基于类云模型聚类的多目标数据关联算法

针对多目标跟踪中的数据关联问题,提出一种基于类云模型c-均值聚类的数据关联算法.该算法采用类云模型c-均值聚类算法对目标有效回波进行聚类,将聚类中心作为目标最终观测值,运用最近邻法对聚类中心与航迹进行关联,用Kalman滤波器进行状态估计.实验结果表明,本算法与联合概率数据互联算法相比,跟踪精度高,计算量小,更适应于工程应用.     

自动人脸跟踪方法研究

针对实时视频监控领域中传统的Camshift算法不能自动跟踪人脸和容易受到肤色相近遮挡等问题,首先采用Adaboost算法实现了人脸的自动检测,同时对于跟踪丢失等情形,通过贪心预测和卡尔曼预测对跟踪偏差进行实时改进,并比较两种算法的优缺点。实验表明前者对跟踪的准确性有较大提高,后者具有较好的实时性,在相近肤色遮挡时仍能实现正确跟踪,并对侧脸也有较好的效果,算法具有较好的鲁棒性。     

多通道Haar-like特征多示例学习目标跟踪

目的 提出一种基于多通道Haar-like特征的多示例学习目标跟踪算法,克服了多示例跟踪算法在处理彩色视频时利用信息少和弱特征不能更换的缺点。方法 首先,针对原始多示例学习跟踪算法对彩色视频帧采用单通道信息或将其简单转化为灰度图像进行跟踪会丢失部分特征信息的缺点,提出在RGB三通道上生成位置、大小和通道完全随机的Haar-like特征来更好地表示目标。其次,针对多示例学习跟踪算法中Haar-like弱特征不能更换,难以反映目标自身和外界条件变化的特点,提出在弱分类器选择过程中,用随机生成的新Haar-like特征实时替换部分判别力最弱的Haar-like特征,从而在目标模型中引入新的信息,以适应目标外观的动态变化。结果 对8个具有挑战性的彩色视频序列的实验结果表明,与原始多示例学习跟踪算法、加权多示例学习跟踪算法、基于分布场的跟踪算法相比,提出的方法不仅获得了最小的平均中心误差,而且平均跟踪准确率比上述3种算法分别高52.85%,34.75%和5.71%,在4种算法中获得最优性能。结论 通过将Haar-like特征从RGB三通道随机生成,并将判别力最弱的部分Haar-like弱特征实时更换,显著提升了原始多示例学习跟踪算法对彩色视频的跟踪效果,扩展了其应用前景。     

Logistic视频字幕增强模型

目的：为提高复杂背景下的视频字幕在OCR中的识别率,需要对提取的视频字幕进行有效地字幕增强。该文首次将Logistic模型应用到视频字幕增强中,提出了基于Logistic模型的融合多帧信息的视频字幕增强方法。方法：对字幕进行检测与跟踪,将出现在连续多帧中的同一字幕片段进行对齐;通过分析字幕片段在多帧中信息,提出字幕背景在时域上的变化特征、背景和字幕文本的固有特征,并将三个特征进行量化与融合,构建适用于字幕增强的Logistic模型,实现对视频字幕的增强。结果：对含阴影或描边效果的特殊复杂背景字幕、普通复杂背景字幕、单一背景字幕分别进行实验,增强后的字幕在OCR软件中的识别正确率分别为81.76%、97.13%、98.19%,与对比方法比较均有一定的提高。结论：实验结果表明,该文方法既可以降低字幕背景的复杂度,又可以提高字幕背景与文本的对比度,从而可以对复杂背景和单一背景下的视频字幕进行有效地增强。     

Biologically adaptive robust mean shift algorithm with Cauchy predator-prey BBO and space variant resolution for unmanned helicopter formation

Visual tracking technology can provide measurement information for unmanned helicopter formation and thus, more attention is being paid to this research area. We propose a novel mean shift （MS） algorithm that is both adaptive and robust for unmanned helicopter formation and apply it to the leading unmanned helicopter tracking. The movement of an unmanned helicopter is very flexible and changeable, which makes the tracking there of more difficulty than for common targets. In creating an algorithm that can adapt to the acceleration of the unmanned helicopter and estimates both the scale and orientation of the movement changes, we combine the traditional MS with the bio-inspired Cauchy predator-prey biogeography-based optimization （CPPBBO） evolutionary algorithm, and also the space variant resolution （SVR） mechanism of the human visual system （MS-CPPBBO-SVR）. To demonstrate the effectiveness and robustness of the proposed method and justify the importance of the CPPBBO algorithm and SVR mechanism at the same time, a series of comparative exper- iments were carried out. The experimental results of the proposed MS-CPPBBO-SVR method are compared with other competitive tracking methods, such as MS, MS with SVR （MS-SVR）, MS-SVR with several other optimization algorithms, and the robust particle filter algorithm. The experimental results demonstrate that our proposed tracking approach, MS-CPPBBO-SVR, is more adaptive, robust and efficient in target tracking than the other methods.     

Multipath effects on vector tracking algorithm for GNSS signal

The vector tracking algorithm uses a single extended Kalman filter （EKF） to predict the time-delays and the Doppler deviations of the GNSS signal, while also estimating the user＇s position, velocity, and clock state. In this paper, the effects of multipath on the tracking performance of the vector delay / frequency lock loop （VDFLL） is studied for better application in the multipath environment. The error expressions of the measurements are given in theory. The tracking error caused by multipath is reduced by VDFLL, which is proved by the tracking error of VDFLL through a new iterating method. The theoretical analysis is verified by the Monte Carlo simulation.     

Adaptive neural network control for a class of continuous stirred tank reactor systems

In this paper, the control problem of continuous stirred tank reactors （CSTR） is studied. The considered CSTR are required to contain unknown functions and unknown dead zone input. An adaptive controller that uses the neural networks （NNs） is provided to solve the unknown terms. The proposed approach overcomes the effect of the dead zone input. The dead zone input in the systems is compensated for by introducing a new Lyapunov form and Young＇s inequality. The backstepping procedure is exploited to implement controller design with adaptation laws. The stability is analyzed using Lyapunov method. The performance is examined for CSTR to confirm the effectiveness of the proposed approach based on computer simulation.     

融入时序和速度信息的自适应更新目标跟踪

目的 针对目标跟踪算法在现实场景的遮挡、光照变化和尺度变化等问题,提出一种融入时序信息和速度信息的多特征融合自适应模型更新目标跟踪算法。方法 通过提取目标的分级深度特征和手工设计方向梯度直方图（histogram of oriented gradients,HOG）特征,以全深度特征组合和深层深度特征与手工设计特征组合的方式构造两个融合特征器,提高在复杂场景下跟踪的稳健性;对融合特征进行可信度计算,选择最可靠融合特征对当前帧目标进行跟踪;在跟踪质量不可靠时,对目标表征模型进行更新,加入时间上下文信息和当前鲁棒表征信息,通过多峰值判定和运动速度判定选择最优目标预测位置作为最终结果。结果 在OTB（object tracking benchmark）2013和OTB2015数据库上进行大量测试,与其他7个算法相比,本文算法总体效果取得最优,且在不同复杂环境下也取得了优秀的跟踪效果,在OTB13和OTB15数据库中,跟踪精度分别为89.3%和83.3%,成功率分别为87%和78.3%。结论 本文算法利用深度特征与手工设计特征进行融合,对跟踪结果进行多峰值分析和运动速度判定,跟踪结果不佳时自适应更新特征进行重跟踪。实验结果表明,本文算法可以有效处理光照变化、背景杂波和遮挡等复杂因素的干扰,有效提升了跟踪质量。     

基于注意力机制和卡尔曼滤波的多目标跟踪

为了解决目前多目标跟踪算法在行人遮挡后无法再次准确跟踪的问题,提出了一种融入注意力机制和改进卡尔曼滤波的多目标跟踪算法,选用联合检测和重识别框架提取特征,同时完成目标检测和重识别任务.设计了并行支路注意力机制,包括空间注意力和通道注意力两个部分,每个部分都采用并行支路的方式完成池化和卷积操作.在跟踪阶段,本文提出了速度先验卡尔曼滤波,实现对行人运动状态更精确的预测.采用CUHK-SYSU数据集对算法进行训练,并在MOT16数据集上做算法的验证和测试.本算法的多目标跟踪准确度(MOTA)达到了65.1％,多目标跟踪精确度(MOTP)达到了78.8％,识别F1值(IDF1)达到62.3％.实验表明,提出的跟踪算法可以有效地提高跟踪的整体性能,实现对目标的持续跟踪.