会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 81. 发明申请
    • METHOD FOR RE-IDENTIFICATION OF OBJECTS
    • 重新识别对象的方法
    • WO2016159199A1
    • 2016-10-06
    • PCT/JP2016/060590
    • 2016-03-24
    • MITSUBISHI ELECTRIC CORPORATION
    • JONES, MichaelMARKS, TimAHMED, Ejaz
    • G06K9/62G06K9/00
    • G06N3/0454G06K9/00771G06K9/6273
    • A method re-identifies objects in a pair of images by applying a convolutional neural network (CNN). Each layer in the network operates on an output of a previous layer. The layers include a first convolutional layer and a first max pooling layer to determine a feature map, a cross-input neighborhood differences layer to produce neighborhood difference maps, a patch summary layer to produce patch summary feature maps, a first fully connected layer to produce a feature vector representing higher order relationships in the patch summary feature maps, a second fully connected layer to produce two scores representing positive pair and negative pair classes, and a softmax layer to produce positive pair and negative pair probabilities. Then, the positive pair probability is output to signal whether the two images represent the same object or not.
    • 一种方法通过应用卷积神经网络(CNN)来重新识别一对图像中的对象。 网络中的每个层对先前层的输出进行操作。 这些层包括第一卷积层和第一最大池层以确定特征图,交叉输入邻域差层以产生邻域差分图,补丁汇总层以产生补丁汇总特征图,第一完全连接的层以产生 表示补丁汇总特征图中的较高阶关系的特征向量,第二完全连接层,以产生表示正对和负对类别的两个分数,以及产生正对和负对概率的softmax层。 然后,输出正对概率以通知两个图像是否表示相同的对象。
    • 82. 发明申请
    • SYSTEM AND METHOD FOR DETERMINING DEPTH IMAGE REPRESENTING DISTANCES TO POINTS OF SCENE
    • 用于确定深度图像代表场景点的距离的系统和方法
    • WO2016136410A1
    • 2016-09-01
    • PCT/JP2016/053294
    • 2016-01-28
    • MITSUBISHI ELECTRIC CORPORATION
    • BOUFOUNOS, Petros T.KADAMBI, Achuta
    • G01S17/89G01S17/10G01S7/484
    • G01S17/89G01S7/484G01S7/4861G01S17/107G01S17/32
    • A system for determining a depth image representing distances to points of a scene (106) includes a light source (110) for generating and directing a set of light pulses toward the scene and a set of sensors (140) for detecting and integrating light reflected from the scene. Each sensor is associated with a point in the scene and generates a signal indicative of an amount of light reflected from the point in the scene and reaching the sensor during a period of time. The system includes at least one processor (150) for coordinating operations of the light source and the set of sensors and for determining the depth image having intensity values representing the distances to the points in the scene. The processor commands the light source to generate the set of light pulses including a first pulse with a first intensity profile having portions (211, 212) of identical lengths with different integrated intensities (221, 222, 223).
    • 一种用于确定表示与场景(106)的距离的深度图像的系统包括用于产生并指向一组光脉冲朝向场景的光源(110),以及用于检测和积分反射的光的一组传感器(140) 从现场。 每个传感器与场景中的点相关联,并且产生一个指示从场景中的点反射的光量并在一段时间内到达传感器的信号。 该系统包括用于协调光源和传感器组的操作的至少一个处理器(150),并且用于确定具有表示到场景中的点的距离的强度值的深度图像。 处理器命令光源产生一组光脉冲,包括具有第一强度分布的第一脉冲,该第一脉冲具有具有不同积分强度(221,222,223)的具有相同长度的部分(211,212)。
    • 88. 发明申请
    • METHOD AND CONFIGURING DEVICE FOR CONFIGURING A RECEIVER
    • 用于配置接收器的方法和配置设备
    • WO2016006334A1
    • 2016-01-14
    • PCT/JP2015/064843
    • 2015-05-19
    • MITSUBISHI ELECTRIC CORPORATIONMITSUBISHI ELECTRIC R&D CENTRE EUROPE B.V.
    • GRESSET, Nicolas
    • H04L25/03H04L25/02
    • H04L25/03057H04L25/0202
    • For configuring a receiver receiving symbol vectors via a linear fading transmission channel, a configuring device : obtains candidate quantized channel matrices (S401); determines, for each candidate quantized channel matrix, pre-filtering parameters by optimizing a utility function according to the pre-filtering parameters in view of said candidate quantized channel matrix and from observations of the linear fading transmission channel, the utility function being representative of the receiver performance: selects one appropriate quantized channel matrix (S402) and the respective pre-filtering parameters according to the results of the utility function determined for each candidate quantized channel matrix and the respectively determined pre-filtering parameters; and configures the receiver (S407) such that the receiver applies, onto the received symbol vectors, a pre-filtering based on the selected pre-filtering parameters determined for the selected appropriate quantized channel matrix.
    • 为了配置接收器经由线性衰落传输信道接收符号向量,配置设备:获得候选量化信道矩阵(S401); 根据所述候选量化信道矩阵和从线性衰落传输信道的观测结果,根据预滤波参数优化效用函数,为每个候选量化信道矩阵确定预滤波参数,该效用函数代表 接收机性能:根据为每个候选量化信道矩阵和分别确定的预滤波参数确定的效用函数的结果,选择一个适当的量化信道矩阵(S402)和各个预滤波参数; 并且配置接收机(S407),使得接收机基于为所选择的适当的量化信道矩阵确定的所选择的预滤波参数,将所接收的符号向量应用于预滤波。
    • 89. 发明申请
    • METHOD FOR GENERATING CODE FOR COHERENT OPTICAL COMMUNICATIONS
    • 用于生成相干光通信代码的方法
    • WO2016002549A1
    • 2016-01-07
    • PCT/JP2015/067812
    • 2015-06-16
    • MITSUBISHI ELECTRIC CORPORATION
    • KOJIMA, KeisukeMILLAR, DavidAKINO, ToshiakiPARSONS, Kieran
    • H04B10/516H04L27/34
    • H04B10/508H04B10/5161H04L27/34
    • A method generates constant modulus multi-dimensional modulations for coherent optical communications by first projecting points in a constellation of the code onto a Poincare sphere or its higher-dimensional hyper-sphere. By using meta-heuristic procedures, nonlinear programming and gradient search methods, constellation points in the hyper-sphere are optimized in certain criteria, such as maximizing the minimum Euclidean distance, minimizing the union bound, minimizing the bit-error rate, minimizing the required signal-to-noise ratio, maximizing the nonlinear fiber reach, maximizing the phase noise tolerance, and maximizing the mutual information. Some methods use parametric unitary space-time block codes such as Grassmannian packing, and filter impulse response as well as unitary rotation over adjacent code blocks to generate near-constant modulus waveform, not only at the symbol timing, but also over the entire time.
    • 一种方法通过将代码星座中的点首先投影到庞加莱球体或其较高维度超球面上来产生用于相干光通信的恒模多维调制。 通过使用元启发式程序,非线性规划和梯度搜索方法,超特征中的星座点在某些标准中得到优化,例如最大化最小欧几里得距离,最小化联合边界,最小化误码率,最小化所需的 信噪比,使非线性光纤达到最大化,最大化相位噪声容限,并最大化互信息。 一些方法使用诸如Grassmannian打包和滤波器脉冲响应之类的参数单一时空块码以及相邻码块的单位旋转,以便不仅在符号定时,而且在整个时间内产生近恒模数波形。