room for improvement in accuracy. This paper proposes a threedimensional matching method based on deep learning: In the feature extraction part, a multi-layer learning parameter guiding feature fusion module is proposed, which can preserve the pixel gradient of the edge when sampling under single channel image guide. Then, the instance whitening noise of the output feature map is calculated, which effectively eliminates image pixel shift and feature similarity through the covariance threshold. In addition to using the traditional SmoothL1 loss function, the algorithm calculates the stereo focus loss by designing the confidence detection network to adjust the cost volume. The algorithm is tested on SceneFlow and Kitti series datasets. Using a multi-layer guiding module, instance bleaching loss, and stereo focus loss simultaneously compared to the original version, the error between test result and Ground Truth in the first frame (D1 Loss) of are reduced by 30.6%(Kitti2015), and the three-pixel error (3PE) is reduced by 6.3%(Kitti2012), which verifies the effectiveness of the algorithm.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Yunxuan Liu, Kai Yang, Jinlong Li, and Zijian Bai "Binocular depth estimation method based on multi-layer guiding feature fusion", Proc. SPIE 12935, Fourteenth International Conference on Information Optics and Photonics (CIOP 2023), 129351D (24 November 2023); https://doi.org/10.1117/12.3005955

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
12 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Convolution

Feature extraction

Feature fusion

Image filtering

Image fusion

Binocular vision

Deep learning

Keywords/Phrases

Search In:

Publication Years