基于颜色空间和面部特征的驾驶员面部定位方法

介绍了一种快速检测和定位驾驶员面部区域的方法。首先将驾驶员面部图像映射到YCbCr颜色空间,建立肤色高斯模型;然后通过似然图自适应阈值分割获得驾驶员面部区域,对获得的二值化面部图像进行形态学操作消除毛刺和小孔;最后对处理后的面部图像进行灰度积分投影,结合面部特征获得驾驶员的面部区域并框定。并试验证明了方法的实时有效性。     

基于眼部特征的疲劳驾驶检测算法研究

疲劳驾驶是交通事故的一个重要致因,文章从驾驶员的面部特征入手,通过肤色特性找到图像中的人脸区域,并采用积分投影法和边缘检测对驾驶员眼睛进行定位,选取眼睛区域面积与人眼图像面积的比值作为眼睛状态评判指标;建立睁眼、半闭眼、闭眼3种状态下的人眼样本,依据3σ准则得到相应的区间范围,从而识别驾驶员的眼部状态,实现一种适合于疲劳驾驶检测系统的改良算法。     

基于车窗区域定位的出租汽车驾驶员蓄须行为的自动检测技术研究

针对出租车驾驶员蓄须行为自动检测技术展开研究,提出了一套基于视频图像的、完整的自动检测方法。首先,根据复杂交通车流的特性,提出一种用车窗区域代替完整车辆的出租车检测方法,同时基于检测到的车窗区域实现对驾驶室的精准定位。其次,利用基于亮度通道的多尺度视网膜（Multi-scale Retinex, MSR） 增强的图像预处理算法实现对图像的光线平衡及细节恢复。最后,通过libfacedetection 算法的精准下巴区域提取、结合非肤色像素点提取以及灰度阈值法的正反验证来提升算法的鲁棒性。利用交通监控视频对整套方法进行测试,研究结果表明,该方法能够对出租车驾驶员蓄须行为进行识别,进而提高执法效能。     

融合Eye Map信息的疲劳检测方法的研究

为降低人眼检测过程的计算量,提出一种改进的融合EyeM ap信息的快速过滤算法。利用基于肤色信息的图像分割,得到感兴趣的人脸区域,缩小了搜索范围,减小了检测的错误接受率;然后使用Haar-Like特征进行级联AdaBoost人眼定位检测,得到潜在的人眼区域,使用EyeMap的快速算法获得准确地人眼区域,再进行二值化处理,计算眼睛的轴向外接矩形以估计出眼睑的闭合程度,结合PERCLOS评判标准进行疲劳状态分析。实验证明：改进算法能快速实时有效地识别驾驶员疲劳时的眼部状态。     

基于灰度形态学的车牌定位算法

根据车牌图像的统计特征,计算形态学滤波器的结构元素,进行图像背景估计;利用闭-开运算,处理残差图像;采用自动搜索种子区域填充算法,得到各个区域;根据区域的几何特征,判断是否为车牌区域;通过对二值化的车牌采用K-means聚类拟合直线方法进行倾斜校正,得到最终的车牌.实验结果表明:该算法简单、迅速,定位准确率高,为后继字符分割和识别奠定了基础.通过对120幅图像定位实验,总有效定位率为96.7%.用Matlab7.0实验时,每张车牌平均定位时间为2.438s,而用VC++实现时,平均定位时间仅为0.139s.     

驾驶人手机通话行为中基于图像特征决策融合的手势识别方法

为鲁棒检测自然环境中驾驶人的通话行为, 提出了一种驾驶人手机通话手势的识别方法。运用Adaboost算法检测驾驶人面部区域, 在YC_gC_r色彩空间中分别对面部肤色亮度分量和色度分量进行稀疏网格间隔采样, 由此建立了肤色的高斯分布模型; 针对驾驶室光照强度的不均匀性, 提出了肤色分量的漂移补偿算法, 建立了适应光照变化的在线肤色模型, 以准确分割左右手部肤色区域; 运用HOG算法获取手部肤色区域的2 376维HOG特征向量, 运用PCA方法将HOG特征降至400维; 同时提取手部肤色区域的PZMs特征, 并采用Relief算法筛选出权重最大的8个PZMs特征向量, 建立了融合PCA-HOG特征和Relief-PZMs特征的通话手势支持向量机分类决策。试验结果表明: 基于PCA-HOG特征的手势识别率为93.1%, 对光照变化的鲁棒性较好, 但易受到手部与头部转动的干扰; 基于Relief-PZMs特征的手势识别率为91.9%, 对于头部与手部姿态的耐受度较好, 但光照鲁棒性较差; 基于PCA-HOG和Relief-PZMs多元特征融合方法的手势识别率达到94.5%, 对光照波动、手部与头部转动等干扰条件具有较好的适应性。      

一种新的肤色区域分割提取方法

提出一种新的肤色区域分割提取方法,并详细叙述了该方法的步骤和实现算法.在人脸检测和识别项目中,应用该算法来提取各肤色区域,经大量实验说明,该方法是有效的.     

一种交互式脸部网格模型调整算法

脸部网格模型的建立是基于人脸合成技术中的关键步骤。提出一个半交互式正面人脸模型调整算法。首先利用区域增长法和模板匹配自动确定图像中人脸及各特征区域的大致位置，然后交互地提取人脸特征点的准确二维位置，最后自动确定脸部姿态和调整模型非特征点，即获得输入人脸模型。该算法简便实用，费时较少，具有实用价值。     

基于改进YCrCb颜色空间的肤色分割

人脸识别是计算机视觉、模式识别、生物特征识别、图像处理等多学科交叉的一门科学,肤色分割是人脸识别研究中一个重要的步骤。针对基于YCrCb颜色空间的肤色分割在强光下分割不准确的现状,提出了一种基于改进的YCrCb空间的肤色分割方法,本文利用改进YCrCb颜色空间提高了强光下肤色分割的准确性和适应性。实验证明该算法减弱了光线强弱对人脸肤色分割的影响,更加准确的找到人脸区域,误检率较低。     

基于改进YCrCb颜色空间的肤色分割

人脸识别是计算机视觉、模式识别、生物特征识别、图像处理等多学科交叉的一门科学,肤色分割是人脸识别研究中一个重要的步骤。针对基于YCrCb颜色空间的肤色分割在强光下分割不准确的现状,提出了一种基于改进的YCrCb空间的肤色分割方法,本文利用改进YCrCb颜色空间提高了强光下肤色分割的准确性和适应性。实验证明该算法减弱了光线强弱对人脸肤色分割的影响,更加准确的找到人脸区域,误检率较低。     

Real-time follow-up head tracking in dynamic complex environments

In the modes of both object motion and camera motion, an enhanced Camshift algorithm, which is based on suppressing similar color features of background and on joint color probability density distribution image, is proposed to real-time track head in dynamic complex environment. The system consists of face detection module, head tracking module and camera control module. When tracking fails, a self-recovery mechanism is introduced. At first the Adaboost face detector based on Haar-like features is implemented to find frontal faces, the false positive is filtered according to the skin color criterion, and the true face is used to initialize the tracking module. In hue saturation value (HSV) colorspace, the hue-saturation (H-S) histogram of face skin and the saturation-value (S-V) histogram of hair are built to produce the joint color probability density distribution image, and this is intended to realize the head tracking with arbitrary pose. During tracking, region of interest (ROI) is introduced, and the color probability density distribution of a specified background area outside the ROI is learned, similar color features in the head are suppressed according to the learning result. The background suppression step is intended to resolve the problem that the tracker maybe fails when the head is distracted by backgrounds having similar colors with the head. A closed loop control model based on speed regulation is applied to drive an active camera to center the head. Once tracking drift or failure is detected, the system stops tracking and returns to the face detection module. Our experimental results show that the presented system is well suitable for tracking head with arbitrary pose in dynamic complex environments, also the active camera can track moving head smoothly and stably. The system is computationally efficient and can run in real-time completely.     

彩色图像中基于几何不变性的数字水印

常规的数字水印算法对几何攻击无能为力,水印算法对几何攻击的鲁棒性被认为是数字水印技术走上实际应用的一个关键点.文中通过对普通的基于几何不变性数字水印算法的分析,提出了一种彩色图像中基于几何不变性的数字水印的嵌入及检测过程,并通过一系列实验证明了该数字水印算法具有很高的抗几何攻击(旋转、缩放等)特性,即提高了鲁棒性.     

公路收费站横杆状态的视频检测方法及应用

以高速公路收费站监控视频为背景,提出一种可自动分析和检测横杆运行状态的方法,从而判断车辆通过与否并统计车辆数。该方法的实现,首先对横杆的颜色特征和轮廓几何特征进行学习;然后对监控视频图像进行颜色和几何特征匹配以滤除非横杆区域;再对得到的疑似横杆区域进行修复;最后计算横杆的重心坐标及其运动状态以判断开闭状态。样本测试显示,新方法能有效克服视频场景中各种干扰因素的影响,提高车辆计数的准确性。     

Region of interest based selective encryption scheme for privacy protection in H.264 video

A selective encryption scheme for region of interest （ROI） of H.264 video is proposed to protect the personal privacy in a video. The most important part of video can be protected with less cost and operation by only encrypting the content of ROIs. Human face regions are selected as ROI and detected by using Gaussian skin color model. Independent ROI encoding is realized with the mechanism of flexible macro-block ordering （FMO）. Frames are divided into grid-like slice-groups which can be combined flexibly to form a required ROI. Both luminance component and chrominance component of the macro-blocks in ROI are modified to achieve good encryption quality and location accuracy. In the process of decryption, the encrypted area is located automatically. There is no need to transmit additional position information of ROIs to the end of decryption. The encrypted video is decrypted correctly with the secret key. JM18.4 software is employed to perform the simulation experiment. Experimental results show the accuracy and effectiveness of our scheme to encrypt and decrypt the ROIs in H.264 video.     

Deformable Registration of MR Brains with Best Features

Introduction Deformable registration is very important formedical image analysis and so far various methodshave been proposed in decades[1-4]. HAMMER reg-istration algorithm[5]defines an attribute vector in-cluding intensity, edge type, and geometric mo-ment invariants as a signature of each point, to re-duce ambiguity in correspondence matching duringthe image registration. Another characteristic ofHAMMER is the hierarchical matching mecha-nism, which helps avoid the warping being trapped…     

Model test for dynamic construction mechanical effect of large-span loess tunnel

It has been a focus of debate for a large time on construction methods for large-span loess tunnel. Reasonable construction method has much effect on stability of tunnel and construction schedule. Deformation and failure of surrounding rock are quite complex. Associating with the large-span loess tunnel of Zhengzhou—Xi’an high-speed passenger rail line in China, large scale model test with geometric proportion 1:20 is applied to study on dynamic mechanical behavior of various construction methods. They include full-face excavation with support and no support, and benching method with support. It is found that pre-deformation and stress accumulation take place ahead of working face. The effects of three construction methods are further studied, particularly in terms of tunnel displacement and stress changes. It is revealed that benching method transfers load to an unexcavated area, limits horizontal deformation, reduces stress concentration effectively, lengthens the distance between location of peak for stress concentration and working face, and consequently increases stability. The model test results not only supply theoretical foundation for determination of reasonable construction method, but also can act as reference for similar tunnel and underground engineering construction.     

一种基于贝叶斯特征的视频图像人脸检测

提出了一种基于仿射运动模型和贝叶斯理论的视频图像人脸检测方法.建立仿射运动模型进行运动估计,提取运动对象区域;对训练图像提取人脸与非人脸的统计特征,利用贝叶斯准则建立概率模型;根据贝叶斯分类器和支持向量机分类器将图像特征分为人脸类与非人脸类,从而检测出视频运动图像中的人脸区域.