Multiclass object classification for real-time video surveillance systems

Object classification in video is an important factor for improving the reliability of various automatic applications in video surveillance systems, as well as a fundamental feature for advanced applications, such as scene understanding. Despite extensive research, existing methods exhibit relatively moderate classification accuracy when tested on a large variety of real-world scenarios, or do not obey the real-time constraints of video surveillance systems. Moreover, their performance is further degraded in multi-class classification problems. We explore multi-class object classification for real-time video surveillance systems and propose an approach for classifying objects in both low and high resolution images (human height varies from a few to tens of pixels) in varied real-world scenarios. Firstly, we present several features that jointly leverage the distinction between various classes. Secondly, we provide a feature-selection procedure based on entropy gain, which screens out superfluous features. Experiments, using various classification techniques, were performed on a large and varied database consisting of ∼29,000 object instances extracted from 140 different real-world indoor and outdoor, near-field and far-field scenes having various camera viewpoints, which capture a large variety of object appearances under real-world environmental conditions. The insight raised from the experiments is threefold: the efficiency of our feature set in discriminating between classes, the performance improvement when using the feature selection method, and the high classification accuracy obtained on our real-time system on both DSP (TMS320C6415-6E3, 600 MHz) and PC (Quad Core Intel® Xeon® E5310, 2 × 4 MB Cache, 1.60 GHz, 1066 MHz) platforms.     

视频监控系统中阴影去除方法的研究

提出了一种基于颜色不变性和建立阴影高斯模型的阴影检测和消除的方法。首先对亮度小于背景的前景根据颜色特征的相近性进行划分,再利用阴影的光谱特性建立高斯模型,去除运动目标的投影。最后利用阴影的空间特性对图像进行后处理,完成阴影的检测与消除。实验结果表明,该方法能够有效地检测和消除阴影。     

Multipath routing and rate-controlled video encoding in wireless video surveillance networks

Video surveillance systems typically consist of many video sources distributed over a wide area, transmitting live video streams to a central location for processing and monitoring. The target of this paper—to bring down the overall system cost and increase feasibility, scalability, and performance—is to propose a new architecture for a wireless video surveillance network, whose telecommunication infrastructure is based on a wireless mesh network, and where video sources are able to estimate network bandwidth and consequently control their output rate. Multipath routing is applied in such a way that at least part of the information arrives at its destination even if a wireless link is shielded (maliciously or not). A case study is considered to discuss the performance of the proposed architecture, analyzing a comparison between single-path and multipath approaches.     

梯度直方图卷积特征的胶囊网络在交通监控下的车型分类

为了充分利用图像信息以提高现有交通监控下车型分类的效果,在胶囊网络的基础上增加梯度直方图卷积（HOG-C）特征提取方法,提出HOG-C特征的胶囊网络模型——HOG-C CapsNet。首先,使用梯度统计特征提取层对图像中的梯度信息进行统计,构建方向梯度直方图（HOG）特征图;其次,使用卷积层提取出图像的颜色信息,把提取出的颜色信息与HOG特征图融合构成HOG-C特征图;最后,输入卷积层提取HOG-C特征图的抽象特征,并通过胶囊网络对提取的抽象特征进行具有三维空间特征表达的胶囊封装,使用动态路由算法实现车型分类。在BIT-Vehicle数据集上对该模型和其他相关模型进行的对比实验中,该模型得到98.17%的准确率、97.98%的平均精确率均值（MAP）、98.42%的平均召回率均值（MAR）和98.20%的综合评价指标。实验结果表明,该模型在交通监控下的车型分类上具有更好的效果。     

多车道路面的大车车型分类特征及提取方法

针对车流量大的多车道路面,提出了以车体结构和颜色为基础的6个能有效区分大客车和大货车的特征,并介绍了提取过程。该过程首先使用Sobel算子与颜色融合的抗干扰水平边缘检测方法提取区间分界线,然后以基于RGB和HSI相结合的颜色分类方法识别区间颜色,进而获得特征的表示。现场实测结果表明,这些特征提取准确率高且速度快,能满足实时性要求。基于这些特征的车型分类系统已应用于实际现场并取得了良好的识别效果。     

改进的应用于交通场景的运动目标检测方法及质量评价*

针对ViBe算法在交通视频检测中出现明显鬼影区域、缓慢目标残影难以消除、检测精确度和鲁棒性不足的问题,本文提出改进算法,利用灰度信息为像素建立生命长度矩阵,使鬼影或残影快速融入背景样本得以消除。结合最大类间方差法设置自适应阈值,加入良好后处理抑制动态噪音。同时本文借鉴分类算法的统计指标,提出质量评价多个要素,以ViBe原算法、混合高斯算法(GMM)、LBP-OTSU相结合的背景差分法和本文改进算法为例,定性、定量对实验结果作出质量评价和分析。实验表明,改进算法在较少帧数内去除了鬼影,抑制了运动目标残影,提高了运动目标检测的准确度和整体性能。     

Designing an intelligent ontological system for traffic light control in isolated intersections

This paper models the traffic light control domain using a fuzzy ontology and applies it to control isolated intersections. Proposing an independent module for reusing traffic light control knowledge is one of the most important purposes of this paper. In this way, software independency increases and other software development activities, such as test and maintenance, are facilitated. The ontology has been developed manually and evaluated by experts. Moreover, the traffic data is extracted and classified from images of intersections using image processing algorithms and artificial neural networks. According to predefined XML schema, this information is transformed to XML instances and mapped onto the fuzzy ontology for firing suitable fuzzy rules using a fuzzy inference engine. The performance of the proposed system is compared with other similar approaches. The comparison shows that it has a much lower average delayed time for each car in each cycle in all traffic conditions as compared with the other ones.     

交通监控中运动目标轨迹的距离计算和聚类

对交通监控中运动目标的轨迹距离计算和聚类方法进行了改进.在轨迹距离计算中,引入目标的空间坐标、运动速度、运动方向和尺寸4个参数,以提高聚类时对不同位置、不同速度、不同方向和不同尺寸运动目标的轨迹的区分能力;针对交通目标运动轨迹比较规律的特点,采用基于统计的方法对K均值的轨迹聚类算法进行初始化,从而可以自适应的确定聚类数目K值和聚类初始中心.在真实场景下,验证了算法的有效性和适用性.