GUMS: A generalized unified model for stereo omnidirectional vision (demonstrated via a folded catadioptric system)

Valenti, Roberto; Xiao, Jizhong

doi:10.1109/iros.2016.7759393

Cited by 6 publications

(2 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are four main types of panoramic vision cameras: pan-tilt rotating, fisheye lens, multi-camera stitching, and catadioptric. In particular, a catadioptric panoramic camera uses a special type of mirror, called a catadioptric mirror, to direct light from different angles onto a single image sensor, thus capturing panoramic images (Jaramillo et al, 2016 ). Consisting mainly of a convex reflecting mirror, an imaging lens, and a photosensitive component (Baker and Nayar, 1998 , 1999 ), the catadioptric panoramic camera avoids the complicated designs associated with optical lens structures and solves the problem of image distortion (Liu et al, 2016 ).…”

Section: Introductionmentioning

confidence: 99%

Monocular catadioptric panoramic depth estimation via improved end-to-end neural network model

Yan,

Liu,

Ding

et al. 2023

Front. Neurorobot.

View full text Add to dashboard Cite

In this paper, we propose a monocular catadioptric panoramic depth estimation algorithm based on an improved end-to-end neural network model. First, we use an enhanced concentric circle approximation unfolding algorithm to unfold the panoramic images captured by the catadioptric panoramic camera and then extract the effective regions. In addition, the integration of the Non-local attention mechanism is exploited to improve image understanding. Finally, a depth smoothness loss strategy is implemented to further enhance the reliability and precision of the estimated depths. Experimental results confirm that this refined algorithm is capable of providing highly accurate estimates of object depth.

show abstract

Section: Introductionmentioning

confidence: 99%

Monocular catadioptric panoramic depth estimation via improved end-to-end neural network model

Yan,

Liu,

Ding

et al. 2023

Front. Neurorobot.

View full text Add to dashboard Cite

show abstract

“…Depth is recovered by triangulating corresponding features on the images. Catadioptric omnidirectional stereovision can be obtained by using a single camera with at least two mirrors [15,16], two cameras and a single mirror [17], two cameras and two mirrors [18,19]. These configurations, however, define passive systems and are mainly used for robotic applications.…”

Section: Introductionmentioning

confidence: 99%