Investigating Optimal Training and Uncertainty Quantification for CNN-based Optical Flow

Optical Flow (OF) techniques provide “dense estimation” flow maps (i.e. pixel-level resolution) of timecorrelated images and thus are appealing to applications requiring high spatial resolutions. OF methods revolve around mathematical descriptions of the image as a collection of features, in which the pixel-level light intensity is the primary variable (Horn and Schunck, 1981). Feature tracking often involves the notion of scale invariance. Traditional OF approaches, merely based on mathematical formulations, have suffered from many challenges, especially when directly applied to images of fluid flows textured with tracer particles (hereafter PIV-like images). Due to the limited number of computationally manageable features and suboptimal regularization methods, successful implementation of past approaches has been limited to highly textured images and small displacement dynamic ranges.

Download Full-text

A Driver’s Visual Attention Prediction Using Optical Flow

Sensors ◽

10.3390/s21113722 ◽

2021 ◽

Vol 21 (11) ◽

pp. 3722

Author(s):

Byeongkeun Kang ◽

Yeejin Lee

Keyword(s):

Visual Attention ◽

Optical Flow ◽

Prediction Models ◽

Apparent Movement ◽

Image Sequences ◽

Motion Information ◽

Accuracy Improvement ◽

Flow Maps ◽

Current State ◽

Motion Features

Motion in videos refers to the pattern of the apparent movement of objects, surfaces, and edges over image sequences caused by the relative movement between a camera and a scene. Motion, as well as scene appearance, are essential features to estimate a driver’s visual attention allocation in computer vision. However, the fact that motion can be a crucial factor in a driver’s attention estimation has not been thoroughly studied in the literature, although driver’s attention prediction models focusing on scene appearance have been well studied. Therefore, in this work, we investigate the usefulness of motion information in estimating a driver’s visual attention. To analyze the effectiveness of motion information, we develop a deep neural network framework that provides attention locations and attention levels using optical flow maps, which represent the movements of contents in videos. We validate the performance of the proposed motion-based prediction model by comparing it to the performance of the current state-of-art prediction models using RGB frames. The experimental results for a real-world dataset confirm our hypothesis that motion plays a role in prediction accuracy improvement, and there is a margin for accuracy improvement by using motion features.

Download Full-text

A Robust Facial Feature Tracking Method Based on Optical Flow and Prior Measurement

International Journal of Cognitive Informatics and Natural Intelligence ◽

10.4018/jcini.2010100105 ◽

2010 ◽

Vol 4 (4) ◽

pp. 62-75

Author(s):

Guoyin Wang ◽

Yong Yang ◽

Kun He

Keyword(s):

Optical Flow ◽

Feature Tracking ◽

Facial Feature ◽

Recognition System ◽

Research Area ◽

Expression Recognition ◽

Tracking Method ◽

Facial Feature Tracking ◽

Recognition Result ◽

Method Of Measurement

Cognitive informatics (CI) is a research area including some interdisciplinary topics. Visual tracking is not only an important topic in CI, but also a hot topic in computer vision and facial expression recognition. In this paper, a novel and robust facial feature tracking method is proposed, in which Kanade-Lucas-Tomasi (KLT) optical flow is taken as basis. The prior method of measurement consisting of pupils detecting features restriction and errors and is used to improve the predictions. Simulation experiment results show that the proposed method is superior to the traditional optical flow tracking. Furthermore, the proposed method is used in a real time emotion recognition system and good recognition result is achieved.

Download Full-text

Generation and Control of Tracer Particles for Optical Flow Investigations in Air

Particle Image Velocimetry: Recent Improvements ◽

10.1007/978-3-642-18795-7_30 ◽

2004 ◽

pp. 417-426 ◽

Cited By ~ 4

Author(s):

C. J. Kähler ◽

B. Sammler ◽

J. Kompenhans

Keyword(s):

Optical Flow ◽

Tracer Particles ◽

And Control

Download Full-text

Consistent Optical Flow Maps for Full and Micro Facial Expression Recognition

Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications ◽

10.5220/0006127402350242 ◽

2017 ◽

Cited By ~ 4

Author(s):

Benjamin Allaert ◽

Ioan Marius Bilasco ◽

Chabane Djeraba

Keyword(s):

Facial Expression ◽

Optical Flow ◽

Facial Expression Recognition ◽

Expression Recognition ◽

Flow Maps

Download Full-text

Tailored optical propulsion forces for controlled transport of resonant gold nanoparticles and associated thermal convective fluid flows

Light Science & Applications ◽

10.1038/s41377-020-00417-1 ◽

2020 ◽

Vol 9 (1) ◽

Author(s):

José A. Rodrigo ◽

Mercedes Angulo ◽

Tatiana Alieva

Keyword(s):

Optical Trap ◽

Marangoni Effect ◽

Fluid Flows ◽

Resonance Wavelength ◽

Noble Metal Nanoparticles ◽

Gradient Force ◽

Phase Gradient ◽

Tracer Particles ◽

Propulsion Force ◽

Convective Fluid

Abstract Noble metal nanoparticles illuminated at their plasmonic resonance wavelength turn into heat nanosources. This phenomenon has prompted the development of numerous applications in science and technology. Simultaneous optical manipulation of such resonant nanoparticles could certainly extend the functionality and potential applications of optothermal tools. In this article, we experimentally demonstrate optical transport of single and multiple resonant nanoparticles (colloidal gold spheres of radius 200 nm) directed by tailored transverse phase-gradient forces propelling them around a 2D optical trap. We show how the phase-gradient force can be designed to efficiently change the speed of the nanoparticles. We have found that multiple hot nanoparticles assemble in the form of a quasi-stable group whose motion around the laser trap is also controlled by such optical propulsion forces. This assembly experiences a significant increase in the local temperature, which creates an optothermal convective fluid flow dragging tracer particles into the assembly. Thus, the created assembly is a moving heat source controlled by the propulsion force, enabling indirect control of fluid flows as a micro-optofluidic tool. The existence of these flows, probably caused by the temperature-induced Marangoni effect at the liquid water/superheated water interface, is confirmed by tracking free tracer particles migrating towards the assembly. We propose a straightforward method to control the assembly size, and therefore its temperature, by using a nonuniform optical propelling force that induces the splitting or merging of the group of nanoparticles. We envision further development of microscale optofluidic tools based on these achievements.

Download Full-text

Simple Hardware Implementation of Motion Estimation Algorithms

Avances en Ciencias e Ingeniería ◽

10.18272/aci.v11i3.1352 ◽

2019 ◽

Vol 11 (3) ◽

Author(s):

Juan Romero ◽

Damien Verdier ◽

Clement Raffaitin ◽

Luis Miguel Procel ◽

Lionel Trojman

Keyword(s):

Optical Flow ◽

Field Programmable Gate Array ◽

Hardware Implementation ◽

Software Implementation ◽

Hardware Description Language ◽

Successful Implementation ◽

Description Language ◽

Estimation Algorithms ◽

Field Programmable ◽

Hardware Description

We present in the following work a hardware implementation of the two principal optical flow methods. The work is based on the methods developed by Lucas & Kanade, and Horn & Schunck. The implementation is made by using a field programmable gate array and Hardware Description Language. To achieve a successful implementation, the algorithms were optimized. The results show the optical flow as a vector field over one frame, which enable an easy detection of the movement. The results are compared to a software implementation to insure the success of the method. The implementation is a fast implementation capable of quickly overcoming a traditional implementation in software.

Download Full-text

APPLYING FEATURE TRACKING TO PARTICLE IMAGE VELOCIMETRY

International Journal of Pattern Recognition and Artificial Intelligence ◽

10.1142/s0218001403002496 ◽

2003 ◽

Vol 17 (04) ◽

pp. 487-504 ◽

Cited By ~ 12

Author(s):

DMITRY CHETVERIKOV

Keyword(s):

Particle Image Velocimetry ◽

Optical Flow ◽

Feature Tracking ◽

Particle Image ◽

Good Alternative ◽

Flow Estimation ◽

Tracking Algorithms ◽

Alternative Approach ◽

Image Velocimetry ◽

Correlation Techniques

Particle Image Velocimetry (PIV) is a popular approach to flow visualization and measurement in hydro- and aerodynamic studies and applications. The fluid is seeded with particles that follow the flow and make it visible. Traditionally, correlation techniques have been used to estimate the displacements of the particles in a digital PIV sequence. These techniques are relatively time-consuming and noise-sensitive. Recently, an optical flow estimation technique developed in machine vision has been successfully used in Particle Image Velocimetry. Feature tracking is an alternative approach to motion estimation, whose application to PIV is proposed and studied in this paper. Two efficient feature tracking algorithms are customized and applied to PIV. The algorithmic solutions of the application are described. In particular, techniques for coherence filtering and interpolation of a velocity field are developed. To assess the proposed and the previous approaches, velocity fields obtained by the different methods are quantitatively compared for numerous synthetic and real PIV sequences. It is concluded that the tracking algorithms offer Particle Image Velocimetry a good alternative to both correlation and optical flow techniques.

Download Full-text