CN103400150B - A method and apparatus for identifying the road edge based on the mobile platform - Google Patents

A method and apparatus for identifying the road edge based on the mobile platform Download PDF

Info

Publication number
CN103400150B
CN103400150B CN 201310351847 CN201310351847A CN103400150B CN 103400150 B CN103400150 B CN 103400150B CN 201310351847 CN201310351847 CN 201310351847 CN 201310351847 A CN201310351847 A CN 201310351847A CN 103400150 B CN103400150 B CN 103400150B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
image
road
edge
obtained
space
Prior art date
Application number
CN 201310351847
Other languages
Chinese (zh)
Other versions
CN103400150A (en )
Inventor
杨国青
李红
逄伟
刘健全
高辉
吴朝晖
Original Assignee
浙江大学
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

本发明公开了一种基于移动平台进行道路边缘识别的方法及装置,包括使用摄像模块来采集获得路面的实时图像信息,使用移动平台对摄像模块获得的实时图像信息进行处理,将YUV空间图像转换成RGB空间图像;对获得的RGB空间图像转换成HSV空间图像,并对其进行图像二值化处理;同时将获得的RGB空间图像转换为灰度图像,对其进行边缘和直线检测;最后依据上述步骤中获得的道路边缘所在区域,对检测到的直线进行进一步筛选,选取在道路边缘区域中的最长直线来作为道路边缘。 The present invention discloses a method and apparatus for performing a road edge detection based on the mobile platform, including the use of an imaging module to capture images in real time road information obtained using real-time image information of the mobile platform module obtains the processed image, the image is converted YUV space RGB space into an image; an image obtained on the RGB space into HSV space image, and subjected to image binarization process; RGB space while the obtained image is converted to grayscale, and its straight edge detection; Finally, according to where the edges of the road area obtained in the above step, the detected straight line screening and selecting the longest straight line of the road to the edge region as the road edge. 该发明解决了现有的道路边缘识别系统算法需占用较大内存,实时性较差,易受外界干扰,且大部分只适用于检测边缘清晰、较为理想的速公路的问题。 The invention solves the problem of the existing road edge recognition system algorithm needs to take up more memory, the real poor, vulnerable to outside interference, and most only apply to the detection of sharp edges, ideal speed highway.

Description

一种基于移动平台进行道路边缘识别的方法及装置 A method and apparatus for identifying the road edge based on the mobile platform

技术领域 FIELD

[0001] 本发明涉及一种道路边缘识别的方法及装置,尤其是一种基于移动平台进行道路边缘识别的方法及装置。 [0001] The present invention relates to a method and apparatus for identifying the road edge, in particular a method and apparatus for a road edge detection based on the moving platform.

背景技术 Background technique

[0002] 目前电子设备中常用的图像色彩模式有如下几类。 [0002] electronic device is currently common image color mode following categories.

[0003] RGB颜色空间:通过对红(R)、绿⑹、蓝⑻三个颜色通道的变化以及它们相互之间的叠加来得到各式各样的颜色的,RGB即是代表红、绿、蓝三个通道的颜色,该标准几乎包括了人类视力所能感知的所有颜色,是目前运用最广的颜色系统之一。 [0003] RGB color space: by superposition between the red (R), green ⑹, changes in Blue ⑻ the three color channels and to obtain their mutual variety of colors, the RGB that is representative of the red, green, and blue color channels, the standard includes almost all the colors the human eye can perceive, is one of the most widely used color system. RGB颜色模型的主要目的是在电子系统中检测、表示和显示图像,比如电视和电脑,同时在传统摄影中也有应用。 The main purpose is to detect the RGB color model in an electronic system, and represents a display image, such as TVs and computers, while also applied in traditional photography.

[0004] YUV颜色空间:YUV与RGB—样,用于表示颜色,两者可以相互转化。 [0004] YUV color space: YUV and RGB- like, used to represent the color, the two can be transformed into each other. YUV是被欧洲电视系统所采用的一种颜色编码方法(属于PAL) JUV主要用于优化彩色视频信号的传输,使其向后兼容老式黑白电视。 YUV color encoding is a method that is used by the European television system (of PAL) JUV mainly for optimizing the transmission of color video signals, so that backward compatibility with older black and white TV. 与RGB视频信号传输相比,它最大的优点在于只需占用极少的带宽。 RGB video signal transmission as compared with its greatest advantage is that occupies only very little bandwidth. 其中中“Y”表示明亮度,也就是灰阶值,作为基带信号。 Wherein in the "Y" represents brightness, i.e. gray level value, as a baseband signal. 而“U”和“V”表示的则是色度,作用是描述影像色彩及饱和度,用于指定像素的颜色。 And "U" and "V" is represented by chromaticity, it is to describe the image color and saturation, color designation for the pixel. HSV颜色空间:HSV模型的三维表示是从RGB立方体演化而来,其中:色调(H),饱和度(S),亮度(V)。 HSV color space: three-dimensional representation of the HSV model is evolved from the RGB cube, wherein: the hue (H), saturation (S), Brightness (V). 设想从RGB沿立方体对角线的白色顶点向黑色顶点观察,可以看到立方体的六边形外形。 Is contemplated to observe black vertex, a hexagonal shape can be seen from the RGB cube along the cube diagonal vertex white. 六边形边界表示色调(H),水平轴表示饱和度(S),亮度(V)沿垂直轴测量。 Denotes that the hue boundary hexagonal (H), the horizontal axis represents saturation (S), Brightness (V) measured along the vertical axis. H参数表示色彩信息,即所处的光谱颜色的位置。 H parameter indicates the color information, i.e., the position in which the spectral colors. 该参数用一角度量来表示,红、绿、蓝分别相隔120度。 This parameter is represented by a corner metric, red, green, and blue are 120 degrees apart. 互补色分别相差180度。 Complementary Color 180 degrees, respectively. 饱和度S为一比例值,范围从〇到1,它表示成所选颜色的纯度和该颜色最大的纯度之间的比率。 S is a ratio of the saturation value, ranging from square to 1, as it represents the ratio between the maximum and the selected color purity of the color purity. s=o时,只有灰度。 When s = o, only gray. V表示色彩的明亮程度,范围从0到1。 V represents the brightness of the color, ranging from 0 to 1. 由于HSV是一种比较直观的颜色模型,所以在许多图像编辑工具中应用比较广泛。 Because HSV is a more intuitive color model, so in many image editing tool used widely.

[0005] 道路边缘识别装置是通过对路面图像的实时处理,来警告车辆驾驶员在车辆驾驶路径中道路边缘的存在,或者提供用于在自主驾驶时路线计划的可行驾驶区域限制的装置。 [0005] road edge identified by means of real-time processing of the road image, to alert the driver of the presence of the vehicle in a road vehicle driving path of the edge, or to provide a viable means for, when the driving region autonomous driving route planning restrictions. 目前常用的道路边缘检测算法及其不足如下。 Currently used road edge detection algorithm and its lack as follows.

[0006] ⑴Hough变换检测直线方法。 [0006] Detection method ⑴Hough linear transformation.

[0007] Hough变换利用图像空间和Hough参数空间的点一线对偶性,把图像空间中的检测问题转换到参数空间。 [0007] Hough transform space using a dot line image and Hough parameter space duality, converting the detection problem in the image space to the parameter space. 通过在参数空间里进行简单的累加统计,然后在Hough参数空间寻找累加器峰值的方法检测直线。 By simple statistics accumulated in the parameter space, and then the accumulator in Hough parameter space to find the peak line detection method.

[0008] 不足:占用较大内存,实时性较差,易受外界干扰,特别对于道路边缘检测时,阴影、障碍物等对Hough检测道路影响较大。 [0008] Inadequate: when a large memory footprint, the real poor, vulnerable to outside interference, especially for road edge detection, shadow, obstacle detection Hough greater impact on the road.

[0009] (2)基于彩色空间HSV的图像分割法。 [0009] (2) Image Segmentation Based on HSV color space.

[0010] 通过将RGB色彩空间的图像转换为HSV空间的图像,可以通过对感兴趣的色彩信息(如S)进行进一步处理,方法主要有直方图阈值法、聚类法、区域增长法和边缘检测法等。 [0010], can be further processed by the color information of interest (e.g., S) by converting the RGB color space to image space, HSV image, methods are histogram method, clustering, and the edge region growing other detection methods. [0011]不足:色彩空间转换后需要选取合适的图像处理方法来检测道路边缘,效果好的方法往往带来实时性的降低。 [0011] disadvantages: the need to select the appropriate image processing method to detect the road edge color space conversion, often brings good way to reduce the effect of real-time. 阈值选取较难把握。 Threshold selection more difficult to grasp.

[0012] (3)道路模版匹配方法。 [0012] (3) Road template matching method.

[0013]用数学模型的模板对车道进行匹配。 [0013] for lane matching template mathematical model.

[0014]不足:较复杂t旲版库能带来较精确的匹配程度,但是内存与实时性下降较为明显。 [0014] less than: t Dae more complex version of the library can bring more accurate matching degree, but the memory and real-time decline more obvious. 较简单的模版库匹配速度快,但精确度不尚。 Simpler template library to match speed, but accuracy is not yet. 阴影和千扰对匹配效果影响较大。 Shadow peaceful disturbance greater impact on the match result.

[0015] (4)基于特征参数的跟踪方法。 [0015] (4) based on feature tracking parameters.

[0016]基于特征参数的跟踪方法是一种参数估计的方法,其主要是用在建立车道标识线模型的基础上进行的,最具代表性的是各种滤波方法,如卡尔曼滤波。 [0016] Based on the characteristic parameter tracking method is a method of parameter estimation, which is mainly used in the establishment of road markings made on the model, the most representative of various filtering methods, such as Kalman filtering. 以及在其基础上发展起来粒子滤波方法等。 Particle filter and a method developed on the basis thereof.

[0017]不足:由于是用先验概率预测后期结果,所以一旦预测出现误差,后期误差会逐渐增大,所以初次检验需要有较高准确率。 [0017] Inadequate: Because it is the result of a prior probability forecast late, so once the prediction error occurs, the late error will gradually increase, so the initial survey needs to have a high accuracy rate. 该方法计算量很大,需要强大的硬件平台。 The method computationally intensive, requires powerful hardware platform.

发明内容 SUMMARY

[0018]为解决上述问题,本发明提供了一种基于移动平台进行道路边缘识别的方法及装置,以解决现有方法对于阴影严重的路面处理效果差、所需内存大而无法应用于移动平台系统(如Android系统)或应用在移动平台后实时性较差的缺点。 [0018] In order to solve the above problems, the present invention provides a method and apparatus based on a mobile platform identification road edge, so as to solve the conventional method for serious shadow poor road handling, large memory requirements can not be applied to a mobile platform system (such as Android system), or after application of the mobile platform real disadvantage of poor.

[0019] 为实现上述目的,本发明的技术方案为: [0019] To achieve the above object, the technical solution of the present invention is:

[0020] —种基于移动平台进行道路边缘识别的方法,包括如下步骤: [0020] - methods for road edge detection based on a moving platform, comprising the steps of:

[0021] S1:使用摄像模块来采集获得路面的实时图像信息,所述实时图像信息为YUV空间图像; [0021] S1: an imaging module to capture in real time using the image information obtained in the road surface, the real time image to a YUV space image;

[0022] S2:使用移动平台对摄像模块获得的实时图像信息进行处理,将yuv空间图像转换成RGB空间图像; [0022] S2: mobile platform real-time imaging means for obtaining image information are processed, converting the image into an RGB space yuv spatial image;

[0023] S3:将步骤S2中获得的RGB空间图像转换成HSV空间图像,对其进行图像二值化处理,具体包括如下步骤: [0023] S3: converting RGB image obtained in step S2 space into HSV space image, the image subjected to the binarization processing, includes the following steps:

[0024] S:31:将步骤S2中获得的RGB空间图像转换成HSV空间图像; [0024] S: 31: convert the RGB image obtained in step S2 space into HSV space image;

[0025] S32:根据所述HSV空间图像的S域选择阈值; [0025] S32: S domain selection threshold according to the HSV space image;

[0026] S33:通过阈值对HSV空间图像进行二值化处理; [0026] S33: binarizing processing on the image by thresholding the HSV space;

[0027] S:34:对步骤S33获得的二值化图像依次进行膨胀、腐蚀处理,获取道路边缘所在阮域; _] S4:将步骤S2中获得的RGB空间图像转换为灰度图像,对其进行边缘和直线检测具体包括如下步骤: ' ' [0027] S: 34: binarized image obtained in step S33 are sequentially inflated, etching, where the edge of the road acquisition Raney domain; _] S4: The image converting RGB space obtained in step S2 into a grayscale image, for and the linear edge detecting which includes the following steps: '

[0029] S41:将步骤S2中获得的RGB空间图像转换为灰度图像; [0029] S41: The image converting RGB space obtained in step S2 is a grayscale image;

[0030] S42:对灰度图像通过边缘提取算子进行边缘检测; [0030] S42: edge detection on the grayscale image by extracting an edge operator;

[0031] S43:对步骤S42检测后得到的图像进行直线检测; [0031] S43: detecting an image obtained after the step S42 for line detection;

[0032] S44:对步骤S43中检测到的直线进行筛选和优化; [0032] S44: step S43, the detected straight lines screening and optimization;

[0033] S5:依据步骤SM中获得的道路边缘所在区域,对步骤S44中检测到的直线进行、 一步筛选,选取在道路边缘区域中的最长直线作为道路边缘。 [0033] S5: Area road edges obtained according to step SM, the detection of step S44 to be a straight line, the screening step, select the longest straight line in the edge region of the road as the road edge. Enter

[0034]进一步的,所述步骤S32中阈值的选择区域为图像偏下方的一个或由多个图开^ 成的近似梯形的几何区域。 Selection region [0034] S32 further threshold value, a step below the image or partial FIG apart by a plurality of approximately trapezoidal ^ into geometric areas. 所述阈值的计算方法是:分别计算每个几何区域中的8域^= 值,选取这些平均值的中位数,所述中位数即为阈值。 The method of calculating the threshold value is: 8 were calculated for each geometric region domain ^ = value, to select the median of these averages, the median is the threshold value. g g

[0035] 进一步的,所述步骤S33包括:将S域在(ab,a+b)范围内的范围的点设为白色,其余设为黑色点,其中a为所述阈值,b为所选取的几个几何区域中求得的S域平均值的最大值与最小值的差值,所述白色点组成区域即为初步检测出的道路所在区域,黑色点组成区域为初步检测出的道路边缘所在区域。 [0035] Further, the step S33 comprises: a point range in the S domain (ab, a + b) ranges to white, and the rest to black dots, wherein a is the threshold value, b is selected area several road geometry area difference determined maximum and minimum values ​​of the average S domain, the region is the initial white dots detected, the black dots area preliminary detected road edges your region.

[0036] 进一步的,所述步骤S42中边缘检测使用Canny算子、Sobel算子。 [0036] Further, the step S42 using Canny edge detection operator, Sobel operator.

[0037] 进一步的,所述步骤S43的直线检测采用Hough变换检测直线方法。 [0037] Further line detection, the Hough transform step S43 line detection method.

[0038] 进一步的,所述步骤S44具体包括:通过对斜率和位移的分析将碎线段进行连接或删除,通过分析前几帧图像检测出的道路边缘来在图像序列中的连续性来进行直线筛选。 [0038] Further, the step S44 comprises: the analysis of the displacement and slope of the broken line segment connecting or deleted, is linearly by analyzing the first few frames road edge image detected by the continuity of the image sequence filter.

[0039] 一种基于移动平台进行道路边缘识别的装置,包括: [0039] An edge detection apparatus for a road based on the moving platform, comprising:

[0040] 用于采集获得路面的实时图像信息的摄像模块,所述摄像模块获得的实时图像信息为YUV空间图像; [0040] The image information obtained in real time for acquiring road camera module, the camera module in real time image information obtained by the YUV space image;

[0041] 用于将摄像模块获得的实时图像信息进行处理的移动平台,所述移动平台将YUV 空间图像转换成RGB空间图像; [0041] The real-time imaging means for obtaining image information of the processed mobile platform, the mobile platform converts the YUV space into RGB image space image;

[0042] 所述移动平台将RGB空间图像转换成HSV空间图像,移动平台根据所述HSV空间图像的S域选择阈值,通过所述阈值对HSV空间图像进行二值化处理,对获得的所述二值化图像依次进行膨胀、腐蚀处理,获取道路边缘所在区域; [0042] The mobile platform converts the RGB image into HSV space image space, according to the mobile internet domain selection threshold S HSV space image binarized by image processing on the HSV space the threshold of the obtained sequentially binarized image is expanded, etching, acquires area road edges;

[0043] 所述移动平台同时将获得的RGB空间图像转换为灰度图像,移动平台对所述灰度图像通过边缘提取算子进行边缘检测,对所述检测后得到的图像进行直线检测,对所述检测到的直线进行筛选和优化; [0043] RGB space conversion of the image while the mobile platform is obtained gray scale image, the mobile platform operator to extract gray-scale image by the edge of the edge detection image obtained after the detection line detection is performed, for the straight line detected screening and optimization;

[0044] 所述移动平台依据获得的所述道路边缘区域对所述检测到的直线进行进一步筛选,选取在道路边缘区域中的最长直线作为道路边缘。 [0044] The mobile platform is obtained according to the road edge region of the detected straight line for further screening, to select the road edge region as the longest straight road edge.

[0045] 进一步的,所述摄像模块通过0TG线与所述移动平台连接进行数据交换。 [0045] Further, the camera module is connected via the mobile platform 0TG line data exchange.

[0046] 进一步的,所述装置还包括信息处理终端,所述信息处理终端用于接收存储移动平台的数据,所述移动平台将最终处理的结果通过网络与信息处理终端进行交互。 [0046] Further, the apparatus further includes an information processing terminal, the information processing terminal for receiving the data stored in the mobile platform, the mobile platform final result of the process to interact with the information processing terminal via a network.

[0047] 根据以上方案,本发明获得了如下有益效果:(1)基于道路边缘特征:以直线为主, 完全可达到实际应用的程度;(2)基于道路阴影特征:在S域中与道路的非阴影区域非常接近,可以依此去除阴影的影响,同时计算量较小,降低对内存的要求;(3)基于实时图像:由于道路在前后图像中是有延续性的,可依此考虑图像连续性,防止前后检测结果出现较大误差;(4)感兴趣区域:图像分割、阈值选取、实时图像检测等都会限定在某些感兴趣区域内,可减小运算量;(5)降低算法复杂度:通过以上策略,算法复杂度非常低,因此可良好的应用于移动平台;(6)适用于复杂环境:即可检测道路标线,亦可检测道路边缘,所以阴影较多的校园环境和高速公路均可使用。 [0047] According to the above embodiment, the present invention achieves the following advantages: (1) based on the road edge features: a linear-based, fully to the extent practical use; (2) based on the road shadow wherein: S field and road the non-hatched region very close to, and so can affect the shadow removal, while the calculation amount is small, to reduce memory requirements; (3) based on the real-time image: As the image before and after the road there is continuity, and so may be considered image continuity, to prevent a large error occurs before and after the detection result; (4) the region of interest: segmentation, thresholding, real-time image detection are limited to certain regions of interest, the amount of computation can be reduced; (5) reducing algorithm complexity: the above strategy, algorithm complexity is very low, it can be good for mobile platforms; (6) applies to complex environments: to detect road markings, can detect the edge of the road, so the more shaded campus environment and highways can be used.

附图说明 BRIEF DESCRIPTION

[0048]图1为本发明的方法流程示意图。 [0048] FIG. 1 is a schematic flowchart of a method of the present invention.

[0049]图2为本发明的基于移动平台进行道路边缘识别的装置示意图。 Road recognition means a schematic view of an edge [0049] FIG. 2 of the present invention is based on mobile internet.

具体实施方式 detailed description

[0050]为了使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本发明进行进一步详细说明。 [0050] To make the objectives, technical solutions and advantages of the present invention will become more apparent hereinafter in conjunction with the accompanying drawings and embodiments of the present invention will be further described in detail. 应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。 It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention.

[0051]如图1所示,本发明的方法步骤如下: [0051] As shown in FIG 1, the method steps of the present invention are as follows:

[0052] S1:使用摄像模块来采集获得路面的实时图像信息,所述实时图像信息为YUV空间图像,具体如下: [0052] S1: an imaging module to capture in real time using the image information obtained in the road surface, the real time image to a YUV space image, as follows:

[0053]直接通过摄像模块获得YUV空间图像的数据,将其从YUV空间转换为RGB空间图像, 设其为img_src。 Data [0053] YUV space image obtained directly by the imaging module, converts YUV from RGB space to image space, which is provided img_src.

[0054] S2:使用移动平台对摄像模块获得的实时图像信息进行处理,将YUV空间图像转换成RGB空间图像,具体如下: [0054] S2: mobile platform real-time imaging means for obtaining image information are processed, converting the image into an RGB space YUV space with the image, as follows:

[0055] 将img_sr c图像从RGB空间转换为HSV空间图像,设为img_hs v。 [0055] The img_sr c image from RGB space to HSV space image, to img_hs v.

[0056] S3:将步骤S2中获得的RGB空间图像转换成HSV空间图像,对其进行图像二值化处理,具体如下: [0056] S3: converting RGB image obtained in step S2 space into HSV space image, the image subjected to the binarization processing, as follows:

[0057]图像空间转换:将img_src图像从RGB空间转换为HSV空间图像,设为img_hs v。 [0057] Image space conversion: img_src the image from RGB space to HSV space image, to img_hs v.

[0058] 选取二值化阈值:依据img_hsv图像的S域选择最佳阈值,阈值选择的区域为图像偏下方的一个或由多个图形组成的近似梯形的区域。 [0058] Select binarization threshold value: selecting the optimal threshold based on the S field img_hsv image, the selected threshold is a partial region below the image area or a plurality of approximately trapezoidal pattern thereof. 最佳阈值计算方法是通过分别计算每个几何区域中的S域平均值,选取这些平均值的中位数,所述中位数即为阈值,设获得的最佳阈值为a。 The optimum threshold is calculated by calculating the geometric mean of each domain area S in, select the median of these averages, the median is the threshold value, the optimum threshold value is obtained provided a.

[0059]通过阈值a进行图像二值化:将S域在(ab,a+b)范围内的点设为白色点,其余设为黑色点,其中:a为所述阈值,b为所选取的邻域值,所述邻域值b为所选取的几个几何区域中求得的S域平均值的最大值与最小值的差值,所述白色点组成区域即为初步检测出的道路所在区域,黑色点组成区域为初步检测出的道路边缘所在区域。 [0059] by a threshold for image binarization: the point in the S domain (ab, a + b) is defined as the range of the white point, black point remaining set, wherein: a is the threshold value, b is selected value neighborhood, the neighborhood of the maximum and minimum value difference S domain of several geometric mean value of the selected area is obtained, b, the white dots is the initial region detected road area, the black dots area where the regional road edge preliminary detected.

[0060] 对所得到的二值化图依次进行膨胀、腐蚀处理。 [0060] FIG binarized obtained sequentially expansion etching. 所述膨胀步骤是将与物体接触的所有背景点合并到该物体中,使边界向外部扩张的过程。 The expansion step is all background point contact with the object in the object incorporated into the process of expanding the boundary to the outside. 可以用来填补物体中的空洞。 It can be used to fill voids in the object. 在该实施例中,膨胀的算法包括:用3x3的结构元素,扫描图像的每一个像素用结构元素与其覆盖的二值图像做“与”操作,如果都为0,则图像的该像素为0,否则为1,该步骤可使二值图像扩大一圈。 In this embodiment, the expansion algorithm comprising: a structural element of a 3x3, binary image of each pixel with a structural element of the scanned image it covers do "and" operation, if is 0, the image of the pixel 0 otherwise 1, step enables the binary image enlarged circle. 腐蚀是一种消除边界点,使边界向内部收缩的过程。 Corrosion is a boundary point of the elimination of the internal process to shrink the boundary. 可以用来消除小且无意义的物体。 It can be used to eliminate small objects and meaningless. 在该实施例中,腐蚀的算法包括:用3x3的结构元素,扫描图像的每一个像素用结构元素与其覆盖的二值图像做“与”操作,如果都为1,则图像的该像素为1,否则为0。 In this embodiment, the etching method comprising: a structural element of a 3x3, binary image of each pixel with a structural element of the scanned image it covers do "and" operation, if is 1, the image of the pixel 1 0 otherwise. 该步骤能使二值图像减小一圈。 This step can be reduced binary image circle. 上述先膨胀后腐蚀的过程称为闭运算,所述闭运算可用来填充物体内细小空洞、连接邻近物体、平滑其边界的同时并不明显改变其面积。 After expansion of corrosion while the above first process is called closing operation, the closing operation is used to fill small voids within the body, adjacent the connector body, a smooth boundary does not significantly change the area. 通过以上膨胀、腐蚀处理步骤,可以有效去除掉道路中的干扰点,并使道路边缘更加平滑。 Through the above expansion, the etching process step, the interference point can be effectively removed in the road, and the road edge smoother.

[0061] S4:将步骤S2中获得的RGB空间图像转换为灰度图像,对其进行边缘和直线检测, 具体如下: [0061] S4: The image converting RGB space obtained in step S2 into a grayscale image, and its straight edge detection, as follows:

[0062] 将img_src转换为灰度图,设为img_gray。 [0062] The img_src converted to grayscale, set img_gray.

[0063] 边缘检测:对img_gray进行边缘检测,边缘检测可使用Canny算子、Sobel算子等其他边缘检测算子。 [0063] Edge detection: detection of img_gray edge, Canny edge detection operator can be used, like other Sobel operator edge detection operator.

[0064] 对边缘检测后所得图形进行直线检测。 [0064] The resultant pattern after edge detection line detection. 本发明使用Hough变换方法检测直线,并对检测到的直线进行筛选和优化,筛选和优化原则可以考虑斜率和位移等将碎线段进行连接或删除,也可考虑道路边缘在图像序列的连续性,依靠之前几帧图像检测出的道路边缘进行直线筛选。 The present invention uses Hough transform to detect straight line detection and screening and to optimize the selection and optimization of the slope and the principles may be considered to be connected to displacement or remove broken line, is also conceivable road edges in the image the continuity of the sequence, relying on the edge of the road a few frames before the image detected by screening a straight line.

[0065]依据步骤33中获得的道路边缘所在区域,对步骤S4中检测到的直线进行进一步筛选,选取在道路边缘区域中的最长直线作为道路边缘。 [0065] By area where the road edges obtained in step 33, the line detection step S4 for further screening to select the longest straight line in the edge region of the road as the road edge.

[0066]如图2^示,一种基于移动平台进行道路边缘识别的装置,包括: [0066] shown in FIG. 2 ^, based on the mobile platform road edge recognition device, comprising:

[0067]用于采集获得路面的实时图像信息的摄像模块,所述摄像模块获得的实时图像信息为YUV空间图像; [0067] The image information obtained in real time for acquiring road camera module, the camera module in real time image information obtained by the YUV space image;

[0068]用于将摄像模块获得的实时图像信息进行处理的移动平台,所述移动平台将YUV 空间图像转换成RGB空间图像; [0068] The camera for real time image processing module obtained by the mobile platform, the mobile platform converts the YUV space into RGB image space image;

[0069] 所述移动平台将RGB空间图像转换成HSV空间图像,移动平台根据所述HSV空间图像的S域选择阈值,通过所述阈值对HSV空间图像进行二值化处理,对获得的所述二值化图像依次进行膨胀、腐蚀处理,获取道路边缘所在区域; [0069] The mobile platform converts the RGB image into HSV space image space, according to the mobile internet domain selection threshold S HSV space image binarized by image processing on the HSV space the threshold of the obtained sequentially binarized image is expanded, etching, acquires area road edges;

[0070] 所述移动平台同时将获得的RGB空间图像转换为灰度图像,移动平台对所述灰度图像通过边缘提取算子进行边缘检测,对所述检测后得到的图像进行直线检测,对所述检测到的直线进行筛选和优化; [0070] RGB space conversion of the image while the mobile platform is obtained gray scale image, the mobile platform operator to extract gray-scale image by the edge of the edge detection image obtained after the detection line detection is performed, for the straight line detected screening and optimization;

[0071] 所述移动平台依据获得的所述道路边缘区域对所述检测到的直线进行进一步筛选,选取在道路边缘区域中的最长直线作为道路边缘。 [0071] The mobile platform is obtained according to the road edge region of the detected straight line for further screening, to select the road edge region as the longest straight road edge.

[0072] 进一步的,所述摄像模块通过0TG线与所述移动平台连接进行数据交换。 [0072] Further, the camera module is connected via the mobile platform 0TG line data exchange.

[0073]进一步的,所述装置还包括信息处理终端,所述信息处理终端用于接收存储移动平台的数据,所述移动平台将最终处理的结果通过网络与信息处理终端进行交互。 [0073] Further, the apparatus further includes an information processing terminal, the information processing terminal for receiving the data stored in the mobile platform, the mobile platform final result of the process to interact with the information processing terminal via a network.

Claims (3)

  1. 1.一种基于移动平台进行道路边缘识别的方法,其特征在于,包括如下步骤: S1:使用摄像模块来采集获得路面的实时图像信息,所述实时图像信息为YUV空间图像; S2:使用移动平台对摄像模块获得的实时图像信息进行处理,将YUV空间图像转换成RGB空间图像; S3:将步骤S2中获得的RGB空间图像转换成HSV空间图像,对其进行图像二值化处理,具体包括如下步骤: S31:将步骤S2中获得的RGB空间图像转换成HSV空间图像; S32:根据所述HSV空间图像的S域选择阈值,阈值的选择区域为图像偏下方的一个或由多个图形组成的近似梯形的几何区域; S33:通过阈值对HSV空间图像进行二值化处理,将S域在(ab,a+b)范围内的范围的点设为白色,其余设为黑色,其中a为所述阈值,b为所选取的几个几何区域中求得的S域平均值的最大值与最小值的差值,所述白色点组成区域即为初步检 1. A method for identifying an edge of a road based on the mobile platform, characterized by comprising the steps of: S1: an imaging module to capture in real time using the image information obtained in the road surface, the real time image to a YUV space image; S2: mobile real-time internet image information obtained by the imaging processing module, converts the image into RGB space YUV space image; S3: converting RGB space image obtained in step S2 into HSV space image, the image subjected to the binarization processing, comprises the steps of: S31: converting RGB space image obtained in step S2 into HSV space image; S32: the S domain selection threshold HSV space image, the selected area threshold as a downward image partial or by a plurality of graphic composition of approximately trapezoidal geometry area; S33: threshold value by HSV space image binarization process, the S point range in the domain (ab, a + b) ranges to white, and the rest to black, wherein a is the threshold value, b the difference between the maximum and the minimum value of the S domain of several geometric average of the selected area obtained in the area of ​​the white dots is the initial sample 出的道路所在区域,黑色点组成区域为初步检测出的道路边缘所在区域; S34:对步骤S33获得的二值化图像依次进行膨胀、腐蚀处理,获取道路边缘所在区域; S4:将步骤S2中获得的RGB空间图像转换为灰度图像,对其进行边缘和直线检测,具体包括如下步骤: S41:将步骤S2中获得的RGB空间图像转换为灰度图像; S42:对灰度图像通过边缘提取算子进行边缘检测,所述边缘提取算子包括Canny算子、 Sobel算子; S43:对步骤S42检测后得到的图像采用Hough变换检测直线方法进行直线检测; S44:对步骤S43中检测到的直线进行筛选和优化; S5:依据步骤S34中获得的道路边缘所在区域,对步骤S44中检测到的直线进行进一步筛选,选取在道路边缘区域中的最长直线作为道路边缘。 A road Area, black dots area Area road edge preliminary detected; S34: binarized image in step S33 is obtained sequentially expanded, etching, acquires Area road edges; S4: the step S2 RGB space image is converted to grayscale images obtained, and its straight edge detection, comprises the following steps: S41: the image converting RGB space obtained in step S2 is a grayscale image; S42: gray image extracted by the edge edge detection operator, said operator comprises Canny edge detection operator, Sobel operator; S43: S42 after the step of detecting the image obtained using the Hough transform for straight line detection method for detection; S44: step S43 of detected selection and optimization of a straight line; S5: By area where the road edges obtained in step S34, step S44 is detected straight line screening and selecting the longest straight line in the edge region of the road as the road edge.
  2. 2. 根据权利要求1所述的基于移动平台进行道路边缘识别的方法,其特征在于,所述步骤S32中的阈值的选取方法包括:分别计算每个几何区域中的S域平均值,选取这些平均值的中位数,所述中位数即为阈值。 The identification method based on the road edge mobile platform according to claim 1, characterized in that the threshold selection method of step S32 comprises: calculate the geometric mean of each domain area S in, select those median average, the threshold value is the median.
  3. 3. 根据权利要求2所述的基于移动平台进行道路边缘识别的方法,其特征在于,所述步骤S44具体包括:通过对斜率和位移的分析将碎线段进行连接或删除,通过分析前几帧图像检测出的道路边缘来在图像序列中的连续性来进行直线筛选。 3. The method of identification based on the road edge mobile platform of claim 2, wherein said step S44 specifically comprises: the analysis of the displacement and slope of the broken line segment connecting or deleted, by several frames prior to analysis road edge image detected by the continuity of the sequence of images is linearly screening.
CN 201310351847 2013-08-14 2013-08-14 A method and apparatus for identifying the road edge based on the mobile platform CN103400150B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201310351847 CN103400150B (en) 2013-08-14 2013-08-14 A method and apparatus for identifying the road edge based on the mobile platform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201310351847 CN103400150B (en) 2013-08-14 2013-08-14 A method and apparatus for identifying the road edge based on the mobile platform

Publications (2)

Publication Number Publication Date
CN103400150A true CN103400150A (en) 2013-11-20
CN103400150B true CN103400150B (en) 2017-07-07

Family

ID=49563768

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201310351847 CN103400150B (en) 2013-08-14 2013-08-14 A method and apparatus for identifying the road edge based on the mobile platform

Country Status (1)

Country Link
CN (1) CN103400150B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103714538B (en) * 2013-12-20 2016-12-28 中联重科股份有限公司 Road edge detection method, apparatus and vehicle
CN104680519B (en) * 2015-02-06 2017-06-23 四川长虹电器股份有限公司 Based puzzle contour and color identification method
CN105549603A (en) * 2015-12-07 2016-05-04 北京航空航天大学 Intelligent road tour inspection control method for multi-rotor-wing unmanned aerial vehicle

Citations (1)

* 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

Patent Citations (1)

* 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

Also Published As

Publication number Publication date Type
CN103400150A (en) 2013-11-20 application

Similar Documents

Publication Publication Date Title
US8498444B2 (en) Blob representation in video processing
US20110164816A1 (en) Image processing apparatus, image processing method, and storage medium
US20100080485A1 (en) Depth-Based Image Enhancement
US20070058863A1 (en) Label Detection
Chung et al. Demosaicing of color filter array captured images using gradient edge detection masks and adaptive heterogeneity-projection
CN101334837A (en) Multi-method integrated license plate image positioning method
CN101334836A (en) License plate positioning method incorporating color, size and texture characteristic
CN102006425A (en) Method for splicing video in real time based on multiple cameras
CN101246551A (en) Fast license plate locating method
US20050196044A1 (en) Method of extracting candidate human region within image, system for extracting candidate human region, program for extracting candidate human region, method of discerning top and bottom of human image, system for discerning top and bottom, and program for discerning top and bottom
JP2000048211A (en) Movile object tracking device
CN102346015A (en) Method for measuring icing thickness of insulator of power transmission line based on video difference analysis
CN102682292A (en) Method based on monocular vision for detecting and roughly positioning edge of road
US20080232715A1 (en) Image processing apparatus
EP2466901A2 (en) Depth data upsampling
US20060153450A1 (en) Integrated image processor
CN104050450A (en) Vehicle license plate recognition method based on video
CN102426649A (en) Simple steel seal digital automatic identification method with high accuracy rate
CN101359365A (en) Iris positioning method based on Maximum between-Cluster Variance and gray scale information
CN102567979A (en) Vehicle-mounted infrared night vision system and multi-source images fusing method thereof
CN102147859A (en) Ship monitoring method
Wu et al. Lane-mark extraction for automobiles under complex conditions
US20100079453A1 (en) 3D Depth Generation by Vanishing Line Detection
CN102147920A (en) Shadow detection method for high-resolution remote sensing image
CN101079149A (en) Noise-possessing movement fuzzy image restoration method based on radial basis nerve network

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
GR01