3D Pose Estimation by Directly Matching Polyhedral Models to Gray Value Gradients

This contribution addresses the problem of pose estimation and tracking of vehicles in image sequences from traffic scenes recorded by a stationary camera. In a new algorithm, the vehicle pose is estimated by directly matching polyhedral vehicle models to image gradients without an edge segment extraction process. The new approach is significantly more robust than approaches that rely on feature extraction since the new approach exploits more information from the image data. We successfully tracked vehicles that were partially occluded by textured objects, e.g., foliage, where a previous approach based on edge segment extraction failed. Moreover, the new pose estimation approach is also used to determine the orientation and position of the road relative to the camera by matching an intersection model directly to image gradients. Results from various experiments with real world traffic scenes are presented.     

基于三帧差法和交叉熵阈值法的车辆检测

考虑运动车辆图像的时间相关性和空间相关性,采用三帧差分法与二维交叉熵阈值法相结合的方法对运动车辆进行检测。利用三帧差分法对视频图像进行运动车辆的检测,得到灰度图像的差分图像,使用二维交叉熵阈值法对差分图像进行二值化。实验结果证明,该方案能够检测出完整的运动车辆,具有良好的可靠性和鲁棒性,满足智能交通系统的要求。     

多任务分段紧凑特征的车辆检索方法

目的 随着公共安全领域中大规模图像监控及视频数据的增长以及智能交通的发展,车辆检索有着极其重要的应用价值。针对已有车辆检索中自动化和智能化水平低、难以获取精确的检索结果等问题,提出一种多任务分段紧凑特征的车辆检索方法,有效利用车辆基本信息的多样性和关联性实现实时检索。方法 首先,利用相关任务之间的联系提高检索精度和细化图像特征,因此构造了一种多任务深度卷积网络分段学习车辆不同属性的哈希码,将图像语义和图像表示相结合,并采用最小化图像编码使学习到的车辆的不同属性特征更具有鲁棒性;然后,选用特征金字塔网络提取车辆图像的实例特征并利用局部敏感哈希再排序方法对提取到的特征进行检索;最后,针对无法获取查询车辆目标图像的特殊情况,采用跨模态辅助检索方法进行检索。结果 提出的检索方法在3个公开数据集上均优于目前主流的检索方法,其中在CompCars数据集上检索精度达到0.966,在VehicleID数据集上检索精度提升至0.862。结论 本文提出的多任务分段紧凑特征的车辆检索方法既能得到最小化图像编码及图像实例特征,还可在无法获取目标检索图像信息时进行跨模态检索,通过实验对比验证了方法的有效性。     

车辆红外图像的计算机仿真

真实的车辆红外图像的生成是红外场景生成和实现其虚拟漫游的关键技术，它也同时在遥感测量、交通管理、车辆设计、国防和军事等众多领域有着广泛地应用。分析了影响地面车辆红外特征的各种环境及内部因素，建立了车身表面热平衡方程，并以此提出车辆三维红外成像模型。还模拟了大气传输过程对红外成像的影响，并运用计算机图形学理论设计适合红外图像的绘制方法。最后给出了地面车辆红外图像的模拟结果。从实验结果看，该方法模拟效果真实，算法可靠。     

基于GPS与GPRS通信的车辆主动防盗报警系统

设计了一套车辆主动报警系统,借助于无线通讯网络数据通道,通过GPRS以短消息和彩信形式作为数据传输方式,实现实时防盗报警功能。系统不仅利用多传感器对车辆状况的实时监控,还增加了现场图像实时采集,并借助于GPS全球定位系统进行车辆位置数据的采集,经过多次试验验证,系统完全能满足实时车辆防盗报警功能要求。     

基于透视点和轮廓的粘连车辆分割方法

为解决视频车辆跟踪中经常出现的多辆车的前景粘连问题,本文提出了一种结合透视原理和车辆轮廓的粘连车辆分割方法。首先,通过分析路面场景中车道信息获取透视点,再利用透视点得到车辆的长方体透视模型,通过判断长方体模型的高宽比和长方体模型内部空缺区域的面积大小进行粘连检测,并根据粘连情况的不同选取相应的分割点搜索方法,从前往后顺序分割出粘连的车辆前景团块。实验表明,此方法对直线道路上多辆车粘连的分割有较好的准确度和适应性。     

Vision-based vehicle detection for a driver assistance system

On the basis of a cost-effective embedded system, this work implements a preceding vehicle detection system by using computer vision technologies. The road scenes are acquired with a monocular camera. The features of the vehicle in front are extracted and recognized by the proposed refined image processing algorithm, and a tracking process based on optical flow is also applied for reducing the complexity of computing. The system also provides the longitudinal distance information for the further function of adaptive cruise control. Moreover, voice alerts and image recording will be activated if the distance is less than the safe range. A statistical base of 100 video road images are tested in our experiments; the natures of the vehicles include sedan, minivan, truck, and bus. The experimental results show that the proportion of correct identifications of proceeding vehicles is above 95.8%, testing on highways in the daytime. Experimental results also indicate that the system correctly identifies vehicles in real time.     

Model-based object tracking in monocular image sequences of road traffic scenes

Moving vehicles are detected and tracked automatically in monocular image sequences from road traffic scenes recorded by a stationary camera. In order to exploit the a priori knowledge about shape and motion of vehicles in traffic scenes, a parameterized vehicle model is used for an intraframe matching process and a recursive estimator based on a motion model is used for motion estimation. An interpretation cycle supports the intraframe matching process with a state MAP-update step. Initial model hypotheses are generated using an image segmentation component which clusters coherently moving image features into candidate representations of images of a moving vehicle. The inclusion of an illumination model allows taking shadow edges of the vehicle into account during the matching process. Only such an elaborate combination of various techniques has enabled us to track vehicles under complex illumination conditions and over long (over 400 frames) monocular image sequences. Results on various real-world road traffic scenes are presented and open problems as well as future work are outlined.     

Intelligent control with new image processing strategy for a mobile vehicle

This paper considers the problem of how to use a camera to recognize the shape of a path without using too much image processing time. A quadrangle image scene is divided into three parts, and the center of gravity of an object in each part is extracted to estimate the shape of the path. A strategy for how to measure distances with a camera is also presented. The idea behind this strategy is first to establish a methematical model describing the relationship between pixel distance in the image scene and real distance in front of the mobile vehicle by experiments, and then to decide the relevant position of the object by means of rotation of the camera. These image processing methods can be used in control problems with mobile vehicles/car-like robots, and can be used in fuzzy control, neural control, and other control strategies. This work was presented, in part, at the 2nd International Symposium on Artificial Life and Robotics, Oita, Japan, February 18–20, 1997     

基于机器视觉的交叉口特种车辆快速通行技术研究

交叉口是道路网络中重要的交通节点,容易产生交通堵塞问题,为了在保证通行安全的情况下提高特种车辆的通行效率,研究基于机器视觉的交叉口特种车辆快速通行技术。优化通行基础采用图像采集及预处理、检测识别和通行控制作为技术框架结构,利用机器视觉技术采集交叉口实时交通图像,通过图像滤波、图像增强等步骤,实现初始图像的预处理。利用Car-YOLO网络识别交叉口通行能力,规划快速通行路线,考虑前车行驶状态,求解特种车辆通行速度,针对车辆所占车道,通过绿灯早启、绿灯周期时间延长等方式控制交叉口信号灯,实现交叉口特种车辆快速通行。实验结果表明:在拥堵和正常通行场景下,优化设计技术的特种车辆通过时间的平均值分别为18.2s、10.1s,事故发生概率分别低于2%、1.4%,具有较好的应用效果。     

视频监控系统中的非线性校正算法

在基于数字图像处理技术的视频监控系统中,车辆信息和道路信息是最为重要的参数,但由于摄像机内外参数的不同将导致图像不同程度的非线性失真,影响正确的交通参数的获得和交通现象的正确判断。基于图像网格的非线性校正算法对此能进行有效的校正,为交通视频监控提供较为准确的信息参数。实验结果表明该算法有较高的精度,并在视频监控系统中有较好地应用。     

一种基于视觉的移动机器人定位系统

具有自主的全局定位能力是自主式稳定机器人传感器系统的一项重要功能,为了实现这个目的,国内外均在不断地研究发展各种定位传感器系统,这里介绍了一种采用光学蝗全方位位置传感器系统,该传感器系统由主动式路标、视觉传感器、图象采集与数据处理系统组成,其视觉传感器和数据处理系统可安装在移动机器人上,然后可通过观测路标物「视角定位的方法,计算出机器人在世界坐标系中的位置和方向,实验证明,该系统可以只的在线定位,     

GPS车辆轨迹对地图精确性的校验

通过全球卫星定位系统（ＧＰＳ） 采集到的车辆定位信息，较精确地记录了车辆在不同时刻的位置数据。本文运用数据采集和图像细化处理的方法，从中提取出了位置比较准确的道路曲线，为数字地图的校准工作提供了精确的图源     

Automatic free parking space detection by using motion stereo-based 3D reconstruction

This paper proposes a free parking space detection system by using motion stereo-based 3D reconstruction. An image sequence is acquired with a single rearview fisheye camera and the view behind the automobile is three-dimensionally reconstructed by using point correspondences. Metric information is recovered from the camera height ratio and free parking spaces are detected by estimating the positions of adjacent vehicles. Since adjacent vehicles are usually located near the epipole, their structures are seriously degraded. To solve this problem, we select point correspondences by using a de-rotation-based method and mosaic 3D structures by estimating a similarity transformation. Unlike in previous work, our system proposes an efficient way of locating free parking spaces in 3D point clouds. Odometry is not used because its accuracy depends largely on road conditions. In the experiments, the system was tested in 154 different parking situations and its success rate was 90% (139 successes in 154 cases). The detection accuracy was evaluated by using ground truth data that was acquired with a laser scanner.     

结合HSV空间的水面图像特征水岸线检测

目的 水岸线可用于无人艇视觉导航、运动状态估计,是无人水面艇自主航行的重要参照特征。水岸线类似于水天线,但由于水波、反光、倒影等因素,内河水岸线检测的背景更复杂,难以采用现有水天线检测技术。通过分析水面图像在HSV空间的特征,发现陆地区域的饱和度值均高于天空和水面区域;在光照较暗时,色彩信息不能使用,亮度图像中的陆地区域相对其他区域较暗,但仍然存在水岸线轮廓特征。基于这一分析结果,提出结合HSV空间的水面图像特征水岸线检测方法。方法 首先将RGB图像经过高斯滤波后变换到HSV空间,依据权重进行HSV空间特征分量选取,接着进行像素点非线性增强;然后在增强的图像上进行区域分割,并将各个区域定义为基底图像;其次分析饱和度图像的行列特性,提取高饱和度的陆地区域,并将其定义为模板图像,将模板图像覆盖在各个基底图像上,按重叠区域面积比选取基底图像;最后通过边缘检测算子检测水岸线。结果 本文采集不同季节,不同光照强度的水面图像进行水岸线检测实验,实验结果表明本文算法可以在不同光照环境下准确检测出水岸线,且轮廓清晰完整,本文算法的实时性可达到1帧/s。结论 本文提出的结合HSV空间的水面图像特征水岸线检测方法,可以在不同的光照环境中有效地检测出轮廓清晰完整的水岸线,验证了水面图像分析的结论,本文算法可适用于无人艇视觉导航中。     

基于对称性的快速车辆检测方法

提出了一种基于车辆边缘对称性和水平阴影的快速车辆检测方法.通过利用被检测图像减去道路背景图像的方法对被检测图像进行预处理,以最大限度去除背景图像中车道线等大量干扰,突出车辆图像,对图像分别做Sobel垂直和水平边缘检测.把垂直边缘图像分割放缩成远景、中景、近景3幅大小相同的图像,利用车辆对称性对分割放缩后3幅图像对称性进行检测,确定图像中所有车辆在图像中的水平位置.同时在水平边缘检测的图像中,利用水平阴影和汽车底部水平边缘确定汽车的垂直位置.实验结果表明,该方法简单、有效,能快速、精准地检测出汽车图像中所有的不同位置和不同大小的车辆.     

Visual Surveillance for Moving Vehicles

An overview is given of a vision system for locating, recognising and tracking multiple vehicles, using an image sequence taken by a single camera mounted on a moving vehicle. The camera motion is estimated by matching features on the ground plane from one image to the next. Vehicle detection and hypothesis generation are performed using template correlation and a 3D wire frame model of the vehicle is fitted to the image. Once detected and identified, vehicles are tracked using dynamic filtering. A separate batch mode filter obtains the 3D trajectories of nearby vehicles over an extended time. Results are shown for a motorway image sequence.     

Comparison of operations of AGVs and ALVs in an automated container terminal

In an automated container terminal, the automated guided vehicles (AGVs) and the automated lifting vehicles (ALVs) are the most popular candidates to be used for transporting containers between the quayside and the storage yard. In this paper, we compare the operational productivities of the two types of vehicles when used in combination with the quay cranes of various performances. We assume a flexible path layout in which the vehicles can move almost freely in any vertical and horizontal directions. The traffic control scheme employed in our simulation finds a minimum- time route and schedules the travel to avoid deadlocks. Simulation experiments show that the ALVs reach the same productivity level as the AGVs using much less number of vehicles due to its self-lifting capability. However, the results also reveal that the AGVs eventually catch up the performance of the ALVs in most cases if the number of vehicles given is large enough. An exception is when the tandem double-trolley QCs are used for loading, in which case the AGVs cannot catch up the ALVs no matter how many more vehicles are added.     

基于高斯混合模型和AdaBoost的夜间车辆检测

针对夜间车辆检测精度相对不高的问题，提出通过构建车头灯对空间几何关系的高斯混合模型（GMM）和采用逆投影车辆样本的AdaBoost分类器准确检测夜间车辆的方法。首先，在交通场景中根据车头灯对的空间位置关系设置逆投影面，通过图像预处理粗定位车灯区域；其次，在逆投影图像下利用车头灯对的空间几何关系构建车灯对的高斯混合模型，初步匹配车头灯对；最后，采用逆投影车辆样本，利用AdaBoost分类器进一步准确检测车辆。实验在3个交通场景的检测结果表明，与原始图像下的AdaBoost方法相比，所提方法的检测率提高了1.93%，漏检率降低了17.83%，误检率降低了27.61%；与D-S （Dempster-Shafer）证据理论方法相比，检测率提高了2.03%，漏检率降低了7.58%，误检率降低了47.51%。所提方法提高了相对检测精度，减少了地面反光和影子等的干扰，满足交通场景中夜间车辆检测的可靠性和准确性的要求。     

Detection of vehicles from traffic scenes using fuzzy integrals

This paper presents an improved vehicle detection algorithm for traffic scene interpretation. In our previous high-level dynamic traffic scene interpretation systems (Dance et al., Picture Interpretation: A Symbolic Approach, World Scientific, New Jersey, 1995; Liu et al., IEEE Trans. Syst. Man Cybernet. Part B (2001) to appear), we used a simple background-subtraction procedure (referred to as Method 0) for vehicle detection, which, although adequate for well-defined image sequences, was not applicable to slightly displaced and noisy images. Our new method (referred to as Method 1) uses the Hough transform to extract the contour lines of the vehicles and morphological operations to reduce noise and improve the shape of the object regions. Finally, in the detection process we compute fuzzy integrals based on the evidence gathered. We have carried out extensive experiments. For all vehicles in the images, including the ones partially occluded and cut off at the image boundary, we were able to achieve a detection rate of 80% (Method 1) compared to 13% (Method 0). For the vehicles that are almost completely visible, the detection rate was 90%. In addition, for a set of 492 images, our new method reduces the number of false alarms from 427 (by Method 0) to only 9.