公开(公告)号：US20170098123A1

公开(公告)日：2017-04-06

申请号：US15379524

申请日：2016-12-15

Applicant: DENSO CORPORATION

Inventor： Yukimasa TAMATSU ,  Kensuke YOKOI ,  Ikuro SATO

IPC: G06K9/00 ,  G06K9/66 ,  G06K9/46

CPC classification number: G06K9/00369 ,  G06K9/00805 ,  G06K9/4628 ,  G06K9/6267 ,  G06K9/66 ,  H04N5/144

Abstract: A detection device has a neural network process section performing a neural network process using parameters to calculate and output a classification result and a regression result of each of frames in an input image. The classification result shows a presence of a person in the input image. The regression result shows a position of the person in the input image. The parameters are determined based on a learning process using a plurality of positive samples and negative samples. The positive samples have segments of a sample image containing at least a part of the person and a true value of the position of the person in the sample image. The negative samples have segments of the sample image containing no person.

公开(公告)号：US20150134578A1

公开(公告)日：2015-05-14

申请号：US14540295

申请日：2014-11-13

Applicant: DENSO CORPORATION

Inventor： Yukimasa TAMATSU ,  Ikuro SATO

IPC: G06N99/00 ,  G06N5/04

CPC classification number: G06N3/084 ,  G06N3/0454

Abstract: A discriminator based on supervised learning includes a data expanding unit and a discriminating unit. The data expanding unit performs data expansion on unknown data which is an object to be discriminated in such a manner that a plurality of pieces of pseudo known data are generated. The discriminating unit applies the plurality of pieces of unknown pseudo data that has been expanded by the data expansion unit to a discriminative model so as to discriminate the plurality of pieces of pseudo unknown data, and integrates discriminative results of the plurality of pieces of pseudo unknown data to perform class classification such that the unknown data is classified into classes.

Abstract translation: 基于监督学习的鉴别器包括数据扩展单元和鉴别单元。 数据扩展单元以作为生成多条伪已知数据的方式对作为被鉴别对象的未知数据进行数据扩展。 识别单元将已经被数据扩展单元扩展的多个未知伪数据片段应用于判别模型，以区分多条伪未知数据，并且将多个伪未知数的判别结果进行积分 用于执行类分类的数据，使得未知数据被分类为类。

公开(公告)号：US20150134583A1

公开(公告)日：2015-05-14

申请号：US14540277

申请日：2014-11-13

Applicant: DENSO CORPORATION

Inventor： Yukimasa TAMATSU ,  Ikuro SATO

IPC: G06N3/08 ,  G06N3/04

CPC classification number: G06N3/08 ,  G06N3/04

Abstract: A learning apparatus performs a learning process for a feed-forward multilayer neural network with supervised learning. The network includes an input layer, an output layer, and at least one hidden layer having at least one probing neuron that does not transfer an output to an uppermost layer side of the network. The learning apparatus includes a learning unit and a layer quantity adjusting unit. The learning unit performs a learning process by calculation of a cost derived by a cost function defined in the multilayer neural network using a training data set for supervised learning. The layer quantity adjusting unit removes at least one uppermost layer from the network based on the cost derived by the output from the probing neuron, and sets, as the output layer, the probing neuron in the uppermost layer of the remaining layers.

Abstract translation: 学习装置对具有监督学习的前馈多层神经网络进行学习处理。 网络包括输入层，输出层和至少一个具有至少一个探测神经元的隐藏层，该探测神经元不将输出传输到网络的最上层。 学习装置包括学习单元和层数量调整单元。 学习单元通过使用针对监督学习的训练数据集计算由多层神经网络中定义的成本函数导出的成本来执行学习过程。 基于由探测神经元的输出导出的成本，层数量调整单元从网络中去除至少一个最上层，并且将剩余层的最上层中的探测神经元设置为输出层。

公开(公告)号：US20150006444A1

公开(公告)日：2015-01-01

申请号：US14317261

申请日：2014-06-27

Applicant: DENSO CORPORATION

Inventor： Yukimasa TAMATSU ,  Ikuro SATO

IPC: G06N3/08 ,  G06N99/00

CPC classification number: G06N3/082

Abstract: When it is determined that a minimum value of a cost function of a candidate structure obtained by a training process of a specified-number sequence is equal to or higher than that of the cost function of the candidate structure obtained by the first step of a previous sequence immediately before the specified-number sequence, a method performs, as a random removal step of the specified sequence, a step of randomly removing at least one unit from the candidate structure obtained by the first step of the previous sequence again. This gives a new generated structure of the target neural network based on the random removal to the first step as the input structure of the target neural network. The method performs the specified-number sequence again using the new generated structure of the target neural network.

Abstract translation: 当确定通过指定数量序列的训练处理获得的候选结构的成本函数的最小值等于或高于由先前的第一步骤的第一步骤获得的候选结构的成本函数的最小值 在指定序列之前的方法中，方法执行作为指定序列的随机去除步骤的步骤，再次从先前序列的第一步骤获得的候选结构中随机地去除至少一个单元。 这给出了基于随机去除第一步的目标神经网络的新生成结构作为目标神经网络的输入结构。 该方法使用目标神经网络的新生成结构再次执行指定序列序列。

公开(公告)号：US20150015596A1

公开(公告)日：2015-01-15

申请号：US14326805

申请日：2014-07-09

Applicant: Denso Corporation

Inventor： Yasutsugu NAGATOMI ,  Tadashi KAMADA ,  Akira TAKAHASHI ,  Yukimasa TAMATSU ,  Ryusuke HOTTA ,  Shohei MORIKAWA

IPC: G06T11/00 ,  G06K9/00

CPC classification number: G06T11/001 ,  B60R1/00 ,  B60R2021/01013 ,  B60R2300/307 ,  B60R2300/802 ,  G06K9/00805 ,  G06T2207/20192 ,  G08G1/166

Abstract: When a driving support apparatus is under automatic driving of a vehicle or an automatic driving button is pressed under manual driving, vicinity image data is acquired from an in-vehicle camera. When a predetermined target object is recognized in the vicinity image data, a visibility reduction process is applied to image data of the recognized target object. The visibility reduction process applies at least one of defocusing; decreasing color information; and decreasing edge intensity, to the image data of the recognized target object. In contrast, any visibility reduction process is not applied to any other image data other than the image data of the recognized target object. An image display apparatus displays the vicinity image by a combination of the image data of the recognized target object of which the visibility is reduced and the other image data of which the visibility is not reduced.

Abstract translation: 在驾驶辅助装置处于自动驾驶状态或在手动驾驶下按下自动驾驶按钮时，从车载摄像机取得附近图像数据。 当在邻近图像数据中识别到预定目标对象时，可视度降低处理被应用于识别的目标对象的图像数据。 可见度降低过程至少适用于散焦; 减少颜色信息; 并将边缘强度降低到识别的目标对象的图像数据。 相比之下，任何可见度降低处理不应用于除了识别的目标对象的图像数据之外的任何其他图像数据。 图像显示装置通过将可见度降低的识别目标对象的图像数据与可见度不降低的其他图像数据的组合来显示附近图像。

公开(公告)号：US20140037138A1

公开(公告)日：2014-02-06

申请号：US13689196

申请日：2012-11-29

Applicant: DENSO IT LABORATORY, INC. ,  DENSO CORPORATION

Inventor： Ikuro SATO ,  Yukimasa TAMATSU ,  Kunihiro GOTO

IPC: G08G1/16 ,  G06K9/00

CPC classification number: G08G1/166 ,  G06K9/00805

Abstract: The moving object recognition system includes: a camera that is installed in a vehicle and captures continuous single-view images; a moving object detecting unit that detects a moving object from the images captured by the camera; a relative approach angle estimating unit that estimates the relative approach angle of the moving object detected by the moving object detecting unit with respect to the camera; a collision risk calculating unit that calculates the risk of the moving object colliding with the vehicle, based on the relationship between the relative approach angle and the moving object direction from the camera toward the moving object; and a reporting unit that reports a danger to the driver of the vehicle in accordance with the risk calculated by the collision risk calculating unit.

Abstract translation: 移动物体识别系统包括：安装在车辆中并捕获连续单视图图像的照相机; 移动物体检测单元，其从由所述照相机拍摄的图像检测移动物体; 相对进给角估计单元，其估计由所述移动物体检测单元相对于所述照相机检测到的所述移动物体的相对接近角度; 碰撞风险计算单元，其基于从相机朝向移动物体的相对进近角度和移动物体方向之间的关系来计算与车辆碰撞的运动物体的风险; 以及报告单元，其根据由碰撞风险计算单元计算的风险向车辆的驾驶员报告危险。