【摘 要】
：
Edge is a significant feature for many computer vision applications.Existing edge detectors obtain a dense edge map depended on the contrast and gradient cues in an image.However, the dense edge map n
【机 构】
：
Department of Automation, University of Science and Technology of China,Anhui, Hefei, 230026
【出 处】
：
第20届中国系统仿真技术及其应用学术年会(20th CCSSTA 2019)
论文部分内容阅读
Edge is a significant feature for many computer vision applications.Existing edge detectors obtain a dense edge map depended on the contrast and gradient cues in an image.However, the dense edge map not only restricts the efficiency of subsequent vision tasks, but also may contain several edges located in background.In this paper, we propose a sparse edge detector by exploiting the boundary adherence of superpxiel.We first obtain original edge response and superpixel segmentation from Structured Edge Detector (SED) and Simple Linear Iterative Clustering (SLIC) separately.Then a simplifying operation is exerted on the original edge map according to the border guidance of superpixel.Extensive experiments on BSDS500 dataset demonstrate that the proposed method achieves a sparse and discriminative edge detector.
其他文献
动力锂电池在新能源电动汽车的应用中面临诸如循环寿命不足、动力不足等挑战,采用超级电容作为补充储能器件可以有效缓解上述问题.本文从仿真建模和优化分析出发,比较锂电池/超级电容混合储能系统的几种常见拓扑结构的优劣.为了使不同拓扑具有可比性,本文首先确定了满足车辆需求的基本配比数量.然后基于动态规划方法,针对NEDC工况,本文给出了不同拓扑下最优功率分配模式的性能表现.对比结果显示,电容/电池半主动式拓
锂离子电池作为一种清洁环保的储能器件,被广泛的用于新能源汽车和微电网等能量存储领域.但是由于其应用环境的不确定加之电池本身是个复杂热电耦合系统,使得对电池老化机理的研究变得更加困难.本文介绍了一种基于电化学模型的锂离子电池老化机理研究方法.主要贡献可以概括为:①基于单粒子模型分析了电池内部反应机理,该模型的参数由原始物理性质参数重组而成,每个参数都具有特定的物理意义.②采用脉冲响应和参数辨识方法获
Graph-Cut RANSAC is an innovative method for outlier rejection in Visual Odometry field.Research aboutoutlier rejectionhas been gradually developedin visual odometer field.These technologies have been
Salient object detection is a fundamental computer vision task, which aims to identify and locate the most interesting objects attracting visual attention of humans in an image.In this work, we regard
For some rodent mammals when they foraging or looking for a target, the positions and headings in their brain cells are distributed as a hexagonal raster shape.Hex-grid maps used in the navigation of
Nowadays due to the advancement of technology, especially in the field of computer science,quality of the maps, resolution as well as the image details, has been remarkable.Although having a more deta
随着深度学习的快速发展,基于深度神经网络的目标检测模型被不断提出并且检测的效率和精度均得到大幅提升.然而,目前的环境目标检测方法在复杂场景下的性能仍然受限.多传感器信息融合被认为是进一步实现目标检测技术突破的重要途径.本文针对环境感知中的车辆检测技术,着重介绍了融合视觉信息和激光信息的多视图目标检测(AVOD)模型的框架,对其工作原理进行说明,探究了决定网络模型性能的关键因素.最后,论文在KITT
Simultaneous localization and mapping (SLAM)has made astonishing progress over the last 30 years and ORB-SLAM system is definitely one of the most essential and popular frameworks so far.The ORB-SLAM
本文描述飞机飞行控制系统电缆设计使用的工具、电线材料选取、线束长度确定,以及基本的画法,供飞机飞行控制系统及其他系统电缆设计时参考.
本文针对智能集群区域探测问题,研究了一种基于质心Voronoi图法的多智能体区域覆盖控制策略.针对质心Voronoi图法的收敛性与初始构型耦合这一问题,采用先分散、后聚拢两步法运动,使得多智能体分布效果明显提升.定义了区域覆盖问题的目标代价函数,以代价函数的函数最优为目标,进行收敛性仿真研究.仿真结果证明了控制策略的可行性和有效性.