A zoom tracking algorithm based on deep learning

Wang, Xuanyin; Ji, Jiayu; Zhu, Yanyu

doi:10.1007/s11042-021-10868-2

Cited by 2 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since first researches in the 1990's, the control law has been improved to keep the element of interest, which is zoomed-in, in sharp focus (Zou et al 2012). The most recent algorithms even rely on deep-learning methods (Wang et al 2021) but the zooming requires about 5 seconds to automatically zoom-in an object, thus not suitable for real-time tracking yet.…”

Section: Adaptive Optics: Moving Lensesmentioning

confidence: 99%

“…Several adaptations of dioptric cameras (Sec. 5), like zooming and focusing, are well-known and, therefore, used in quantity of applications, including surveillance, animal observation, artistic photography and fashion (Maître 2017;Wang et al 2021;Roberts 1995). In entertainment and arts, two camera adaptations are able to make real landscapes like models and the other way around.…”

Section: Applications Of Adaptive Camerasmentioning

confidence: 99%

See 1 more Smart Citation

A Survey on Adaptive Cameras

Ducrocq,

Caron

2024

Int J Comput Vis

View full text Add to dashboard Cite

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

show abstract

Section: Adaptive Optics: Moving Lensesmentioning

confidence: 99%

Section: Applications Of Adaptive Camerasmentioning

confidence: 99%

A Survey on Adaptive Cameras

Ducrocq,

Caron

2024

Int J Comput Vis

View full text Add to dashboard Cite

show abstract

A Novel Fuzzy Controller for Visible-Light Camera Using RBF-ANN: Enhanced Positioning and Autofocusing

Zhou

Xue

et al. 2022

Sensors

View full text Add to dashboard Cite

To obtain high-precision for focal length fitting and improve the visible-light camera autofocusing speed, simultaneously, the backlash caused by gear gaps is eliminated. We propose an improved RBF (Radical Basis Function) adaptive neural network (ANN) FUZZY PID (Proportional Integral Derivative) position closed-loop control algorithm to achieve the precise positioning of zoom and focus lens groups. Thus, the Levenberg–Marquardt iterative algorithm is used to fit the focal length, and the improved area search algorithm is applied to achieve autofocusing and eliminate backlash. In this paper, we initially adopt an improved RBF ANN fuzzy PID control algorithm in the position closed-loop in the visible-light camera position and velocity double closed-loop control system. Second, a similar triangle method is used to calibrate the focal length of the visible-light camera system, and the Levenberg–Marquardt iterative algorithm is used to fit the relation of the zoom potentiometer code values and the focal length to achieve the zoom position closed-loop control. Finally, the improved area search algorithm is used to achieve fast autofocusing and acquire clear images. The experimental results show that the ITAE (integrated time and absolute error) performance index of the improved RBF ANN fuzzy PID control algorithm is improved by more than two orders of magnitude as compared with the traditional fuzzy PID control algorithm, and the settling time is 6.4 s faster than that of the traditional fuzzy PID control. Then, the Levenberg–Marquardt iterative algorithm has a fast convergence speed, and the fitting precision is high. The quintic polynomial fitting results are basically consistent with the sixth-degree polynomial. The fitting accuracy is much better than that of the quadratic polynomial and exponential. Autofocusing requires less than 2 s and is improved by more than double that of the traditional method. The improved area search algorithm can quickly obtain clear images and solve the backlash problem.

show abstract

A zoom tracking algorithm based on deep learning

Cited by 2 publications

References 15 publications

A Survey on Adaptive Cameras

A Survey on Adaptive Cameras

A Novel Fuzzy Controller for Visible-Light Camera Using RBF-ANN: Enhanced Positioning and Autofocusing

Contact Info

Product

Resources

About