公开(公告)号：US09542606B2

公开(公告)日：2017-01-10

申请号：US14635540

申请日：2015-03-02

Applicant: DENSO CORPORATION

Inventor： Hirotake Ishigami ,  Naoki Kawasaki ,  Syunya Kumano ,  Shunsuke Suzuki ,  Satoshi Osanai ,  Tetsuya Takafuji

IPC: G06K9/00 ,  G06T7/00 ,  G06T7/60

CPC classification number: G06K9/00798 ,  G06T7/13 ,  G06T7/60 ,  G06T2207/20061 ,  G06T2207/30256

Abstract: In a lane line recognition apparatus, an abnormality determiner is configured to determine whether or not a width of a travel lane defined by left-side and right-side lane lines is abnormal, and a lane-line recognizer is configured to, when the width of the travel lane is not abnormal, recognize both of the left-side and right-side lane lines in a both-side line recognition mode, and when the width of the travel lane is abnormal, recognize one of the lane lines in a one-side line recognition mode. In the one-side line recognition mode, the lane-line recognizer is configured to, for each of the left-side and right-side lane lines, calculate two or more parameters of the lane line, and then integrate recognition results for the respective two or more parameters, and based on the integrated recognition results of the left-side and right-side lane lines, select one of the lane lines to be recognized.

Abstract translation: 在车道线识别装置中，异常判定器被配置为确定由左侧车道线和右侧车道线限定的行驶车道的宽度是否异常，并且车道线识别器被配置为当宽度 行驶车道不正常的情况下，在双方侧线路识别模式下识别左侧车道线路和右侧车道线路两侧，并且当行驶车道的宽度异常时，识别一条车道线路 边线识别模式。 在单面线识别模式中，车道线识别器被配置为对于左侧车道线和右侧车道线中的每一条，计算车道线的两个或更多个参数，然后将各个车道线的识别结果集成 两个或多个参数，并且基于左侧和右侧车道线的集成识别结果，选择要被识别的车道线之一。

公开(公告)号：US09460352B2

公开(公告)日：2016-10-04

申请号：US14792858

申请日：2015-07-07

Applicant: DENSO CORPORATION

Inventor： Tomohiko Tsuruta ,  Shunsuke Suzuki ,  Naoki Kawasaki ,  Osamu Shimomura

IPC: G06K9/00 ,  G06K9/52 ,  G06T7/60 ,  G06K9/62 ,  G06T7/00 ,  G06K9/46

CPC classification number: G06K9/00798 ,  G06K9/52 ,  G06K9/6215 ,  G06K2009/4666 ,  G06T7/60 ,  G06T2207/30256

Abstract: In a lane boundary line recognition device, an extraction unit extracts lane boundary line candidates from image acquired by an in-vehicle camera. A position estimation unit estimates a position of each lane boundary line based on drive lane information containing a number of drive lanes on a roadway and a width of each drive lane when (a) and (b) are satisfied, (a) when an own vehicle drives on an own vehicle lane specified by the drive lane specifying unit, and (b) when the lane boundary line candidate corresponds to lane boundary lines of the own vehicle lane. A likelihood calculation unit increases a likelihood of the lane boundary line candidate when a distance between a position of the lane boundary line candidate and an estimated position of the lane boundary line candidate obtained by the drive lane boundary line position estimation unit is within a predetermined range.

Abstract translation: 在车道边界线识别装置中，提取单元从车载摄像机获取的图像中提取车道边界线候选。 当满足（a）和（b）时，位置估计单元基于包含道路上的多个驱动车道的驱动车道信息和每个驱动车道的宽度来估计每个车道边界线的位置，（a）当自身 由车道指定单元指定的车辆车道上的车辆驱动，以及（b）当车道边界线候补对应于本车道的车道边界线时。 当车道边界线候补的位置与由行驶车道边界线位置估计单元获得的车道边界线候选者的估计位置之间的距离在预定范围内时，似然度计算单元增加车道边界线候选的可能性 。

公开(公告)号：US20160196474A1

公开(公告)日：2016-07-07

申请号：US14965041

申请日：2015-12-10

Applicant: DENSO CORPORATION

Inventor： Yuu Maeda ,  Shunsuke Suzuki ,  Naoki Kawasaki

IPC: G06K9/00 ,  G06T5/00

CPC classification number: G06K9/00798 ,  G06T5/002 ,  G06T2207/10024 ,  G06T2207/30256

Abstract: In an image processing apparatus, an image data acquisition unit acquires image data including R-pixels which are red pixels, holding red luminance values, B-pixels which are blue pixels, holding blue luminance values, G-pixels which are green pixels, holding green luminance values, and C-pixels which are clear pixels, holding luminance values in a wavelength range including wavelength ranges corresponding to red, blue and green. In the apparatus, a determination unit determines whether or not an inequality given by Valc−Valx

Abstract translation: 在图像处理装置中，图像数据获取单元获取包含红色像素，保持红色亮度值的R像素，保持蓝色亮度值的蓝色像素的B像素，保持蓝色亮度值的G像素，保持绿色像素的G像素的图像数据 绿色亮度值和作为清晰像素的C像素，保持包括对应于红色，蓝色和绿色的波长范围的波长范围内的亮度值。 在该装置中，确定单元确定是否满足由Valc-Valx <α给出的不等式，其中Valx是作为R，R和G像素之一的感兴趣像素的亮度值 图像数据Valc是最靠近感兴趣像素的C像素或像素的亮度值，α是常数，并且如果不等式被满足，则校正单元将亮度值Valx校正为小于亮度 价值Valc。

公开(公告)号：US20160173831A1

公开(公告)日：2016-06-16

申请号：US14964131

申请日：2015-12-09

Applicant: DENSO CORPORATION

Inventor： Yusuke Akamine ,  Naoki Kawasaki ,  Shunsuke Suzuki

IPC: H04N7/18 ,  G06T7/00 ,  G06K9/00

CPC classification number: G06K9/00798 ,  G06K9/00805

Abstract: A lane boundary line recognition apparatus, mounted to an own vehicle, images a road surface ahead of the own vehicle and acquires an image. Edge points are extracted from the image. A lane boundary line candidate is extracted based the edge points. A lane boundary line probability of the lane boundary line candidate is calculated. A lane boundary line candidate of which the lane boundary line probability exceeds a predetermined threshold is recognized as a lane boundary line. A solid object is recognized in an image. A lane boundary line being hidden by the solid object is detected. When a lane boundary line being hidden is detected, the lane boundary line probability is suppressed in at least a part of an area outside of the hidden lane boundary line, compared to that when the lane boundary line being hidden is not detected.

Abstract translation: 安装在自身车辆上的车道边界线识别装置对自身车辆前方的路面进行成像并获取图像。 从图像中提取边缘点。 基于边缘点提取车道边界线候选。 计算车道边界线候选的车道边界线概率。 车道边界线概率超过预定阈值的车道边界线候选被识别为车道边界线。 在图像中识别实体。 检测由固体物体隐藏的车道边界线。 当检测到隐藏的车道边界线时，与隐藏车道边界线以外的区域的至少一部分相比，未检测到车道边界线被隐藏时，车道边界线概率被抑制。

公开(公告)号：US20160055383A1

公开(公告)日：2016-02-25

申请号：US14830200

申请日：2015-08-19

Applicant: DENSO CORPORATION

Inventor： Yusuke Akamine ,  Naoki Kawasaki ,  Tomohiko Tsuruta ,  Shunsuke Suzuki

IPC: G06K9/00 ,  G06K9/66

CPC classification number: G06K9/00798 ,  G06K9/033

Abstract: In a lane boundary line recognition device, a calculation section calculates a degree of uncertainty which affects a correct recognition of white lines on a roadway of a vehicle. A learning section updates a learning value of the degree of uncertainty. A recognition suppression section suppresses execution of a recognition process of recognizing white lines on the roadway when the updated learning value is more than a threshold value. An environment change judgment section judges whether or not a road environment has changed. A learning resetting section resets the learning value of the degree of uncertainty to a previous learning value when the detection result of the environment change judgment section indicates an occurrence of change of the road environment.

Abstract translation: 在车道边界线识别装置中，计算部分计算影响车辆道路上的白线的正确识别的不确定度。 学习部分更新不确定程度的学习价值。 当更新的学习值大于阈值时，识别抑制部分抑制识别道路上的白线的识别处理的执行。 环境变化判断部判断道路环境是否变化。 当环境变化判断部分的检测结果指示道路环境的变化的发生时，学习重置部分将不确定度的学习值重置为先前的学习值。

公开(公告)号：US20160042236A1

公开(公告)日：2016-02-11

申请号：US14783060

申请日：2014-04-04

Applicant: DENSO CORPORATION

Inventor： Kazuhisa Ishimaru ,  Shunsuke Suzuki ,  Tetsuya Takafuji

IPC: G06K9/00 ,  H04N7/18 ,  G06T7/00

CPC classification number: G06K9/00798 ,  G06T7/12 ,  G06T2207/20076 ,  G06T2207/30256 ,  G08G1/167 ,  H04N7/183

Abstract: This invention is provided with: a camera for capturing the image of a travel path; an edge point extraction unit for extracting edge points on the basis of the brightness of an image captured by the camera; a candidate line extraction unit for extracting, on the basis of the succession of the extracted edge points, a candidate line for a boundary line demarcating the travel path; a frequency calculation unit for calculating, on the basis of edge points belonging to the candidate line extracted by the candidate line extraction unit, the frequency distribution of the edge points for a parameter that specifies the width of the boundary line; a probability generation unit for calculating, on the basis of the frequency distribution calculated by the frequency calculation unit, the distribution for the probability that the candidate line at the parameter is the boundary line; and a boundary line recognition unit for recognizing the boundary line on the basis of the probability distribution calculated by the probability generation unit.

Abstract translation: 本发明提供有：摄像机，用于拍摄行进路径的图像; 边缘点提取单元，用于基于由照相机拍摄的图像的亮度提取边缘点; 候选行提取单元，用于根据所提取的边缘点的连续性，提取划分行驶路径的边界线的候选行; 频率计算单元，用于根据候选行提取单元提取的属于候选行的边缘点，计算用于指定边界线宽度的参数的边缘点的频率分布; 概率生成单元，其基于由所述频率计算单元计算出的频率分布，计算所述参数的候补行为边界线的概率的分布; 以及边界线识别单元，用于基于由概率生成单元计算的概率分布来识别边界线。