基于Level Sets的实时视频跟踪算法

需积分: 9 11 浏览量更新于2024-09-12 收藏 1.02MB PDF 举报

实时跟踪使用 Level Sets Level Sets 是一种常用于图像处理和计算机视觉领域的数学工具，它可以用来实现图像分割、目标跟踪和形状变换等任务。近年来，Level Sets 方法在视频跟踪领域中的应用日益广泛，然而传统的 Level Sets 方法存在计算复杂度高、速度慢等问题。本文提出了一种实时 Level Sets 视频跟踪算法，通过使用简单的链表操作和高斯滤波来实现快速的曲线演化和边界平滑正则化。该算法可以实时跟踪视频中的目标，且可以轻松控制曲线的拓扑结构。 Level Sets 方法的基本思想是将图像分割问题转换为一个曲线演化问题，即将图像中的目标边界表示为一个曲线，然后通过曲线的演化来追踪目标的运动。传统的 Level Sets 方法使用偏微分方程来描述曲线的演化，但是这种方法存在计算复杂度高的问题。本文提出的算法使用链表操作来实现曲线的演化，而不是使用偏微分方程。这使得算法的计算速度大大加快，同时也能够轻松控制曲线的拓扑结构。此外，算法还使用高斯滤波来实现边界平滑正则化，从而提高了跟踪的准确性。在视频跟踪应用中，本文的算法可以实时跟踪视频中的目标，且可以轻松控制曲线的拓扑结构。这使得算法可以广泛应用于视频监控、自动驾驶、机器人视觉等领域。 Level Sets 方法在视频跟踪领域中的应用具有广泛的前景，包括： 1. 视频监控：Level Sets 方法可以用来实时跟踪视频中的目标，从而实现智能视频监控系统。 2. 自动驾驶：Level Sets 方法可以用来跟踪视频中的目标，以实现自动驾驶系统中的目标检测和跟踪。 3. 机器人视觉：Level Sets 方法可以用来跟踪视频中的目标，以实现机器人视觉系统中的目标检测和跟踪。本文提出的实时 Level Sets 视频跟踪算法可以广泛应用于视频监控、自动驾驶、机器人视觉等领域，具有广泛的前景和应用价值。

Real-time Tracking Using Level Sets

Yonggang Shi, W. Clem Karl

∗

Electrical and Computer Engineering Department

Boston University

Boston, MA 02215

email:{yshi,wckarl}@bu.edu

Abstract

In this paper we propose a novel implementation of

the level set method that achieves real-time level-set-based

video tracking. In our fast algorithm, the evolution of the

curve is realized by simple operations such as switching

elements between two linked lists and there is no need to

solve any partial differential equations. Furthermore, a

novel procedure based on Gaussian ﬁltering is introduced

to incorporate boundary smoothness regularization. By re-

placing the standard curve length penalty with this new

smoothing procedure, further speedups are obtained. An-

other advantage of our fast algorithm is that the topology

of the curves can be controlled easily. For the tracking of

multiple objects, we extend our fast algorithm to maintain

the desired topology for multiple object boundaries based

on ideas from discrete topology. With our fast algorithm, a

real-time system has been implemented on a standard PC

and only a small fraction of the CPU power is used for

tracking. Results from standard test sequences and our real-

time system are presented.

1. Introduction

Real-time tracking of object boundaries is an impor-

tant task in many vision applications such as video surveil-

lance, video conferencing, and human-computer interac-

tion, etc. Parametric contours have been applied success-

fully to achieve real-time performance[12, 7, 6, 19], but

they have difﬁculties in handling topological changes such

as the merging and splitting of object regions. For this pur-

pose, the level set method[15] is a more powerful technique

and various models have been proposed[2, 16, 4, 14, 9, 21].

But the high computational cost of the level set method has

∗

This work was partially supported by The Air Force Ofﬁce of Scien-

tiﬁc Research under Grant F49620-03-1-0257, The National Institutes of

Health under Grant NINDS 1 R01 NS34189, and The Engineering research

centers program of the NSF under award EEC-9986821

limited its popularity in real-time scenarios. In this pa-

per, we propose a novel approach to implement the level

set method. Our approach does not need to solve any par-

tial differential equations (PDEs), thus reducing the compu-

tation dramatically compared with optimized narrow band

techniques proposed before. With our approach, real-time

level-set-based video tracking can be achieved.

Tracking models using level sets can be classiﬁed as

edge-based or region-based. In [16], a geodesic model that

combines motion and edge information was proposed. Such

edge-based models can be traced back to the snake model

in [13]. Based on morphing images, an early work on

region-based tracking was proposed in [2]. Using the dif-

ference between the current frame and a reference back-

ground, a region-based model was proposed in [4]. By

generalizing the region competition[22] idea, statistical ap-

proaches become quite popular recently. In [14, 21], the

feature distributions of both the object and background re-

gions were used for tracking. In [9], a predeﬁned distri-

bution for the object region was tracked by minimizing a

Kullback-Leibler or Bhattacharyya distance.

Considerable work has been done to improve the speed

of level-set-based curve evolution. The basic idea has been

to limit the solution of the level set PDE to a narrow band

around the zero level set[1, 17, 20, 16], but issues such as

narrow band construction, reinitialization and step size con-

trol still make existing level set methods computationally

too expensive for real-time tracking.

The novel implementation we propose in this paper

avoids the above problems. Our approach is based on the

key observation that implicitly represented curves can be

moved by simply switching elements between two linked

lists. We also propose a Gaussian ﬁltering process to replace

curvature-based smoothing, and this further speeds up our

algorithm. Another advantage of our method is that ideas

for controlling topological changes in [11] can be incorpo-

rated into our algorithm easily, which can be important for

the tracking of multiple objects and medical imaging appli-

cations. With our fast level set implementation, we have

下载后可阅读完整内容，剩余7页未读，立即下载

SYUANBOL

粉丝: 0
资源: 14

基于Level Sets的实时视频跟踪算法

A Twofold Siamese Network for Real-Time Object Tracking

Adaptive background mixture models for real-time tracking(stauffer_cvpr98_track)代码

Adaptive Color Attributes for Real-Time Visual Tracking算法流程

In Defense of Color-based Model-free Tracking

towards real-time multi-object tracking

帮翻译《BlazePose: On-device Real-time Body Pose tracking》这篇论文

mediapipe hands: on-device real-time hand tracking

flow-shift tracking

搜索基于openmv的英文文献

分别详细介绍以下的GCC编译选项的功能原理： -fno-var-tracking-assignments-toggle -fno-var-tracking-uninit -fvariable-expansion-in-unroller -fno-tree-partial-pre -funconstrained-commons -fno-unroll-all-loops -funroll-loops -funsafe-math-optimizations -fno-vpt

最新资源