公开(公告)号：EP1655688A3

公开(公告)日：2008-12-10

申请号：EP05077317.5

申请日：2005-10-11

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Kiselewich, Stephen J.

IPC: G06K9/46 ,  G06K9/00 ,  G06K9/52

CPC classification number: G06K9/00369 ,  B60R21/01538 ,  G06K9/4614 ,  G06K9/527

Abstract: A stream of images including an area occupied by at least one object are processed to extract wavelet coefficients, and the extracted coefficients are represented as wavelet signatures that are less susceptible to misclassification due to noise and extraneous object features. Representing the wavelet coefficients as wavelet signatures involves sorting the coefficients by magnitude, setting a coefficient threshold based on the distribution of coefficient magnitudes, truncating coefficients whose magnitude is less than the threshold, and quantizing the remaining coefficients.

公开(公告)号：EP1883050A2

公开(公告)日：2008-01-30

申请号：EP07075564.0

申请日：2007-07-05

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Kiselewich, Stephen J. ,  Shen, Junqiang

IPC: G06T7/00

CPC classification number: G06T7/0008 ,  G06K9/3241 ,  G06K2209/23

Abstract: The empty vs. non-empty status of a cargo container (10) is detected based on boundary analysis of a wide-angle image obtained by a monocular vision system (14). The wide-angle image is warped (56) to remove distortion created by the vision system optics (18a), and the resulting image is edge-processed (58) to identify the boundaries of the container floor (10e). If package boundaries are detected within the floor space (82), or a large foreground package is blocking the floor boundaries (86), the cargo status is set to non-empty (84). If floor boundaries (24a, 24b) are detected and no package boundaries are detected within the floor space (82, 86), the cargo status is set to empty (88).

Abstract translation: 基于由单目视觉系统（14）获得的广角图像的边界分析来检测货物集装箱（10）的空 - 非空状态。 广角图像被扭曲（56）以去除由视觉系统光学器件（18a）产生的变形，并且所得到的图像被边缘处理（58）以识别容器底板（10e）的边界。 如果在楼面空间（82）内检测到包装边界，或者大的前台包装阻挡了楼层边界（86），货物状态被设置为非空（84）。 如果检测到楼层边界（24a，24b），并且在地板空间（82,86）内没有检测到包装边界，则将货物状态设置为空（88）。

公开(公告)号：EP1703462A1

公开(公告)日：2006-09-20

申请号：EP06075529.5

申请日：2006-03-07

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Taylor, Ronald M. ,  Perisho, Robert A. Jr.

IPC: G06T5/40

CPC classification number: G06T5/009 ,  G06T5/40

Abstract: A system (20) and method for enhancing the contrast within an image (22). An enhanced image (24) can be generated in a real-time or substantially real-time manner from an initial image (22). The pixel values (38) of the initial image (22) can be used to populate a histogram (40) or otherwise serve as the basis for subsequent processing. A valley (44) can be identified within the range of pixel values (38) for use as a stretch metric (48) used by a stretch heuristic (46) to expand the contrast of the pixel values (38) in the initial image (22) by expanding the range of pixel values (38) associated with the pixels (36) in the histogram (40). In some embodiments, the initial image (22) is first divided into image regions (52) that are each associated with individualized processing. A bilinear interpolation step (56) can then be performed to smooth the integrated image after the individualized processing is used to stretch the pixels (36) within the individual image regions (52).

Abstract translation: 一种用于增强图像（22）内的对比度的系统（20）和方法。 可以从初始图像（22）以实时或基本上实时的方式生成增强图像（24）。 初始图像（22）的像素值（38）可以用于填充直方图（40）或以其他方式用作后续处理的基础。 可以在像素值（38）的范围内识别谷（44），以用作由拉伸启发式（46）使用的拉伸度量（48），以扩展初始图像中的像素值（38）的对比度 通过扩展与直方图（40）中的像素（36）相关联的像素值（38）的范围来实现。 在一些实施例中，初始图像（22）首先被划分成各自与个体化处理相关联的图像区域（52）。 然后可以执行双线性插值步骤（56）以在使用个体化处理来拉伸各个图像区域（52）内的像素（36）之后平滑集成图像。

公开(公告)号：EP1758058B1

公开(公告)日：2009-11-18

申请号：EP06076559.1

申请日：2006-08-10

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Bauson, William A. ,  Perisho Jr., Robert A.

IPC: G06T5/00

CPC classification number: G06T5/009 ,  G06T5/50 ,  G06T2207/20016

公开(公告)号：EP1679639B1

公开(公告)日：2008-07-23

申请号：EP05077935.4

申请日：2005-12-20

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Kieselewich, Stephen J.

IPC: G06K9/00

CPC classification number: G06K9/3241

公开(公告)号：EP1903535A1

公开(公告)日：2008-03-26

申请号：EP07075793.5

申请日：2007-09-12

Applicant: Delphi Technologies, Inc.

Inventor： Dhua, Arnab S. ,  Zhang, Yan ,  Kiselewich, Stephen J.

IPC: G08G1/16 ,  G06T7/00 ,  B60R21/013 ,  B60R21/00

CPC classification number: G06K9/3241 ,  B60W30/08 ,  G06K9/00805 ,  G06K2209/23 ,  G06T7/70 ,  G08G1/16 ,  G08G1/166

Abstract: A method (100) for identifying objects in an electronic image is provided. The method (100) includes the steps of providing an electronic source image (10) and processing the electronic source image to identify edge pixels. The method (100) further includes the steps of providing an electronic representation of the edge pixels (10') and processing the electronic representation of the edge pixels (10') to identify valid edge center pixels. The method (100) still further includes the step of proving an electronic representation of the valid edge center pixels. Each valid edge center pixel represents the approximate center of a horizontal edge segment of a target width. The horizontal edge segment is made up of essentially contiguous edge pixels. The method (100) also includes the steps of determining symmetry values of test regions (46,48,50,95) associated with valid edge center pixels, and classifying the test regions (46,48,50,95) based on factors including symmetry.

Abstract translation: 提供了一种用于识别电子图像中的对象的方法（100）。 方法（100）包括提供电子源图像（10）和处理电子源图像以识别边缘像素的步骤。 方法（100）还包括提供边缘像素（10'）的电子表示并处理边缘像素（10'）的电子表示以识别有效边缘中心像素的步骤。 方法（100）还包括证明有效边缘中心像素的电子表示的步骤。 每个有效的边缘中心像素表示目标宽度的水平边缘片段的大致中心。 水平边缘段由基本相邻的边缘像素组成。 方法（100）还包括以下步骤：确定与有效边缘中心像素相关联的测试区域（46,48,50,95）的对称值，以及基于以下因素对测试区域（46,48,50,95）进行分类： 对称。

公开(公告)号：EP1679639A1

公开(公告)日：2006-07-12

申请号：EP05077935.4

申请日：2005-12-20

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Kieselewich, Stephen J.

IPC: G06K9/00

CPC classification number: G06K9/3241

Abstract: An object classification method (100a) for a collision warning system is disclosed. The method includes the steps of capturing (10) a video frame (25) with an imaging device and examining a radar-cued potential object location (50) within the video frame (25), extracting (12) orthogonal moment features from the potential object location (50), extracting (14) Gabor filtered features from the potential object location (50), and classifying (16) the potential object location (50) into one of a first type of image (18a, 18b) or a second type of image (18c, 18d) in view of the extracted orthogonal moment features and the Gabor filtered features.

Abstract translation: 公开了一种用于碰撞预警系统的物体分类方法（100a）。 该方法包括以下步骤：利用成像装置捕获（10）视频帧（25）并检查视频帧（25）内的雷达提示的潜在对象位置（50），从电位提取（12）正交矩特征 对象位置（50），从潜在对象位置（50）提取（14）Gabor滤波特征，并将潜在对象位置（50）分类为第一类型的图像（18a，18b）或第二 鉴于提取的正交矩特征和Gabor滤波特征，图像类型（18c，18d）。

公开(公告)号：EP1703462B1

公开(公告)日：2012-01-11

申请号：EP06075529.5

申请日：2006-03-07

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Taylor, Ronald M. ,  Perisho, Robert A. Jr.

IPC: G06T5/40

CPC classification number: G06T5/009 ,  G06T5/40

公开(公告)号：EP1973058A2

公开(公告)日：2008-09-24

申请号：EP08152835.8

申请日：2008-03-17

Applicant: Delphi Technologies, Inc.

Inventor： Kiselewich, Stephen J. ,  Zhang, Yan ,  Park, Su-Birm

IPC: G06K9/00

CPC classification number: G06K9/00362

Abstract: A method of object classification (10) including the steps of providing an image device (28), providing selecting predetermined imaging features to be extracted from an image produced by the imaging device (28) based upon a desired classification of an object (14), and obtaining an image (16) by the imaging device (28) of at least one object in a field of view of the imaging device (28). The method (10) further includes the steps of extracting at least one feature from the image (18), wherein the at least one feature corresponds to the predetermined imaging features, determining a value for each of the extracted at least one feature (21), and classifying the object based upon the at least one feature that could be extracted from the image (22).

Abstract translation: 对象分类（10）包括提供图像设备（28）提供选择预定成像特征的步骤的方法，从由基于对象的期望的分类所述成像装置（28）产生的图像中提取（14） ，并获得由所述至少一个对象的摄像装置（28）的图像（16）在视摄像装置（28）的一个字段。 的方法（10）进一步包括从所述图像（18）中提取至少一个特征的步骤worin所述至少一个特征对应于所述预定成像特征的，确定性的采矿对于每个提取的至少一个特征的值（21） 和分类基于所述至少一个特征，可以从图像（22）中提取对象。

公开(公告)号：EP1655688A2

公开(公告)日：2006-05-10

申请号：EP05077317.5

申请日：2005-10-11

Applicant: Delphi Technologies, Inc.

Inventor： Zhang, Yan ,  Kiselewich, Stephen J.

IPC: G06K9/52 ,  B60R21/01

CPC classification number: G06K9/00369 ,  B60R21/01538 ,  G06K9/4614 ,  G06K9/527

Abstract: A stream of images including an area occupied by at least one object are processed to extract wavelet coefficients, and the extracted coefficients are represented as wavelet signatures that are less susceptible to misclassification due to noise and extraneous object features. Representing the wavelet coefficients as wavelet signatures involves sorting the coefficients by magnitude, setting a coefficient threshold based on the distribution of coefficient magnitudes, truncating coefficients whose magnitude is less than the threshold, and quantizing the remaining coefficients.

Abstract translation: 处理包括至少一个对象占据的区域的图像流以提取小波系数，并且将所提取的系数表示为由于噪声和无关对象特征而不易于错误分类的小波特征。 将小波系数表示为小波特征涉及按系数排序，设定基于系数幅度分布的系数阈值，幅度小于阈值的截断系数，以及量化剩余系数。