CN103699899B - Equidistant curve model based on lane line detection method - Google Patents

Equidistant curve model based on lane line detection method Download PDF

Info

Publication number
CN103699899B
CN103699899B CN 201310717643 CN201310717643A CN103699899B CN 103699899 B CN103699899 B CN 103699899B CN 201310717643 CN201310717643 CN 201310717643 CN 201310717643 A CN201310717643 A CN 201310717643A CN 103699899 B CN103699899 B CN 103699899B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
point
line
step
corresponding
lane
Prior art date
Application number
CN 201310717643
Other languages
Chinese (zh)
Other versions
CN103699899A (en )
Inventor
付梦印
王新宇
杨毅
朱昊
屈新
Original Assignee
北京理工大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

本发明公开一种基于等距曲线模型的车道线检测方法,采用方法可识别车载摄像机图像中的多条车道线,识别范围大,识别结果不限于直线,圆弧等简单曲线,对变曲率半径的车道线也可进行识别;且可对每条车道线的种类进行识别,区分实线与虚线,识别准确率高。 The present invention discloses a lane line equidistant curve model-based detection method, a method may identify a plurality of lanes lines onboard camera images, a large range of recognition, the recognition result is not limited to simple linear curves, arcs, etc., to change the radius of curvature the lane line can be identified; and may be identified for each lane mark type, distinguishing the solid and dashed lines, high recognition accuracy. 该方法包括对车载摄像机获得的车辆前方的全景图像进行逆投影变换的步骤、对地面俯视图进行滤波及二值化预处理的步骤、对二值化预处理后图像进行Hough变换的步骤以及在Hough空间内进行车道线及其种类识别的步骤。 The method comprises the panoramic image obtained by a camera on board the vehicle in front of the inverse projective transformation step, a plan view of the ground is filtered and binarized pretreatment step, after the step of Hough transform of the binarized image and the pre-Hough and the step of identifying the type of the lane line is a space.

Description

基于等距曲线模型的车道线检测方法 Equidistant curve model based on lane line detection method

技术领域 FIELD

[0001] 发明设及一种车道线检测方法,具体设及一种基于等距曲线模型的车道线检测方法,属于图像处理技术领域。 [0001] The invention is provided, and one lane line detection method is provided, and one specific method for detecting a lane line based on equidistant curve model, which belongs to technical field of image processing.

背景技术 Background technique

[0002] 现今社会,汽车已经成为人类不可或缺的交通工具之一。 [0002] modern society, the car has become an indispensable tool for human transportation. 随着汽车使用率的不断扩增,其安全问题日益突出,来自美国和中国交通部口的相关资料显示,汽车驾驶的过程中车辆因为偏离车道而产生的事故是交通事故的一个重要部分。 With the amplification of automobile usage, security issues become increasingly prominent, relevant information from the US Department of Transportation and the China ports show the process of driving a car accident because the vehicle deviates from the lane produced is an important part of traffic accidents. 另一个报告指出,近51%的交通事故与驾驶员没有意识到车辆偏离了本车道有很大关系。 Another report noted that nearly 51 percent of traffic accidents and the driver did not realize the vehicle deviates from the lane a great relationship. 因此,需要一种可识别出车道位置的设备,在车辆快要偏离车道时给驾驶员预警信号。 Therefore, a device can identify the position of the lane, the driver warning signal when the vehicle is about to deviate from the lane.

[0003] 作为未来智能交通系统(ITS)的重要部分,无人驾驶汽车将在人们的生活中起到越来越重要的作用。 [0003] As an important part of the future of intelligent transportation systems (ITS), and unmanned vehicles will play an increasingly important role in people's lives. 车道线检测系统作为无人车环境感知的重要部分,在结构化道路环境的感知中起到核屯、作用。 Lane detection system as an important part of the unmanned vehicle environment perception, nuclear Tuen played the role in the perception of structural road environment. 随着无人车水平的不断进步,其智能化程度不断提高,其决策系统需要得到更多路面信息W做出更智能的决策,运对车道线检测系统提出了更高的要求,需要车道线检测系统能够提供能加全面的车道线信息,包括多条车道线的位置,形状,类型等。 With the continuous improvement of the level of unmanned vehicles, and continuously improve its intelligence, its decision-making decision-making system needs to be more information on W pavement to make more intelligent, transport of lane detection system put forward higher requirements, require lane line the detection system can be added to provide comprehensive information lane line, comprising a plurality of lane line position, shape, type and the like.

[0004] 目前,基于视觉的车道线检测技术大多限于单车道左右两侧车道线的检测,检测范围较小。 [0004] Currently, vision-based lane detection technology is mostly limited to the detection of the left and right lane lines on both sides of the lane, the detection range is small. 而对车道线的形状,往往使用直线或特定曲线近似,如圆弧,抛物线等。 The shape of the lane line, tend to use a particular linear or curve approximation, such as a circular arc, parabola. 当检测范围扩大后,近似在较复杂的路面上,特别是变曲率的弯道上,该种检测技术并不能很好的表示车道线的实际位置和形状,甚至影响识别率。 When the expanded detection range, approximately in the more complex road, especially on the curvature of the curve of variation of the technique made not well represent the actual position and shape of the lane line, even affect the recognition rate. 另外,现有技术大多没有对车道线的类型进行识别,如实线和虚线。 Further, the prior art generally do not have to identify the type of the lane line, solid line and broken line. 运些问题在一定程度上制约了车道线检测技术在无人车等对检测结果要求较高的领域的应用。 Shipped some problems to some extent, restricted the lane detection technology in unmanned vehicles, and other test results demanding application areas.

[0005] 由此可见,先进的车道线检测系统在W上领域发挥着举足轻重的作用,而车道线检测算法则是检测系统的核屯、,决定了检测系统的性能,对提高有人驾驶车的安全性,无人驾驶车的智能性有着很高的应用价值。 [0005] Thus, advanced lane detection systems play on the field W pivotal role, and lane detection algorithm is Tuen nuclear detection system ,, determines the performance of the detection system, to improve the manned vehicles security, unmanned vehicles have high intelligence value.

发明内容 SUMMARY

[0006] 有鉴于此,本发明提供一种基于等距曲线模型的车道线检测方法,车道线识别范围大,且对每条车道线的种类都进行识别,实时性好,识别准确率高。 [0006] Accordingly, the present invention provides a lane line equidistant curve model-based detection methods, large lane line recognition range, and each of lane mark types are identified, real-time, high recognition accuracy.

[0007] 该方法的具体步骤为: [0007] The specific steps of the process are:

[000引步骤一:对车载摄像机获得的车辆前方的全景图像进行逆投影变换,得到地面俯视图; [Cited Step 000 a: panoramic image obtained by a camera on board the vehicle in front of the inverse projective transformation to obtain a plan view of the ground;

[0009] 步骤二:对步骤一中得到的地面俯视图进行滤波及二值化预处理; [0009] Step two: on the ground obtained in step a plan view filtering and binarization pretreatment;

[0010] 步骤Ξ:对二值化预处理后的地面俯视图进行化U曲变换,得到的化U曲空间;所述化U曲空间的横轴为目,纵轴为。 [0010] Step Ξ: on the ground after binarized preprocessing plan view of the U curve for conversion of U curved space obtained; the horizontal axis of the U curve for the head space, the vertical axis represents.

[0011] 步骤四:采用下述步骤找到化U曲空间中所有车道线在各个角度上的切线对应点: [0011] Step Four: the following procedure is found in all of the U curve lane line tangent space corresponding points on the respective angles:

[0012] (401)在化ugh空间中找到亮度值最大的像素点作为全局最大值点A,令全局最大值点A所在车道线为主车道线; [0012] (401) find the maximum luminance value of the pixel ugh space as a global maximum point A, so that the global maximum point where the lane line A main lane line;

[001引(402)在化U曲空间中找到与全局最大值点A在同一纵轴上,其它车道线的切线对应点,作为相应车道线的主切线对应点; [001 primer (402) to find the global maximum points on the same vertical axis A, corresponding to the tangent point of the other lane line, a main cut line points as respective lane line curved in the U of space;

[0014] (403)在化U曲空间内找到主车道线的所有切线对应点,形成切线对应点队列;所述主车道线的所有切线对应点为一串连通的亮点; [0014] (403) found in all spatial curvature of the U corresponding to the tangent point of the main lane line is formed corresponding to the tangent point array; highlights all the points corresponding to the tangent line of the main lane of a string of communication;

[0015] (404)判断切线对应点队列中像素点的个数是否大于设定值,若大于表明主车道线为曲线,进入步骤(405);否则表明主车道线为直线,进入步骤六; [0015] (404) determines whether the number of pixels corresponding to the tangent point array than the set value, if greater than a curved line indicates that the main lane, go to step (405); otherwise the lane line is a straight line indicates that the main proceeds to step six;

[0016] (405)将主车道线切线对应点队列中的所有切线对应点整体上下平移至步骤(402)中找到的每条车道线的主切线对应点处,得到化U曲空间内相应车道线的所有切线对应点; [0016] (405) corresponding to all the tangent point across the vertical level of the main lane line tangent point corresponding to the queue to step (402) corresponding to the tangent point of the main line is found in each lane to give the corresponding space of the curved lane U All the points corresponding to the tangent line;

[0017] 步骤五:依据化U曲空间中每条车道线的所有切线对应点,计算出地面俯视图中相应车道线切线的切点坐标,将同一条车道线上的所有切点连接得到车道线; [0017] Step Five: All corresponding points based on the tangent space of the U curve for each lane mark, calculated ground plan view corresponding lane line tangent to tangent point coordinates, the points are connected with all the cut line to obtain a lane line lane ;

[0018] 步骤六:计算每条车道线在二值化后图像中亮线的平均长度值及方差:若亮线的平均长度值符合设定标准且方差小于设定值,则表面该车道线为虚线;否则为实线。 [0018] Step Six: the average calculated for each lane line length value and the variance of the binarized image in the bright line: If the value of the average length of the bright line meets set criteria and variance is less than the set value, the surface of the lane line as a dashed line; otherwise a solid line.

[0019] 在所述步骤(402)中采用下述步骤找到与全局最大值点A在同一纵轴上,其它车道线的切线对应点: [0019] The following procedure was used to find the global maximum points on the same longitudinal axis A, other lane line tangent to the corresponding point in step (402):

[0020] (4021)在化U曲空间中全局最大值点A所在纵轴上,找出设定范围内所有像素点中亮度值最大的像素点作为局部最大值点B;则主车道宽度即为全局最大值点A与局部最大值点B之间的距离H; [0020] (4021) in the curved space of the U on the global maximum point where the longitudinal axis A, find the maximum pixel within the set range of the luminance values ​​of all pixels as a local maximum point B; lane width i.e. the main a global maximum point and the local maximum distance H between the point B;

[0021] (4022)¾线段|AB|为基准,将全局最大值点A所在纵轴划分为多条宽度为Η的线段;然后W除全局最大值点A和局部最大值点BW外的每个等分点为中屯、,划分设定区域的矩形范围作为车道线候选区域; [0021] (4022) ¾ segment | AB | as a reference, the global maximum point A where the longitudinal axis is divided into a plurality of width segments Η; then, in addition to the global maximum W per point A and the local maximum point of the BW Tun aliquots point in rectangular area ,, divided region is set as the lane line candidate region;

[0022] (4023)找出每个车道线候选区域内亮度值最大的像素点,将其亮度值与设定的阔值比对,保留大于该阔值的像素点,作为相应车道线的主切线对应点。 [0022] (4023) find the maximum pixel within each lane line candidate region luminance value, the luminance value which is set wider than the value of retained pixels greater than the width value, corresponding to a main lane line the corresponding tangent point.

[0023] 所述步骤(403)中,采用下述步骤找到主车道线的所有切线对应点: [0023] The step (403), the following procedure is to find the main lane line tangent to all corresponding points:

[0024] (4031)将全局最大值点A作为当前点;同时建立切线对应点队列,将全局最大值点A存入切线对应点队列中;下述中W化U曲空间的横轴方向为左右、纵轴方向为上下; [0024] (4031) the global maximum point A as the current point; while establishing a corresponding tangent point queue, the global maximum point corresponding to the tangent point A is stored in the queue; below the horizontal axis W of the curved space is U left and right, up and down the longitudinal axis;

[0025] (4032)在当前点的上、右上、右、右下、下五个相邻像素点中找到除切线对应点队列中已存在像素点W外的亮度值最大的像素点; [0025] (4032) at the current point, top right, right, lower right, the five neighboring pixels found in the largest pixel luminance value of pixels outside the tangent line W exists in addition to the corresponding points in the queue;

[0026] (4033)将步骤(4032)中找到的像素点的亮度值与设定的阔值比对:若其亮度值大于该阔值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4032);若不大于,则进入(4034); [0026] (4033) The step (4032) of the luminance values ​​of pixels found to match the width value setting: If the value is greater than the width of the brightness value, the pixels corresponding to the tangent point array is added; then, the pixel as the current point, returns to (4032); if not greater than, the entry (4034);

[0027] (4034)将全局最大值点A作为当前点; [0027] (4034) the global maximum of the current point as the point A;

[00%] (4035)在当前点的上、左上、左、左下、下五个相邻像素点中找到除切线对应点队列中已存在像素点W外的亮度值最大的像素点; [00%] (4035) at the current point, the upper left, left, lower left, the five neighboring pixels found in the largest pixel luminance value of pixels outside the tangent line W exists in addition to the corresponding points in the queue;

[0029] (4036)将步骤(4035)中找到的像素点的亮度值与设定的阔值对比:若其亮度值大于该阔值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4035);若不大于,则进入步骤(404)。 Wide comparison with the set value of luminance pixel values ​​found in [0029] (4036) The step (4035): If the value is greater than the width of the brightness value, the pixels corresponding to the tangent point array is added; then as the current pixel point, it returns to (4035); if yes, it proceeds to step (404).

[0030] 在所述步骤一中:首先划定地面俯视图的范围,然后采用下述公式进行逆投影变换处理,得到所划定的地面俯视图范围内每个像素点的RGB值: [0030] In step a: first scoping plan view of the ground, then the following equation using an inverse projective transformation to give a top surface delimited RGB values ​​of each pixel in FIG range:

[0031] Pck = R*kRekRT 邮(Pwk 巧) [0031] Pck = R * kRekRT Post (Qiao PWK)

[0032] 其中: [0032] wherein:

Figure CN103699899BD00071

[0033] [0033]

Figure CN103699899BD00072

Pwk为地面俯视图中像素点k的RGB 值;Pck为在全景图像中与像素点k对应的像素点的RGB值,Φ为像素点k在摄像头球面坐标系下的侧倾角度,Θ为像素点k在摄像头球面坐标系下的俯仰角度,Φ为像素点k在摄像头球面坐标系下的旋转角度;h为车载摄像机的光屯、距离地面的高度。 Pwk to ground RGB value of the pixel point k plan view; Pck is the RGB values ​​of pixel points in the panoramic image and the pixels corresponding to k, Φ pixel point k roll angle at the imaging head plane coordinate system, Θ is the pixel k head pitch angle at the imaging plane coordinate system, [Phi] is the rotation angle k pixels at the imaging surface of the head coordinate system; H Tun light vehicle camera, height from the ground.

[0034] 所述步骤二中采用二维高斯核对地面俯视图进行滤波处理,提取地面俯视图中车辆行驶方向上具有设定宽度的亮线;所述二维高斯核的横向采用二次高斯函数F(m),纵向采用一次高斯函数G(n),二维高斯核中坐标为(m,n)的像素点的值Kvm,n为:Kvm,n,=F(m)G (η);所述: A top [0034] In step two of the two-dimensional Gaussian filter processing for the collation FIG surface, a top surface to extract the bright line drawing having a set width in the vehicle traveling direction; the transverse dimensional Gaussian quadratic Gaussian kernel function F ( m), using a vertical Gaussian function G (n), a two-dimensional Gaussian kernel with coordinates (m, n) pixel value Kvm, n is an: Kvm, n, = F (m) G (η); the above:

Figure CN103699899BD00073

[0035] 其中m为二维高斯核的横向变量,η为二维高斯核的纵向变量,Ον为常数,即所设定的宽度。 [0035] wherein m is a two-dimensional Gaussian kernel variable lateral, longitudinal variable [eta] is a two-dimensional Gaussian kernel, Ον is constant, i.e., set width.

[0036] 所述步骤二中对滤波后的图像进行二值化预处理时,二值化阔值通过直方图分析方式获得,将图像中的所有像素点按亮度值排序,保留图像中亮度值在设定比例内的像素点。 When [0036] Step II of the filtered image is binarized pretreatment wide binary value obtained by histogram analysis mode, all the pixels in the image sorted by the luminance value, the luminance value of image retention pixels within the set ratio.

[0037] 所述步骤五中每次计算地面俯视图中相应车道线切线的切点坐标时采用同一条车道线上相邻的Ξ条切线,车道线上相邻的Ξ条切线对应为化U曲空间内相邻的Ξ个切线对应点,分别为(町^,91^)、片1山1,91,^)、(〇,讯91,山,通过下述公式得到第1条车道线第占'个切线对应点(。^,01^)在地面俯视图中的切点坐标(扣^,71^): [0037] The fifth step of calculating each adjacent ground using the same lane line Ξ tangent plan view when cut point coordinates corresponding to the lane line tangent, Ξ adjacent lane line tangent to the corresponding curve of U Ξ space adjacent tangents corresponding points, respectively (cho ^, 91 ^), 1,91 Hill sheet 1, ^), (square, News 91, the mountain, to give an article of a lane line by the following formula accounting 'tangents corresponding points (^, ^ 01.) at the ground contact point coordinates in plan view (buckle ^, 71 ^):

[00;3 引 [00; 3 lead

Figure CN103699899BD00074

[0039] [0039]

Figure CN103699899BD00081

[0040] 所述步骤(4023)中设定的阔值均为全局最大值点A亮度值的0.2倍。 [0040] The step (4023) values ​​are set to 0.2 times the width of the point A global maximum luminance value.

[0041] 所述步骤(4033)中设定的阔值均为全局最大值点A亮度值的0.2倍。 [0041] The step (4033) values ​​are set to 0.2 times the width of the point A global maximum luminance value.

[0042] 在所述步骤(4021)中,设全局最大值点A的坐标为(μ、Θα),则所述设定范围内所有像素点指:横轴坐标为ΘA,纵轴坐标在[ΓΑ+2.5,ΓΑ+4.5 ]及[rΑ-2.5,ΓΑ-4.5 ]范围内的所有像素点; [0042] In the step (4021), the set point A global maximum coordinates (μ, Θα), is set within a range of all the pixels means: coordinates theta] a horizontal axis, and the vertical axis coordinates [ ΓΑ + 2.5, ΓΑ + 4.5] and [rΑ-2.5, ΓΑ-4.5] all of the pixels within the range;

[0043] 所述步骤(4022)中,所述设定区域的矩形范围为:W相应等分点为中屯、纵向± 0.5m,横向±0.3°的矩形范围。 [0043] The step (4022), the rectangular area is set in the range: W corresponding equant 0.5m, the rectangular range of the village, lateral longitudinal ± a ± 0.3 °.

[0044] 有益效果: [0044] beneficial effects:

[0045] (1)采用该种方法车道线识别范围大,识别结果不局限于直线、圆弧等简单曲线, 还能够可识别变曲率半径的车道线,且能够对车道线的种类进行识别,能够区分实线与虚线;识别准确率局。 [0045] (1) the method using a large range of the lane mark recognition, the recognition result is not limited to a simple linear curves, arc, etc., can also identify the lane line can be variable curvature radius, and is capable of identifying the type of the lane line, able to distinguish between the solid and dashed lines; Bureau recognition accuracy.

[0046] (2)采用该种方法在化U曲空间内找到主车道线的所有切线对应点后,基于等距曲线模型,直接将主车道线的所有切线对应点平移至每条车道线的主切线对应点处,便可得到化U曲空间内该车道线的所有切线对应点,计算效率高。 After [0046] (2) The method to find all the points corresponding to the tangent line of the main lane in the space of the U curve, equidistant curves model, directly to all the main lane line tangent translated to corresponding points of each lane mark corresponding to the tangent point of the main, can be obtained for all the points corresponding to the lane line tangent to the inner space of the U curve, computational efficiency.

[0047] (3)采用二维高斯核滤波,具有一定的抗干扰性,提高检测精度。 [0047] (3) a two-dimensional Gaussian kernel filter, having a certain degree of immunity, improve the detection accuracy.

附图说明 BRIEF DESCRIPTION

[0048] 图1为该方法的流程图; [0048] FIG. 1 is a flowchart for a method;

[0049] 图2为滤波并二值化后的地面俯视图; [0049] Figure 2 is filtered and binarized ground plan view;

[0050] 图3为划分候选区域的示意图; [0050] FIG. 3 is a schematic diagram of the candidate region is divided;

[0051 ]图4为寻找主车道线所有切线对应点的示意图; [0051] FIG. 4 is a main road to find the tangent line corresponds to all points schematic;

[0052] 图5为计算每条切线切点的示意图。 [0052] FIG. 5 is a schematic cut each tangent point is calculated.

具体实施方式 detailed description

[0053] 下面结合附图并举实施例,对本发明进行详细描述。 [0053] The following embodiments in conjunction with the accompanying drawings and embodiments, the present invention will be described in detail.

[0054] 本发明提供一种基于等距曲线模型的车道线检测方法,能够解决现有技术提供的车道线识别方法识别车道线范围窄,数量少,形状限制多,不能对车道线种类进行识别等问题。 [0054] The present invention provides a method for detecting a lane line based on equidistant curve model can be solved lane line recognition method for recognizing a lane line prior art to provide a narrow range, a small number, shape more restrictions, can not be recognized lane mark type And other issues.

[0055] 采用该方法进行车道线检测的具体步骤为: [0055] The specific procedure adopted for the lane line detection method is:

[0056] 步骤一:安装在车辆顶部的车载摄像机获得车辆前方的全景图像。 [0056] Step a: an onboard camera mounted on top of the vehicle front of the vehicle to obtain the panoramic image. 本实施例中,车载摄像机所获得全景图像的像素是3500X 1750,颜色模式为RGB,全景图像所采用的坐标系是球面坐标系。 In this embodiment, the panoramic image pixel is obtained by vehicle-mounted camera 3500X 1750, the color mode is RGB, the panoramic image using the coordinate system is a spherical coordinate system. 然后通过下述步骤对所获得的全景图像进行逆投影变换处理,得到地面俯视图。 Then a projective transformation processing inverse to the panoramic image obtained by the following steps, to give a top surface in FIG.

[0化7] (101)划分地面俯视图的范围: [0 of 7] (101) into the ground plan of FIG range:

[0058] W车体中屯、为原点,车辆的行驶方向为y向,向前为正;车辆的左右方向为X向,向右为正,建立局部坐标系;其中X轴和y轴的单位均为m。 [0058] W Tun vehicle body, the traveling direction of the origin of the vehicle is y-direction, a positive forward; right and left direction of the vehicle X direction, the right is positive, the establishment of local coordinate system; wherein the X-axis and y-axis units are m. 地面俯视图的范围为:x轴[-20,40], y轴[-30,30],共3600平方米的范围;地面俯视图的像素为750 X 750。 FIG ground plan in the range of: x-axis [-20,40], y-axis [-30,30], the range of a total 3,600 square meters; ground plan view of the pixel 750 X 750. 地面俯视图的范围确定后,该图中每个像素点的在局部坐标系下的坐标值已知。 After a plan view of the ground range is determined, the coordinates of the drawing points for each pixel in the local coordinate system values ​​are known.

[0059] (102)采用公式(1)进行逆投影变换处理,得到所划分的地面俯视图范围内每个像素点的RGB值。 [0059] (102) using Equation (1) inverse projective transformation to give the divided ground plan RGB values ​​of each pixel in the range in FIG.

[0060] Pck = R*kRekRT 邮(Pwk 巧) (1) [0060] Pck = R * kRekRT Post (Qiao PWK) (1)

[006。 [006. 其中; among them;

Figure CN103699899BD00091

[0062] [0062]

Figure CN103699899BD00092

Pwk为地面俯视图中像素点k的RGB 值,Pck为像素点k在全景图像中对应像素点的RGB值,Φ为像素点k在全景图像中对应像素点在球面坐标系下的侧倾角度,Θ为像素点k在全景图像中对应像素点在球面坐标系下的俯仰角度,Φ为像素点k在全景图像中对应像素点在球面坐标系下的旋转角度;k=[l,7502],h为车载摄像机的光屯、距离地面的高度。 Pwk to ground plan RGB values ​​of FIG pixel point k, Pck pixel point k corresponding to the RGB values ​​of the pixels in the panoramic image, [Phi] is the pixel k corresponding pixel roll angle in a spherical coordinate system in the panoramic image, Θ pixel point k corresponding to the panoramic image pixel pitch angle in a spherical coordinate system, Φ pixel point k corresponding to the rotation angle of the pixels in the spherical coordinate system in the panoramic image; k = [l, 7502], Tuen h light-vehicle camera height from the ground. 所述侧倾指沿车辆的左右方向偏离中屯、的角度,所述俯仰指沿车辆的行驶方向偏离中屯、的角度,所述旋转指沿竖直方向转动的角度。 Refers to an angle around the roll direction of the vehicle deviates from the village, and refers to an angle with the pitch direction of the vehicle deviates from the village, and the rotating means rotates the angle in the vertical direction.

[0063] 步骤二:采用二维高斯核对步骤一中得到的地面俯视图进行滤波处理,提取地面俯视图中y向上具有设定宽度的亮线,采用该种方法滤波具有一定的抗干扰性,提高识别精度。 [0063] Step Two: using two-dimensional Gaussian ground collation obtained in step a plan view of a filtering process to extract the y-floor plan view of a set of bright line width upwardly, with this method has certain anti-interference filter, to improve the recognition accuracy. 其中高斯核的横向采用二次高斯函数 Wherein the Gaussian kernel transverse Gaussian quadratic

Figure CN103699899BD00093

从向为一次高斯函数 Once from the Gaussian function

Figure CN103699899BD00094

刚二维高斯核中坐标为(m,η)的像素点的值Kvm, η为: K™,n = F(m)G(n);其中m为二维高斯核的横向变量,η为二维高斯核的纵向变量,Ον为常数,即所设定的宽度,本实施例中6〇v=Wv,Wv为车道线的常规宽度。 Kvm pixel values ​​of just two-dimensional Gaussian kernel with coordinates (m, η) is, η as: K ™, n = F (m) G (n); where m is a lateral two-dimensional Gaussian kernel variable, η is variable longitudinal dimensional Gaussian kernel, Ον is constant, i.e., set width, in the embodiment of the present embodiment 6〇v = Wv, Wv conventional lane line width.

[0064] 然后对滤波后的图像进行二值化预处理,二值化阔值通过直方图分析方式获得, 将图像中的所有像素点按亮度值排序,仅保留图像中最亮的5%的像素点。 [0064] The image is then filtered binarized pretreatment wide binary value obtained by histogram analysis mode, all the pixels in the image sorted by the luminance value, the image remains only the brightest 5% pixel. 滤波及二值化处理后的地面俯视图如图2所示。 Ground filtered and binarization processing shown in plan view in FIG. 2.

[0065] 步骤Ξ:对二值化预处理后图像进行化U曲变换 [0065] Step Ξ: preprocessing of binarized image conversion curve of U

[0066] 在进行化u曲变换时,Hou曲空间的横轴为目,分辨率为0.2°,范围为[-45°,45°]; 化U曲空间的纵轴为r,分辨率为0.08米,范围为[-80,80]米。 [0066] During transformation of u curved abscissa Hou space is curved mesh, a resolution of 0.2 °, the range of [-45 °, 45 °]; longitudinal axis of the curved space is U r, resolution 0.08 m, in the range [-80,80] m. 化U曲变换时采用公式(2): When using a formula of the transformation curve U (2):

[0067] r=xcos目+ysin目(2) [0067] r = xcos + ysin mesh mesh (2)

[0068] 步骤四:采用下述步骤找到化U曲空间中所有车道线在各个角度上的切线对应点。 [0068] Step Four: the following procedure is found in all of the U curve lane line tangent space corresponding points on the respective angles. 原理为:若车道线为直线,则对应在化ugh空间中为一个点;若车道线为曲线,则对应在Hou曲空间中为一条曲线,且该曲线上的一个点对应车道线的一条切线;因此将在化U曲空间中与车道线对应的点称为切线对应点。 Principle: if the lane line is a straight line, corresponds in of ugh space to a point; if the lane line is a curve, corresponds in Hou curved space as a curve and a tangent of a point corresponding to the lane mark on the curve ; therefore the curvature of the U space and referred to as a lane line corresponding to the tangent point corresponding points.

[0069 ] (401)在化ugh空间中找到亮度值最大的像素点作为全局最大值点A,为描述方便, 将全局最大值点A所在车道线称为主车道线。 [0069] (401) found in the space of ugh maximum luminance value of pixels as the global maximum point A, for ease of description, the global maximum point A where the lane line called a main lane line.

[0070] (402)设全局最大值点A的坐标为(γα、Θα)。 [0070] (402) A set of global maximum point coordinates (γα, Θα). 常规的车道宽度在[2.5m,4.5m]之间,为此在Hough空间内,从横轴坐标为ΘA,纵轴坐标在[rA+2.5, rA+4.5 ]及[rA-2.5, rA-4.5 ]范围内的所有像素点中找出亮度值最大的像素点作为局部最大值点Β(γβ、Θβ),将局部最大值点B在化ugh空间内的纵轴坐标与全局最大值点A在化ugh空间内的纵轴坐标相减并取绝对值得到当前车道宽度H,即H=|rA-n|。 Conventional lane width in [2.5m, 4.5m] between, for this purpose in the Hough space, the coordinates theta] a horizontal axis, and the vertical axis coordinates [rA + 2.5, rA + 4.5] and [rA-2.5, rA- 4.5] in the range of all the pixels in the pixel find the maximum brightness value as the local maximum point Β (γβ, Θβ), the longitudinal axis B of the local maximum point coordinates of the point a in the global maximum of space ugh ugh the longitudinal axis of the coordinate space, and the absolute value of the subtraction current lane width obtained H, i.e., H = | rA-n |.

[0071] (403)依据步骤(402)计算得到的车道宽度,沿化U曲空间的纵轴方向划分车道线候选区域;划分车道线候选区域时,首先W IAB I为基准,将r=M的直线划分为多条宽度为Η 的线段;然后W除全局最大值点A和局部最大值点BW外的每个等分点为中屯、,划分纵向± 0.5m,横向±0.3° (避免图像采集或处理过程中出现的崎变)的矩形范围作为车道线候选区域,如图3所示。 [0071] (403) according to step (402) calculating a lane width obtained by dividing lane line candidate area along the longitudinal direction of the curved space U; when dividing lane line candidate area, the first W IAB I as a reference, the r = M a straight line is divided into a plurality of width segments Η; and W each aliquot global maximum point except the point a and the local maximum point of the BW ,, divided Tun longitudinal ± 0.5m, lateral ± 0.3 ° (to avoid of distortion) image acquisition or processing occurs in the coordinate lane line candidate area, as shown in FIG.

[0072] (404)找出车道线每个候选区域内亮度值最大的像素点,将其亮度值与设定的阔值(0.2倍的全局最大值)对比,保留大于该阔值的像素点,作为对应车道线的主切线对应点;(若该车道线为直线,则该像素点点对应为车道线;若该车道线为曲线,该像素点对应为该车道线的主切线); (Global maximum of 0.2 times) [0072] (404) find the maximum point of the lane line pixel luminance value for each candidate region, which is the width of the luminance value and the set value comparison, the retention is greater than the pixel width value as a main cut line point corresponding to the lane line; (if the lane line is a straight line, then the pixel bit corresponding to the lane line; if the lane line is a curve, the corresponding pixels of the main lane that a tangent line);

[0073] (405)采用下述步骤在化U曲空间内找到主车道线的所有切线对应点,如图4所示; [0073] (405) uses the following steps to find all points corresponding to the tangent line of the main lane in the space of the U curve, shown in Figure 4;

[0074] (4051)将全局最大值点A作为当前点;同时建立切线对应点队列,将全局最大值点A存入切线对应点队列中;下述中W化U曲空间的横轴方向为左右、纵轴方向为上下; [0074] (4051) the global maximum point A as the current point; while establishing a corresponding tangent point queue, the global maximum point corresponding to the tangent point A is stored in the queue; below the horizontal axis W of the curved space is U left and right, up and down the longitudinal axis;

[0075] (4052)在当前点的上、右上、右、右下、下五个相邻像素点中找到除切线对应点队列中已存在像素点W外的亮度值最大的像素点; [0075] (4052) at the current point, top right, right, lower right, the five neighboring pixels found in the largest pixel luminance value of pixels outside the tangent line W exists in addition to the corresponding points in the queue;

[0076] (4053)将其亮度值与设定的阔值(全局最大值的0.2倍)对比:若大于该阔值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4052);若不大于,进入(4054); [0076] (4053) which is the width of the luminance value and the set value (0.2 times the global maximum) comparison: if the value is greater than the width, added to the pixel corresponding to the tangent point array; then the current pixel as a point, returns to (4052); if not greater than, entering (4054);

[0077] (4054)将全局最大值点A作为当前点; [0077] (4054) the global maximum of the current point as the point A;

[0078] (4055)在当前点的上、左上、左、左下、下五个相邻像素点中找到除切线对应点队列中已存在像素点W外的亮度值最大的像素点; [0078] (4055) at the current point, the upper left, left, lower left, the five neighboring pixels found in the largest pixel luminance value of pixels outside the tangent line W exists in addition to the corresponding points in the queue;

[0079] (4056)将其亮度值与设定的阔值(全局最大值的0.2倍)对比:若大于该阔值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4055);若不大于,贝U 进入(4057)。 [0079] (4056) which is the width of the luminance value and the set value (0.2 times the global maximum) comparison: if the value is greater than the width, added to the pixel corresponding to the tangent point array; then the current pixel as a point, returns to (4055); if not greater than, shellfish into the U (4057).

[0080] (4057)判断切线对应点队列中像素点的个数是否大于设定值,若大于表明主车道线为曲线,进入步骤(406);否则表明主车道线为直线,进入步骤六。 [0080] The number (4057) corresponds to the tangent point determination pixel is greater than the set value of the queue, if more than a curved line indicates that the main lane, go to step (406); otherwise the lane line is a straight line indicates that the main proceeds to step six.

[0081] (406)将主车道线的所有切线对应点整体上下平移至步骤(404)中得到的每条车道线的主切线对应点处,即:使主车道线上的主切线对应点(全局最大值点A)与相应车道线的主切线对应点重合,便可得到化U曲空间内该车道线的所有切线对应点; [0081] (406) corresponding to all the tangent point across the vertical level of the main line of the lane to step (404) corresponding to the tangent point of the main line in each lane obtained, namely: the corresponding tangent point of the main line of the main lane ( global maximum point a) coinciding with the main corresponding point corresponding to the tangent line of the lane, can be obtained corresponding to the tangent points of all the lane line space within the U-curve;

[0082] 步骤五:依据化U曲空间中每条车道线的切线对应点,计算出地面俯视图中每条切线的切点坐标,将同一条车道线上的所有切点连接得到车道线。 [0082] Step Five: based on the tangent space of the U curve corresponding points of each lane mark, calculate ground coordinates of the plan view of each cut tangent, tangent point to connect with all of the obtained lane line lane line.

[0083] 计算切点时使用同一条车道线上相邻的Ξ条切线,车道线上相邻的Ξ条切线对应为Hou曲空间内相邻的;个切线对应点,如图5所示;分别为(ri,w,9i,w),采用下述公式(3)计算: Use the same adjacent lane line calculation of the tangent point [0083] Ξ tangents, adjacent lane line corresponding to a Cascade tangent to the curved space adjacent to Hou; tangents corresponding points, shown in Figure 5; respectively (ri, w, 9i, w), using the following equation (3) is calculated:

Figure CN103699899BD00111

[0086] 其中(XI, 为第i条车道第j个切线对应点在局部坐标系下的坐标值。 [0086] where (XI, for the i th cut line lanes j-th point coordinate in the local coordinate system.

[0087] 步骤六:计算每条车道线在二值化后图像中的亮线的平均长度值及方差,并据此判断车道线是否为虚线:若亮线的平均长度值符合设定标准且方差小于设定值(保证每段长度均匀)则表面该车道线为虚线;否则为实线。 [0087] Step Six: the average calculated for each lane line length value and the variance of the bright lines in the image after binarization, and judge whether the lane line is a broken line: If the average length value of bright lines meet set criteria and variance is less than the set value (to ensure a uniform segment length) of the surface of the lane line is a broken line; otherwise a solid line.

[0088] 该种车道线识别方法既可用于驾驶辅助警告系统,为驾驶员的安全可靠驾驶提供支持,更可用于无人驾驶车辆环境感知系统,使无人驾驶车辆能够更好的对路面信息进行理解与决策,实用性强,具有较强的应用价值。 [0088] The kind lane detection method can be used to drive an auxiliary warning system, to support the safe driving of the driver, but also for unmanned vehicle environment-aware system that allows unmanned vehicle on the road to better information be understanding and decision-making, practical, with a strong value.

[0089] 综上所述,W上仅为本发明的较佳实施例而已,并非用于限定本发明的保护范围。 [0089] In summary, the present invention is only preferred embodiments of the W, not intended to limit the scope of the present invention. 凡在本发明的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本发明的保护范围之内。 Any modification within the spirit and principle of the present invention, made, equivalent substitutions, improvements, etc., should be included within the scope of the present invention.

Claims (10)

  1. 1. 基于等距曲线模型的车道线检测方法,其特征在于: 步骤一:对车载摄像机获得的车辆前方的全景图像进行逆投影变换,得到地面俯视图; 步骤二:对步骤一中得到的地面俯视图进行滤波及二值化预处理; 步骤三:对二值化预处理后的地面俯视图进行Hough变换,得到的Hough空间;所述Hough空间的横轴为9,纵轴为r; 步骤四:采用下述步骤找到Hough空间中所有车道线在各个角度上的切线对应点: (401) 在Hough空间中找到亮度值最大的像素点作为全局最大值点A,令全局最大值点A 所在车道线为主车道线; (402) 在Hough空间中找到与全局最大值点A在同一纵轴上,其它车道线的切线对应点, 作为相应车道线的主切线对应点; (403) 在Hough空间内找到主车道线的所有切线对应点,形成切线对应点队列;所述主车道线的所有切线对应点为一串连通的亮点; (404) 判断切 1. The lane line detection method equidistant curve model based, wherein: step a: the panoramic image of the vehicle-vehicle camera obtained in front of the inverse projective transformation to obtain the ground plan view; Step two: on the ground in step a obtained in plan view filtering and binarization pretreatment; step three: the ground after binarized preprocessing plan view for Hough transform, obtained Hough space; Hough space is the horizontal axis 9, the vertical axis represents R & lt; step four: using the steps to find all the Hough space correspond to a lane line tangent points on the respective angles: (401) find the maximum luminance value of pixels as the global maximum point in the Hough space a, so that the global maximum point where the lane line a main lane line; (402) to find the global maximum points on the same vertical axis a, corresponding to the tangent point of the other lane line, a main cut line points as respective lane line in Hough space; (403) found in the Hough space All the points corresponding to the tangent line of the main lane, corresponding to the tangent point array is formed; all corresponding points of the tangent line of the main lane is a series of highlights communication; (404) determines the cut 对应点队列中像素点的个数是否大于设定值,若大于表明主车道线为曲线,进入步骤(405);否则表明主车道线为直线,进入步骤六; (405) 将主车道线切线对应点队列中的所有切线对应点整体上下平移至步骤(402)中找到的每条车道线的主切线对应点处,得到Hough空间内相应车道线的所有切线对应点; 步骤五:依据Hough空间中每条车道线的所有切线对应点,计算出地面俯视图中相应车道线切线的切点坐标,将同一条车道线上的所有切点连接得到主车道线; 步骤六:计算每条车道线在二值化后图像中亮线的平均长度值及方差:若亮线的平均长度值符合设定标准且方差小于设定值,则表明该车道线为虚线;否则为实线。 If the corresponding number of pixel points in the point array is greater than the set value, if greater than a curved line indicates that the main lane, go to step (405); otherwise the lane line is a straight line indicates that the main proceeds to step six; (405) the main lane line tangent All of the main cut line tangent point corresponding to the whole upper and lower flat points to step (402) of each corresponding point in the queue lane line found to give all the corresponding points in Hough space tangent respective lane lines; step five: based on Hough space All points in each lane corresponding to the tangent line, tangent point coordinates to calculate the ground plan view the respective lane line tangent to connect the main lane line obtained with all lane line tangent point; step six: calculating in each lane line binarized image of the bright line and the variance of the average length value: If the average length value of bright lines meet set criteria and variance is less than the set value, it indicates that the lane line is a broken line; otherwise a solid line.
  2. 2. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:在所述步骤(402) 中采用下述步骤找到与全局最大值点A在同一纵轴上,其它车道线的切线对应点: (4021) 在Hough空间中全局最大值点A所在纵轴上,找出设定范围内所有像素点中亮度值最大的像素点作为局部最大值点B;则主车道宽度即为全局最大值点A与局部最大值点B 之间的距离H; (4022) 以线段| AB |为基准,将全局最大值点A所在纵轴划分为多条宽度为H的线段;然后以除全局最大值点A和局部最大值点B以外的每个等分点为中心,划分设定区域的矩形范围作为车道线候选区域; (4023) 找出每个车道线候选区域内亮度值最大的像素点,将其亮度值与设定的阈值比对,保留大于该阈值的像素点,作为相应车道线的主切线对应点。 2. A method for detecting a lane line based on the model of offset curves claimed in claim 1, wherein: the step (402) employed in said step to find the global maximum points on the same longitudinal axis A, other lane line tangent to corresponding points: (4021) in the Hough space on a global maximum point where the longitudinal axis a, find the maximum pixel within the set range of the luminance values ​​of all pixels as a local maximum point B; the main lane global maximum width is the distance H between the point a and the local maximum point B; (4022) line segments | AB | as a reference, the global maximum point a where the longitudinal axis is divided into a plurality of line width is H; each aliquot was then global maximum point other than the point a and the local maximum point B as the center, divide the rectangular region as range setting lane line candidate region; (4023) to find the brightness of each of the lane line candidate areas the maximum value of pixels, the brightness threshold value and the set alignment, retention pixels greater than the threshold value, a corresponding point corresponding to the tangent as the main lane line.
  3. 3. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤(403) 中,采用下述步骤找到主车道线的所有切线对应点: (4031) 将全局最大值点A作为当前点;同时建立切线对应点队列,将全局最大值点A存入切线对应点队列中;下述中以Hough空间的横轴方向为左右、纵轴方向为上下; (4032) 在当前点的上、右上、右、右下、下五个相邻像素点中找到除切线对应点队列中已存在像素点以外的亮度值最大的像素点; (4033) 将步骤(4032)中找到的像素点的亮度值与设定的阈值比对:若其亮度值大于该阈值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4032);若不大于,则进入(4034); (4034) 将全局最大值点A作为当前点; (4035) 在当前点的上、左上、左、左下、下五个相邻像素点中找到除切线对应点队列中已存在像素点以外 3. A method for detecting a lane line based on the model of offset curves claimed in claim 1, wherein: said step (403), the following procedure is to find all the corresponding tangent point of the main lane line: (4031) The a global maximum value point as the current point; while establishing a corresponding tangent point queue, the global maximum point corresponding to the tangent point a is stored in the queue; following to Hough space is the horizontal axis about the vertical axis direction up and down; ( 4032) at the current point, top right, right, lower right, the five neighboring pixels found in the largest pixel luminance value of pixels existing outside the other cut line point array; (4033) the step (4032 ) threshold luminance pixel value and the set point for the ratio found: if the brightness value is greater than the threshold value, the pixel corresponding to the tangent point array is added; the pixel as the current point and then returns to (4032) ; if so, then entry (4034); (4034) the global maximum of the current point as the point a; (4035), left upper, left, lower left, the five neighboring pixels found on the current inter tangent point a corresponding point in the queue already exists outside pixels 的亮度值最大的像素点; (4036) 将步骤(4035)中找到的像素点的亮度值与设定的阈值对比:若其亮度值大于该阈值,将该像素点加入切线对应点队列中;然后将该像素点作为当前点,返回至(4035);若不大于,则进入步骤(404)。 The luminance value of the maximum pixel; threshold comparison luminance value and the set of pixel point values ​​found in (4036) The step (4035): If the brightness value is greater than the threshold value, the pixel is added cut line point array; the pixel as the current point and then returns to (4035); if yes, it proceeds to step (404).
  4. 4. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:在所述步骤一中:首先划定地面俯视图的范围,然后采用下述公式进行逆投影变换处理,得到所划定的地面俯视图范围内每个像素点的RGB值: Pck=R<t>kR0kRTij)k (Pwk+T) | cos# 0 ij 1 8 0 j I 0 I 其中:双只=| 0 I 0 I 0 cos 辦:sin 9k | J-| 0 | I sin# 0 cos# ! 10 -sin與:cos傲| 卜為| 4. A method for detecting a lane line based on the model of offset curves claimed in claim 1, wherein: in said step a: first scoping plan view of the ground, then the following equation using an inverse projective transformation process, delineated obtained RGB value of each pixel of the ground point in a plan view range: Pck = R <t> kR0kRTij) k (Pwk + T) | cos # 0 ij 1 8 0 j I 0 I wherein: double only = | 0 I 0 I 0 cos do: sin 9k | J- | 0 | I sin # 0 cos # 10 -sin with:! cos proud | BU is |
    Figure CN103699899BC00031
    Pwk为地面俯视图中像素点k的RGB值;Pck为在全景图像中与像素点k对应的像素点的RGB值,巾为全景图像中与像素点k对应的像素点在摄像头球面坐标系下的侧倾角度,9为全景图像中与像素点k对应的像素点在摄像头球面坐标系下的俯仰角度3为全景图像中与像素点k对应的像素点在摄像头球面坐标系下的旋转角度;h为车载摄像机的光心距离地面的高度。 Pwk to ground plan RGB values ​​of FIG pixel point k; Pck the RGB values ​​of pixel points in the panoramic image and the pixels corresponding to k, towel to pixel panoramic image with pixels corresponding to k at the imaging head plane coordinate system roll angle, 9 pixel panoramic image with pixel k corresponding to the pitch angle of the imaging head plane coordinate system 3 is a pixel panoramic image with pixel k corresponding to the rotational angle at the imaging head plane coordinate system; H height from the ground to the optical center-board camera.
  5. 5. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤二中采用二维高斯核对地面俯视图进行滤波处理,提取地面俯视图中车辆行驶方向上具有设定宽度的亮线;所述二维高斯核的横向采用二次高斯函数F(m),纵向采用一次高斯函数G (n ),二维高斯核中坐标为(m,n )的像素点的值Kvm, "为:Kvm, n = F(m)G(n );所述: 5. A method for detecting a lane line based on the model of offset curves claimed in claim 1, wherein: said step two two-dimensional Gaussian filter checking process ground plan view, a top surface to extract the vehicle traveling direction in the drawing is provided with bright line of a given width; the lateral two-dimensional Gaussian quadratic Gaussian function F (m) of the core, the longitudinal using a Gaussian function G (n), a two-dimensional Gaussian kernel with coordinates (m, n) of the pixel value Kvm, "as: Kvm, n = F (m) G (n); said:
    Figure CN103699899BC00032
    其中m为二维高斯核的横向变量,n为二维高斯核的纵向变量,〇v为常数,即所设定的宽度。 Wherein m is a two-dimensional Gaussian kernel laterally variable, n is a variable longitudinal dimensional Gaussian kernel, 〇v is constant, i.e., set width.
  6. 6. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤二中对滤波后的图像进行二值化预处理时,二值化阈值通过直方图分析方式获得,将图像中的所有像素点按亮度值排序,保留图像中亮度值在设定比例内的像素点。 6. A method for detecting a lane line based on the model of offset curves claimed in claim 1, wherein: said step two of the filtered image binarization preprocessing, binarization threshold values ​​by a histogram analysis way to get all the image pixels by the luminance values ​​are sorted to retain the image pixel luminance values ​​in the set proportions.
  7. 7. 如权利要求1所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤五中每次计算地面俯视图中相应车道线切线的切点坐标时采用同一条车道线上相邻的三条切线,车道线上相邻的三条切线对应为Hough空间内相邻的三个切线对应点,分别为(ri,j, 0^)、(^上1,01^1)、(^|1,01| 1),通过下述公式得到第1条车道线第」个切线对应点出+ 0^)在地面俯视图中的切点坐标&^,7^) : Using a lane line at the same time the fifth step of each ground contact point coordinates corresponding plan view lane line tangent calculation: 7. The lane detection model based on the offset curves to claim 1, characterized in that adjacent three tangent, tangent lane line adjacent three points corresponding to three cut line in the Hough space adjacent, respectively (ri, j, 0 ^), (^ 1 ^ on 1,01), ( ^ | 1,01 | 1), the following equation is obtained by a lane line of the article "a corresponding point tangents ^ + 0) point coordinates cut in the ground plan view of the & ^, 7 ^):
    Figure CN103699899BC00041
  8. 8. 如权利要求2所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤(4023)中设定的阈值均为全局最大值点A亮度值的0.2倍。 Lane line detection based on equidistant curve model as claimed in claim 2, wherein: said step (4023) are set 0.2 times the threshold luminance value A global maximum points.
  9. 9. 如权利要求3所述的基于等距曲线模型的车道线检测方法,其特征在于:所述步骤(4033)中设定的阈值均为全局最大值点A亮度值的0.2倍。 Lane line detection based on equidistant curve model as claimed in claim 3, wherein: said step (4033) are set 0.2 times the threshold point A global maximum luminance value.
  10. 10. 如权利要求2所述的基于等距曲线模型的车道线检测方法,其特征在于:在所述步骤(4021)中,设全局最大值点A的坐标为(r A、0A),则所述设定范围内所有像素点指:横轴坐标为9 a,纵轴坐标在[rA+2.5,rA+4.5 ]及[rA_2.5,rA_4.5 ]范围内的所有像素点; 所述步骤(4022)中,所述设定区域的矩形范围为:以相应等分点为中心纵向±0.5m,横向±0.3°的矩形范围。 Lane line detection based on equidistant curve model as claimed in claim 2, wherein: in said step (4021), the coordinates of the global maximum set point A is (r A, 0A), the set within the range of all pixels refers to: the horizontal axis coordinates 9 a, longitudinal coordinates [rA + 2.5, rA + 4.5] and [rA_2.5, rA_4.5] all of the pixels within the range; the step (4022), the rectangular area is set in the range of: dividing points corresponding to the central longitudinal ± 0.5m, a transverse rectangular area of ​​± 0.3 °.
CN 201310717643 2013-12-23 2013-12-23 Equidistant curve model based on lane line detection method CN103699899B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201310717643 CN103699899B (en) 2013-12-23 2013-12-23 Equidistant curve model based on lane line detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201310717643 CN103699899B (en) 2013-12-23 2013-12-23 Equidistant curve model based on lane line detection method

Publications (2)

Publication Number Publication Date
CN103699899A true CN103699899A (en) 2014-04-02
CN103699899B true CN103699899B (en) 2016-08-17

Family

ID=50361421

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201310717643 CN103699899B (en) 2013-12-23 2013-12-23 Equidistant curve model based on lane line detection method

Country Status (1)

Country Link
CN (1) CN103699899B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105528801A (en) * 2015-12-09 2016-04-27 中国西电电气股份有限公司 Generation method of switch equipment mechanical stroke curve enveloping line

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102682292A (en) * 2012-05-10 2012-09-19 清华大学 Method based on monocular vision for detecting and roughly positioning edge of road
CN103136341A (en) * 2013-02-04 2013-06-05 北京航空航天大学 Lane line reconstruction device based on Bezier curve
CN103389733A (en) * 2013-08-02 2013-11-13 重庆市科学技术研究院 Vehicle line walking method and system based on machine vision

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102682292A (en) * 2012-05-10 2012-09-19 清华大学 Method based on monocular vision for detecting and roughly positioning edge of road
CN103136341A (en) * 2013-02-04 2013-06-05 北京航空航天大学 Lane line reconstruction device based on Bezier curve
CN103389733A (en) * 2013-08-02 2013-11-13 重庆市科学技术研究院 Vehicle line walking method and system based on machine vision

Also Published As

Publication number Publication date Type
CN103699899A (en) 2014-04-02 application

Similar Documents

Publication Publication Date Title
Garcia-Garrido et al. Fast traffic sign detection and recognition under changing lighting conditions
US20090295917A1 (en) Pixel-based texture-less clear path detection
US20100098297A1 (en) Clear path detection using segmentation-based method
US20100098295A1 (en) Clear path detection through road modeling
Liu et al. Rear vehicle detection and tracking for lane change assist
US20080137908A1 (en) Detecting and recognizing traffic signs
US20100014714A1 (en) Camera-based lane marker detection
CN101196980A (en) Method for accurately recognizing high speed mobile vehicle mark based on video
Suhr et al. Sensor fusion-based vacant parking slot detection and tracking
CN101894271A (en) Visual computing and prewarning method of deviation angle and distance of automobile from lane line
Sivaraman et al. A review of recent developments in vision-based vehicle detection
CN1945596A (en) Vehicle lane Robust identifying method for lane deviation warning
Huang et al. An intelligent strategy for checking the annual inspection status of motorcycles based on license plate recognition
CN102708356A (en) Automatic license plate positioning and recognition method based on complex background
US20130201334A1 (en) Illumination Invariant and Robust Apparatus and Method for Detecting and Recognizing Various Traffic Signs
CN102632839A (en) Back sight image cognition based on-vehicle blind area early warning system and method
CN101608924A (en) Method for detecting lane lines based on grayscale estimation and cascade Hough transform
Nieto et al. Road environment modeling using robust perspective analysis and recursive Bayesian segmentation
Jiang et al. Lane and obstacle detection based on fast inverse perspective mapping algorithm
Wang et al. A vision-based road edge detection algorithm
Kühnl et al. Monocular road segmentation using slow feature analysis.
CN102722704A (en) Method and system for recognizing vehicle license plate by integrating video dynamic tracking
Kuo et al. Vision-based vehicle detection for a driver assistance system
CN102184413A (en) Automatic vehicle body color recognition method of intelligent vehicle monitoring system
CN102722705A (en) Method for detecting multi-lane line on basis of random sample consensus (RANSAC) algorithm

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model