WO2020088453A1 - 用于车道检测的图像处理方法及装置 - Google Patents

用于车道检测的图像处理方法及装置 Download PDF

Info

Publication number
WO2020088453A1
WO2020088453A1 PCT/CN2019/114014 CN2019114014W WO2020088453A1 WO 2020088453 A1 WO2020088453 A1 WO 2020088453A1 CN 2019114014 W CN2019114014 W CN 2019114014W WO 2020088453 A1 WO2020088453 A1 WO 2020088453A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
area
sub
region
threshold
Prior art date
Application number
PCT/CN2019/114014
Other languages
English (en)
French (fr)
Inventor
宫原俊二
Original Assignee
长城汽车股份有限公司
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
Application filed by 长城汽车股份有限公司 filed Critical 长城汽车股份有限公司
Priority to EP19879551.0A priority Critical patent/EP3872690A4/en
Publication of WO2020088453A1 publication Critical patent/WO2020088453A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Definitions

  • the invention relates to the field of intelligent transportation and image processing, in particular to an image processing method and device for lane detection.
  • AD Autonomous Driving
  • ADAS Advanced Driver Assistance System
  • lane detection can be performed by supporting AD / ADAS through radar, visual camera (hereinafter referred to as camera), lidar, ultrasonic sensor, etc.
  • the camera is most widely used because it can obtain the same two-dimensional image information as human vision .
  • the captured image has various information about the detection object and the environment, and there is a large amount of this information, and not all are related to lane detection . Therefore, in the lane detection process, image processing of the image captured by the camera should be considered to extract the required features.
  • FIG. 1 is a schematic flowchart of a conventional lane detection. As shown in Figure 1, it usually includes the following image processing steps: 1) image acquisition; 2) color selection; 3) analysis area determination; 4) differential image or sub-sampled image; 5) smoothing, edge detection, peripheral processing image; 6 ) Set a threshold to get a ternary image; 7) Extract features from the ternary image, including edge processing, Hough transform, lane estimation, etc.
  • a single threshold is usually set to classify the difference image by pixel. For example, if the gray value of a point in the image is greater than the set threshold, the point is set to white, otherwise it is set to black, Thus, the corresponding ternary image is obtained.
  • this threshold segmentation method can get a ternary image, in practice, the lane is often composed of dotted lines and solid lines. If a lower threshold is used, the edges caused by the dotted lines can be detected, but the noise may be picked up a lot If you use a higher threshold, you may miss the edge caused by the dotted line, that is, lose important feature points.
  • the present invention aims to propose an image processing method and device for lane detection to at least partially solve the above technical problems.
  • An image processing method for lane detection includes: acquiring a road image captured by a camera on a vehicle; determining a rectangular area covered with a lane in the road image; performing differential processing on the rectangular area to obtain a corresponding differential image ; Divide the differential image into multiple sub-regions that individually include the solid or dotted lines of the lane; determine a threshold that is suitable for the pixels of each sub-region; and process the differential image of the corresponding sub-region based on the determined threshold To obtain the corresponding ternary image or binary image.
  • dividing the difference image into a plurality of sub-regions including solid lines or dotted lines of lanes separately includes dividing the difference image into a left sub-region and a right sub-region based on the center line of the rectangular region, and Let the left sub-region and the right sub-region include a dotted line or a solid line of the lane, respectively.
  • the determination of the threshold value adapted to the pixels of each sub-region includes: performing pixel statistical analysis on the difference image of each sub-region to obtain the statistical characteristics of each sub-region, and determining the The threshold of the pixels corresponding to the sub-region.
  • the image processing method further includes: determining an analysis area and a statistical area of the road image, wherein the analysis area is used to perform acquisition of the The area of the operation of the area of the ternary image or the binary image, the statistical area is an area for performing the operation of determining the threshold by statistical analysis, and the analysis area is covered by the rectangular area.
  • the statistical area is covered by the rectangular area, and the area of the statistical area is smaller than the rectangular area.
  • the image processing method for lane detection according to the present invention includes a multi-region and multi-threshold method for the road image Different thresholds are set in different areas of lines or dotted lines to obtain corresponding ternary images or binary images, thus solving the problem of difficult to pick up solid lines and dotted lines by a single threshold in traditional lane detection, and avoiding important feature points Lost, and helps to suppress noise points.
  • Another object of the present invention is to propose an image processing device for lane detection to at least partially solve the above technical problems.
  • An image processing device for lane detection includes: an image acquisition module for acquiring a road image captured by a camera on a vehicle; an area division module for determining a rectangular area covered with a lane in the road image; differential processing A module for performing differential processing on the rectangular area to obtain a corresponding differential image; a sub-region determining module for dividing the differential image into a plurality of sub-regions including solid lines or dotted lines of lanes separately; a threshold determining module , Used to determine a threshold suitable for the pixels of each sub-region; and a threshold processing module, used to process the differential image of the corresponding sub-region based on the determined threshold to obtain the corresponding ternary image or binary image.
  • the sub-region determining module is used to divide the difference image into a plurality of sub-regions including solid lines or dotted lines of lanes separately: dividing the difference image into left based on the center line of the rectangular region
  • the sub-region and the right sub-region, and the left sub-region and the right sub-region respectively include a dotted line or a solid line of a lane.
  • the threshold determination module is used to determine a threshold that is adapted to the pixels of each sub-region includes: performing pixel statistical analysis on the difference image of each sub-region to obtain the statistical characteristics of each sub-region, based on The statistical characteristics determine the threshold value that is adapted to the pixels of the corresponding sub-region.
  • the area dividing module is further used to: determine an analysis area and a statistical area of the road image, wherein the analysis area is an area for performing an operation of acquiring an area of the ternary image or binary image,
  • the statistical area is an area for performing an operation of determining a threshold by statistical analysis, and the analysis area is covered by the rectangular area.
  • the statistical area is covered by the rectangular area, and the area of the statistical area is smaller than the rectangular area.
  • Another aspect of the present invention also provides a machine-readable storage medium having instructions stored on the machine-readable storage medium. The instructions are used to cause the controller to execute the image processing method for lane detection described above.
  • the image processing device for lane detection and the machine-readable storage medium have the same advantages as the image processing method for lane detection described above relative to the prior art, and will not be repeated here.
  • Figure 1 is a schematic diagram of a conventional lane detection process
  • FIG. 2 is a schematic flowchart of an image processing method for lane detection according to an embodiment of the present invention
  • Figure 3 (a) is a road image in the example of the present invention.
  • Figure 3 (c) is a differential image filtered by the Sobel filter in the example of the present invention.
  • FIG. 3 (d) is a ternary image obtained by using a single threshold to process differential images in the example of the present invention
  • FIG. 3 (e) is a ternary image obtained by processing a differential image with a double threshold in the example of the present invention
  • FIG. 4 (b) is a schematic diagram of the difference image based on the horizontal line corresponding to FIG. 4 (a);
  • FIG. 5 is a schematic diagram of an exemplary neutron region determination scheme of the present invention.
  • 6 (a) and 6 (b) are schematic diagrams of lanes extracted using a single threshold and a double threshold in an embodiment of the present invention
  • 7 (a) and 7 (b) are the lane estimation results finally obtained by using the single threshold and the double threshold in the embodiment of the present invention.
  • FIG. 8 is a schematic structural diagram of an image processing device for lane detection according to an embodiment of the present invention.
  • Image acquisition module 120 Area division module
  • Threshold value determination module 160 Threshold value processing module
  • the image processing method may include the following steps:
  • Step S210 Acquire a road image captured by a camera on the vehicle.
  • the road image refers to an image captured by the camera and having a lane in front of the vehicle.
  • Step S220 Determine a rectangular area covered with a lane in the road image.
  • the rectangular area refers to an arbitrary rectangular simple area covered with a lane in the road image
  • FIG. 3 (a) is a road image in the example of the present invention, where Rect_area represents a rectangular area.
  • steps 2) to 6) shown in FIG. 1 can be applied to this rectangular area.
  • the rectangular area in the embodiment of the present invention is schematic, and other shapes of areas similar to the definition thereof should also be included in the scope of the embodiment of the present invention.
  • the image processing method may further include: determining an analysis area and a statistical area of the road image.
  • the analysis area is an area for performing an operation of acquiring an area of the ternary image or a binary image
  • the statistical area is an area for performing an operation of determining a threshold through statistical analysis
  • the statistical area is A subset of the rectangular area and covers the analysis area.
  • FIG. 3 (b) is a schematic diagram of area division in an example of the present invention, where Rect_area represents a rectangular area, Analysis_area represents an analysis area, and Statistical_area represents a statistical area.
  • the setting of the analysis area can reduce the difference processing and threshold processing involved in the subsequent steps from the rectangular area to the analysis area, which can not only ensure accurate lane information, but also help reduce the amount of image processing data.
  • the analysis area can be selected by the relative values of the camera image in the vertical and horizontal directions, more specifically, the ratio of the analysis area in the vertical and horizontal directions and the camera specifications. Many methods for determining the analysis area are disclosed in the prior art, so this part of the content is conventional for those skilled in the art, and will not be repeated here.
  • the analysis area may be executed, for example: a) Since vehicles and road-side features (such as walls, electronic signal poles, etc.) will bring noisy signals, delete the noisy area on top of the rectangular area in the analysis area; b) Provide proper edge distribution on the road; c) Provide simple distribution calculation.
  • vehicles and road-side features such as walls, electronic signal poles, etc.
  • the setting of the statistical area is to prepare for threshold determination in subsequent steps, which belongs to a rectangular area that is relatively easy to perform statistical analysis, so that the threshold can be determined through statistical analysis.
  • the statistical area is a subset of the rectangular area, that is, the statistical area Statistical_area is reduced from the rectangular area Rect_area to avoid a peripheral area that tends to have noise / unnecessary areas.
  • Step S230 Perform differential processing on the rectangular area to obtain a corresponding differential image.
  • a Sobel filter may be used to perform differential filtering on the rectangular area to obtain a corresponding differential image.
  • One typical Sobel filter is shown in equation (1) below.
  • 3 (c) is a differential image filtered by the Sobel filter in the example of the present invention, in which positive and negative edges of the lane are created, in which the brightness value of the right pixel is higher than that of the left pixel The brightness value is high, and in the negative edge, the brightness value of the left pixel is higher than that of the right pixel.
  • the edge is a tiny area where the brightness (or grayscale) of the image is abrupt or discontinuous, that is, the boundary of two areas with relatively different brightness values. Therefore, if the original image changes from dark to bright, the differential image is positive (for example, expressed as a positive number in formula (1)), and the corresponding boundary line is a positive edge.
  • the differential image is negative (for example Expressing the negative number in formula (1))
  • the corresponding boundary line is a negative edge
  • the brightness change of the original image and the positive and negative of the corresponding differential image can be determined by setting a threshold.
  • the following uses a specific lane detection as an example. In conventional lane detection, the difference image as shown in FIG. 3 (c) is divided by a single threshold. For example, if the threshold is threshold, ddd (m, n) and ttt (m, n) are the difference image and the ternary image, respectively, then
  • the lane is composed of solid lines and dotted lines. If a low threshold is used in equation (2), edges caused by dotted lines can be detected, but noise may be picked up in a large amount, and if high thresholds are used, edges caused by dotted lines may be missed. The reason for this phenomenon is explained below with reference to FIGS. 4 (a) and 4 (b), where FIG.
  • FIG. 4 (a) is the original image with a single threshold horizontal line in the example of the present invention, where line A and line B represent two Different horizontal lines, for example, pixels in the horizontal line can be expressed by aaa (m, n), and the horizontal line can correspond to ⁇ aaa (m, 1), aaa (m, 2) ... aaa (m, NN) ⁇ , where NN Is the endpoint of the horizontal axis, and line A and line B represent two horizontal lines with different m values, where aaa represents the original image.
  • FIG. 4 (b) is a schematic diagram of the difference image based on the horizontal line corresponding to FIG.
  • the solid line (refer to the right half of Figure 4 (b)) may have many difference points, some of which may have large difference values;
  • the solid line (refer to the right half of FIG. 4 (b)) may generally have stronger edges than the dotted line (refer to the left half of FIG. 4 (b)).
  • the embodiment of the present invention is different from the conventional single-threshold scheme, which inherits from the above steps S210 to S230, and improves the subsequent threshold processing scheme.
  • Step S240 Divide the difference image into a plurality of sub-regions that individually include solid lines or dotted lines of the lane.
  • the difference image may be divided into a left sub-region and a right sub-region based on the center line of the rectangular region, and the left sub-region and the right sub-region include lane points respectively Line or solid line.
  • FIG. 5 shows this preferred sub-region determination scheme.
  • the rectangular region Rect_area is correspondingly divided into a left sub-region Rect_area_left including a dotted line and a right sub-region Rect_area_right including a solid line, and a statistical region Statistical_area It is correspondingly divided into a left sub-region Statistical_area_left including a dotted line and a right sub-region Statistical_area_right including a solid line.
  • the left and right sides of the image can be processed independently, that is, the solid line and the dotted line can be processed independently.
  • step S250 a threshold value adapted to the pixels of each sub-region is determined.
  • determining the threshold suitable for the pixels of each sub-region includes: performing pixel statistical analysis on the difference image of each sub-region to calculate the threshold suitable for the pixels of the sub-region .
  • the pixel statistical analysis of the statistical area Statistical_area in each sub-region can be performed to obtain the statistical characteristics of the corresponding statistical area, and the most appropriate threshold is finally determined based on the statistical characteristics.
  • the embodiment of the present invention provides a method for automatically determining a threshold based on statistical characteristics, including using any one of the following to automatically determine a threshold: a) Obtaining the cumulative distribution of the difference image of the corresponding area through statistical analysis, and then for example, the cumulative value 98% is the threshold to be determined; b) The number of edges in the ternary image in the horizontal line is obtained through statistical analysis, for example, the number of edges is 5 or 6, and the threshold point to be determined corresponds to the ternary or binary image of the horizontal line The number of edges in is 5 or 6.
  • step S260 the difference image corresponding to the sub-region is processed based on the determined threshold to obtain the corresponding ternary image. That is, for each sub-region, a threshold value adapted to obtain a ternary image is used, instead of using a unified single threshold to obtain a ternary image.
  • FIG. 3 (e) is a ternary image obtained by using a double threshold processing difference image in the example of the present invention.
  • the image processing method of the embodiment of the present invention can make the dotted line stronger than the solid line, which is beneficial to subsequent image processing such as feature point extraction on the ternary image.
  • the ternary image may be further subjected to feature extraction, edge processing, Hough transform, etc. to complete the lane estimation.
  • the embodiment of the present invention takes the double threshold value as an example.
  • FIG. 6 (a) and FIG. 6 (b) are schematic diagrams of lanes extracted by using single threshold and double threshold respectively. The effect has been improved.
  • 7 (a) and 7 (b) are the results of the lane estimation finally obtained by using single threshold and double threshold respectively. It can be seen that the range of the estimated solid line and dotted line is longer in the double threshold than in the single threshold.
  • the image processing method for lane detection sets different thresholds for different areas including solid lines or dotted lines in a road image to obtain a corresponding ternary image through a multi-region and multi-threshold method Or binary image, which solves the problem of difficult to pick up solid lines and dotted lines through a single threshold in the detection of passing lanes, avoids the loss of important feature points, and helps to suppress noise points.
  • an embodiment of the present invention also provides an image processing device for lane detection.
  • 8 is a schematic structural diagram of an image processing device according to an embodiment of the present invention.
  • the image processing apparatus may include: an image acquisition module 110 for acquiring a road image captured by a camera on a vehicle; and an area division module 120 for determining a rectangular area covered with a lane in the road image
  • the difference processing module 130 is used to perform difference processing on the rectangular area to obtain a corresponding difference image
  • the sub-region determination module 140 is used to divide the difference image into a plurality of sub-lines including solid lines or dotted lines of lanes Area; threshold determination module 150, used to determine the threshold value adapted to the pixels of each sub-area; and threshold processing module 160, used to process the differential image of the corresponding sub-area based on the determined threshold to obtain the corresponding three Meta image or binary image.
  • the differential processing module 130 includes a Sobel filter that performs differential filtering on the rectangular area to obtain a corresponding differential image.
  • the sub-region determination module 140 is configured to divide the difference image into a plurality of sub-regions including solid lines or dotted lines of lanes separately: based on the center line of the rectangular region, The difference image is divided into a left sub-region and a right sub-region, and the left sub-region and the right sub-region respectively include a dotted line or a solid line of a lane.
  • the threshold determination module 150 is used to determine a threshold that is suitable for the pixels of each sub-region includes: performing pixel statistical analysis on the difference image of each sub-region to obtain the Statistical characteristics, based on which the threshold value adapted to the pixels of the corresponding sub-region is determined.
  • the area dividing module 120 is further used to determine an analysis area and a statistical area of the road image, wherein the analysis area is an area for performing acquisition of the ternary image or binary image The area of the operation, the statistical area is an area for performing an operation of determining a threshold by statistical analysis, and the analysis area is covered by the rectangular area.
  • the statistical area is covered by the rectangular area, and the area of the statistical area is smaller than the rectangular area.
  • image processing device has the same or similar implementation details and effects as the image processing method embodiment described above, so it will not be repeated here.
  • the processing of the ternary image in the embodiments of the present invention is adapted to the binary image.
  • the embodiment of the present invention mainly uses the ternary image as an example, and the processing of the binary image will not be described in detail.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code .

Abstract

一种用于车道检测的图像处理方法及装置。该方法包括：获取车辆上的摄像头捕获的道路图像（S210）；确定道路图像中覆盖有车道的矩形区域（S220）；对矩形区域进行差分处理以获取对应的差分图像（S230）；将差分图像划分为单独包括车道的实线或点线的多个子区域（S240）；确定与每一子区域的像素相适配的阈值（S250）；以及基于所确定的阈值处理对应子区域的差分图像以获取相对应的三元图像（S260）。该方法通过多区域多阈值的方式，为包括实线或点线的不同区域确定不同的阈值，再基于确定的阈值获取对应的三元图像，从而解决了传递车道检测中难以通过单个阈值拾取实线和点线的问题，并有利于避免丢失重要特征点以及抑制噪声点。

Description

用于车道检测的图像处理方法及装置 技术领域
本发明涉及智能交通及图像处理领域，特别涉及一种用于车道检测的图像处理方法及装置。
背景技术
目前，具有AD(Autonomous driving，自主驾驶)功能或ADAS(Advanced Driver Assistance System，高级驾驶辅助系统)的车辆已开始逐步推向市场，极大地促进了智能交通的发展。对于AD/ADAS，车道检测至关重要，其是AD/ADAS进行判断的重要条件。
现有技术中，可通过雷达、视觉摄像头(以下简称摄像头)、激光雷达、超声波传感器等支持AD/ADAS来进行车道检测，其中摄像头因能够获得与人类视觉一样的二维图像信息而应用最为广泛。但是，在利用摄像头进行车道检测时，虽然可以成功捕获关于车道的图像，但其捕获的图像中具有关于检测对象和环境的各种信息，而这些信息数量很多，且并不是都与车道检测相关。因此，在车道检测过程中，应考虑对摄像头捕获的图像进行图像处理以提取出需要的特征。
其中，图1是常规的车道检测的流程示意图。如图1所示，通常包括以下图像处理步骤：1)图像获取；2)颜色选择；3)分析区域确定；4)差分图像或子采样图像；5)平滑、边缘检测、外围处理图像；6)设置阈值得到三元图像；7)从三元图像中提取特征，包括边缘处理、霍夫(Hough)变换、车道估计等。
其中，在步骤6)中，通常是设置单个阈值来把差分图像按像素点进行分类，例如若图像中某点灰度值大于设定阈值，是把该点设置为白色，反之设置为黑色，从而获得了对应的三元图像。这种阈值分割方法虽然能得到三元图像，但在实际中，车道往往由点线和实线共同构成，如果使用较低阈值，可检测到由点线引起的边缘，但是噪声可能被大量拾取，而如果使用较高的阈值，则又有可能错过由点线引起的边缘，即丢失重要特征点。
到目前为止，现有技术中没有有效的方法来避免重要特征点的丢 失以及抑制图像中的大量噪声点。
发明内容
有鉴于此，本发明旨在提出一种用于车道检测的图像处理方法及装置，以至少部分地解决上述技术问题。
为达到上述目的，本发明的技术方案是这样实现的：
一种用于车道检测的图像处理方法，包括：获取车辆上的摄像头捕获的道路图像；确定所述道路图像中覆盖有车道的矩形区域；对所述矩形区域进行差分处理以获取对应的差分图像；将所述差分图像划分为单独包括车道的实线或点线的多个子区域；确定与每一子区域的像素相适配的阈值；以及基于所确定的阈值来处理对应子区域的差分图像以获取相对应的三元图像或二值图像。
进一步的，将所述差分图像划分为单独包括车道的实线或点线的多个子区域包括：基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。
进一步的，所述确定与每一子区域的像素相适配的阈值包括：在每一子区域的差分图像上进行像素统计分析以得到每一子区域的统计特性，基于所述统计特性确定与对应子区域的像素相适配的阈值。
进一步的，在确定所述道路图像中覆盖有车道的矩形区域之后，所述图像处理方法还包括：确定所述道路图像的分析区域和统计区域，其中所述分析区域是用于执行获取所述三元图像或二值图像的区域的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述分析区域被所述矩形区域覆盖。
进一步的，所述统计区域被所述矩形区域覆盖，且所述统计区域的面积小于所述矩形区域。
相对于现有技术，本发明所述的用于车道检测的图像处理方法具有以下优势：本发明所述的用于车道检测的图像处理方法通过多区域多阈值的方式，为道路图像中包括实线或点线的不同区域设置不同的阈值以获得对应的三元图像或二值图像，从而解决了传统车道检测中难以通过单个阈值拾取实线和点线的问题，且避免了重要特征点的丢 失，并有利于抑制噪声点。
本发明的另一目的在于提出一种用于车道检测的图像处理装置，以至少部分地解决上述技术问题。
为达到上述目的，本发明的技术方案是这样实现的：
一种用于车道检测的图像处理装置，包括：图像获取模块，用于获取车辆上的摄像头捕获的道路图像；区域划分模块，用于确定所述道路图像中覆盖有车道的矩形区域；差分处理模块，用于对所述矩形区域进行差分处理以获取对应的差分图像；子区域确定模块，用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域；阈值确定模块，用于确定与每一子区域的像素相适配的阈值；以及阈值处理模块，用于基于所确定的阈值来处理对应子区域的差分图像以获取相对应的三元图像或二值图像。
进一步的，所述子区域确定模块用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域包括：基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。
进一步的，所述阈值确定模块用于确定与每一子区域的像素相适配的阈值包括：在每一子区域的差分图像上进行像素统计分析以得到每一子区域的统计特性，基于所述统计特性确定与对应子区域的像素相适配的阈值。
进一步的，所述区域划分模块还用于：确定所述道路图像的分析区域和统计区域，其中所述分析区域是用于执行获取所述三元图像或二值图像的区域的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述分析区域被所述矩形区域覆盖。
进一步的，所述统计区域被所述矩形区域覆盖，且所述统计区域的面积小于所述矩形区域。
本发明另一方面还提供一种机器可读存储介质，该机器可读存储介质上存储有指令，该指令用于使得控制器执行上述的用于车道检测的图像处理方法。
所述用于车道检测的图像处理装置及所述机器可读存储介质与 上述用于车道检测的图像处理方法相对于现有技术所具有的优势相同，在此不再赘述。
本发明的其它特征和优点将在随后的具体实施方式部分予以详细说明。
附图说明
构成本发明的一部分的附图用来提供对本发明的进一步理解，本发明的示意性实施方式及其说明用于解释本发明，并不构成对本发明的不当限定。在附图中：
图1是常规的车道检测的流程示意图；
图2是本发明实施例的用于车道检测的图像处理方法的流程示意图；
图3(a)是本发明示例中的道路图像；
图3(b)是本发明示例中的区域划分示意图；
图3(c)是本发明示例中经Sobel滤波器滤波后的差分图像；
图3(d)是本发明示例中采用单阈值处理差分图像得到的三元图像；
图3(e)是本发明示例中采用双阈值处理差分图像得到的三元图像；
图4(a)为本发明示例中具有单阈值水平线的原始图像；
图4(b)为图4(a)所对应的基于水平线的差分图像示意图；
图5为本发明示例中子区域确定方案的示意图；
图6(a)及图6(b)分别为本发明实施例中采用单阈值和双阈值所提取的车道的示意图；
图7(a)及图7(b)分别为本发明实施例中采用单阈值和双阈值所最终获得的车道估计结果；以及
图8是本发明实施例所述的用于车道检测的图像处理装置的结构示意图。
附图标记说明：
110、图像获取模块        120、区域划分模块
130、差分处理模块     140、子区域确定模块
150、阈值确定模块     160、阈值处理模块
具体实施方式
需要说明的是，在不冲突的情况下，本发明中的实施方式及实施方式中的特征可以相互组合。
下面将参考附图并结合实施方式来详细说明本发明。
图2是本发明实施例的用于车道检测的图像处理方法的流程示意图。如图2所示，该图像处理方法可以包括以下步骤：
步骤S210，获取车辆上的摄像头捕获的道路图像。
其中，该道路图像是指所述摄像头所捕获的具有车辆前方的车道的图像。
步骤S220，确定所述道路图像中覆盖有车道的矩形区域。
其中，所述矩形区域是指道路图像中覆盖有车道的任意矩形状的简单区域，图3(a)是本发明示例中的道路图像，其中Rect_area表示矩形区域。另外，图1中所示出的步骤2)至步骤6)可应用于此矩形区域。需说明的是，本发明实施例中的矩形区域是示意性的，与之定义类似的其他形状的区域也应包括在本发明实施例的范围内。
在优选的实施例中，在确定所述矩形区域之后，所述图像处理方法还可以包括：确定所述道路图像的分析区域和统计区域。其中所述分析区域是用于执行获取所述三元图像或二值图像的区域的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述统计区域是所述矩形区域的子集，并覆盖分析区域。图3(b)是本发明示例中的区域划分示意图，其中Rect_area表示矩形区域，Analysis_area表示分析区域，Statistical_area表示统计区域。
在此，分析区域的设定可使后续步骤中所涉及的差分处理、阈值处理等从矩形区域减少至分析区域，既能保证得到准确的车道信息，又有利于减少图像处理的数据量。一般地，分析区域可通过摄像头图像垂直和水平方向上的相对值来进行选择，更为具体的是通过分析区域在垂直和水平方向的比率以及摄像头规格来确定。现有技术中公开了很多确定分析区域的方法，因此这部分内容对于本领域技术人员是 常规的，在此不再赘述。本发明实施例中，分析区域例如可执行：a)由于车辆和道路侧特体(如墙壁、电子信号杆等)等会带来嘈杂信号，则在分析区域中删除矩形区域顶部的嘈杂区域；b)在道路上提供适当的边缘分布；c)可提供简单的分布计算。
另外，统计区域的设定是为后续步骤中进行阈值确定做准备，其属于相对易于进行统计分析的矩形区域，从而可通过统计分析确定阈值。
需说明的是所述统计区域是所述矩形区域的子集，即是使统计区域Statistical_area从矩形区域Rect_area减少，以避免倾向于具有噪声/不必要区域的外围区域。
步骤S230，对所述矩形区域进行差分处理以获取对应的差分图像。
在优选的实施例中，可采用Sobel滤波器对所述矩形区域进行差分滤波处理，以得到对应的差分图像。其中一个典型的Sobel滤波器如下述的方程(1)所示。
Figure PCTCN2019114014-appb-000001
其中，图3(c)是本发明示例中经Sobel滤波器滤波后的差分图像，其中创建了车道的正边缘和负边缘，在所述正边缘中，其右像素的亮度值比左像素的亮度值高，在所述负边缘中，其左像素的亮度值比右像素高。其中，边缘是图像中亮度(或灰度)发生突变或不连续的微小区域，即是两个具有相对不同亮度值特征的区域的边界线。因此，若原始图像由暗变亮，则差分图像为正(例如表达为公式(1)的正数)，对应边界线为正边缘，若原始图像由亮变暗，则差分图像为负(例如表达公式(1)中的负数)，对应边界线为负边缘，而原始图像的亮度未发生变化时，差分图像可以为零。需说明的是，现有技术中关于边缘检测技术的介绍较多，本领域技术人员可参考相关现有技术以理解正边缘和负边缘。进一步地，原始图像的亮度变化与对应的差分图像的正负可通过阈值设定来确定，下面以具体的车道检测为例。在常规车道检测中，通过单阈值来分割如图3(c)所示的差分图像。举例而言，设阈值为threshold，ddd(m，n)和ttt(m，n) 分别是差分图像和三元图像，则
ddd(m,n)<-threshold  ttt(m,n)＝-1，对应三元图像中的负边缘；
ddd(m,n)>+threshold  ttt(m,n)＝1，对应三元图像中的正边缘；
否则，ttt(m,n)＝0      (2)
其中，例如当ttt(m，n)＝-1时对应为黑色(通过黑色表示“暗”)，当ttt(m，n)＝1时对应为白色(通过白色表示“亮”)，则得到的三元图像如3(d)所示。
可知，最终得到的三元图像显著地取决于阈值。但是，再次参考图3(a)，可知车道是由实线和点线组成的。如果在等式(2)中使用低阈值，则可以检测到由点线引起的边缘，但是噪声可能被大量拾取，而如果使用高阈值，则可能错过由点线引起的边缘。下面通过图4(a)及图4(b)来说明造成这种现象的原因，其中图4(a)为本发明示例中具有单阈值水平线的原始图像，其中line A和line B表示两条不同的水平线，例如水平线中的像素可通过aaa(m，n)表达，而水平线可对应表示{aaa(m，1)、aaa(m，2)……aaa(m,NN)}，其中NN为水平轴的端点，而line A和line B则表示m值不相同的两条水平线，其中aaa表示原始图像。图4(b)为图4(a)所对应的基于水平线的差分图像示意图，其中line A和line B分别都具有高阈值high_threshold和低阈值low_threshold，举例而言，结合上述公式(2)，对于正边缘，如果ddd(m，n)大于阈值，则三元图像ttt(m，n)将为“1”，否则为“0”。即，如果选择高阈值，则三元图像具有少量的“1”点，从而有较少数量的三元图像点。因此，从图4(b)可以看出造成上述的“使用低阈值，则可以检测由点线引起的边缘，但是噪声可能被大量拾取，而如果使用高阈值，则可能错过由点线引起的边缘”的现象的原因主要有两点：
1)实线(参考图4(b)的右半部分)可能有许多差分点，其中一些可能有很大的差分值；
2)实线(参考图4(b)的右半部分)通常可以具有比点线(参 考图4(b)的左半部分)更强的边缘。
据此，本发明实施例不同于常规的单阈值方案，其承接于上述步骤S210-步骤S230，改进了后续的阈值处理方案。
步骤S240，将所述差分图像划分为单独包括车道的实线或点线的多个子区域。
在优选的实施例中，可基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。其中，图5示出了这种优选的子区域确定方案，如图5所示，矩形区域Rect_area被对应划分为包括点线的左子区域Rect_area_left以及包括实线的右子区域Rect_area_right，统计区域Statistical_area被对应划分为包括点线的左子区域Statistical_area_left以及包括实线的右子区域Statistical_area_right。据此，可独立地处理图像的左侧和右侧，也即是可以独立地处理实线和点线。
需说明的是，在此涉及的基于中心线划分左、右子区域的方案是示例性的，在实践中，可根据实际情况划分为多个区域。
步骤S250，确定与每一子区域的像素相适配的阈值。
在优选的实施例中，确定与每一子区域的像素相适配的阈值包括：在每一子区域的差分图像上进行像素统计分析，以计算出与该子区域的像素相适配的阈值。
其中，在划分了统计区域Statistical_area的情况下，可对每一子区域中属于统计区域Statistical_area的部分进行像素统计分析以得到对应统计区域的统计特性，并基于统计特性来最终确定最为合适的阈值。其中，本发明实施例给出了基于统计特征自动确定阈值的方法，包括可以使用以下任意一者来自动确定阈值：a)通过统计分析获得对应区域的差分图像的累积分布，再例如将累积值的98％作为所要确定的阈值；b)通过统计分析获取水平线中三元图像中的边缘数，例如边缘数为5或6，而对应要确定的阈值点要保证水平线三元图像或二值图像中的边缘数为5或6。
步骤S260，基于所确定的阈值来处理对应子区域的差分图像以 获取相对应的三元图像。即，对每一子区域，采用与之相适配的阈值来获取三元图像，而不再在是采用统一的单阈值来获取三元图像。
其中，图3(e)是本发明示例中采用双阈值处理差分图像得到的三元图像，相比于图3(d)中采用单阈值得到的三元图像，可知点线得到增强，且点线的边缘点多于图3(d)。因此，采用本发明实施例的图像处理方法，可使得点线比实线更强，有利于后续对三元图像进行特征点提取等图像处理。
需说明的是，参考图3(e)，图像的左半部分有一些噪声点，这可能是阈值较低而形成的，可通过调整阈值来进行处理。
更进一步地，在经由步骤S210-步骤S260得到三元图像后，可进一步地对三元图像进行特征提取、边缘处理及Hough变换等以完成车道估计。本发明实施例以双阈值为例，图6(a)及图6(b)分别是采用单阈值和双阈值所提取的车道的示意图，可知采用双阈值后在点线的更远点的识别效果得到了改善。图7(a)及图7(b)分别是采用单阈值和双阈值所最终获得的车道估计结果，可知估计的实线和点线的范围在双阈值中比在单阈值中更长。
综上所述，本发明实施例的用于车道检测的图像处理方法通过多区域多阈值的方式，为道路图像中包括实线或点线的不同区域设置不同的阈值以获得对应的三元图像或二值图像，从而解决了传递车道检测中难以通过单个阈值拾取实线和点线的问题，避免了重要特征点的丢失，并有利于抑制噪声点。
基于与上述的图像处理方法相同的发明思路，本发明实施例还提供了一种用于车道检测的图像处理装置。图8是本发明实施例所述的图像处理装置的结构示意图。如图8所示，所述图像处理装置可以包括：图像获取模块110，用于获取车辆上的摄像头捕获的道路图像；区域划分模块120，用于确定所述道路图像中覆盖有车道的矩形区域；差分处理模块130，用于对所述矩形区域进行差分处理以获取对应的差分图像；子区域确定模块140，用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域；阈值确定模块150，用于确定与每一子区域的像素相适配的阈值；以及阈值处理模块160，用于基于 所确定的阈值来处理对应子区域的差分图像以获取相对应的三元图像或二值图像。
在优选的实施例中，所述差分处理模块130包括Sobel滤波器，该Sobel滤波器对所述矩形区域进行差分滤波处理，以得到对应的差分图像。
在优选的实施例中，所述子区域确定模块140用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域包括：基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。
在优选的实施例中，所述阈值确定模块150用于确定与每一子区域的像素相适配的阈值包括：在每一子区域的差分图像上进行像素统计分析以得到每一子区域的统计特性，基于所述统计特性确定与对应子区域的像素相适配的阈值。
在优选的实施例中，所述区域划分模块120还用于：确定所述道路图像的分析区域和统计区域，其中所述分析区域是用于执行获取所述三元图像或二值图像的区域的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述分析区域被所述矩形区域覆盖。
在更为优选的实施例中，所述统计区域被所述矩形区域覆盖，且所述统计区域的面积小于所述矩形区域。
需说明的是，本发明实施例所述的图像处理装置与上述关于图像处理方法的实施例的实施细节及效果相同或相似，故在此不再赘述。
还需说明的是，本发明实施例中对三元图像的处理均适应于二值图像，本发明实施例主要以三元图像为例，不再对二值图像的处理进行赘述。
以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，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，例如适应性改变步骤的执行顺序以及功能模块间的连接关系，均应包含在本发明的保护范围之内。
本领域技术人员可以理解实现上述实施例方法中的全部或部分 步骤是可以通过程序来指令相关的硬件来完成，该程序存储在一个存储介质中，包括若干指令用以使得单片机、芯片或处理器(processor)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
此外，本发明实施例的各种不同的实施例之间也可以进行任意组合，只要其不违背本发明实施例的思想，其同样应当视为本发明实施例所公开的内容。

Claims (13)

  1. 一种用于车道检测的图像处理方法，其特征在于，所述用于车道检测的图像处理方法包括：
    获取车辆上的摄像头捕获的道路图像；
    确定所述道路图像中覆盖有车道的矩形区域；
    对所述矩形区域进行差分处理以获取对应的差分图像；
    将所述差分图像划分为单独包括车道的实线或点线的多个子区域；
    确定与每一子区域的像素相适配的阈值；以及
    基于所确定的阈值来处理对应子区域的差分图像以获取相对应的三元图像或二值图像。
  2. 根据权利要求1所述的用于车道检测的图像处理方法，其特征在于，将所述差分图像划分为单独包括车道的实线或点线的多个子区域包括：
    基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。
  3. 根据权利要求1所述的用于车道检测的图像处理方法，其特征在于，所述确定与每一子区域的像素相适配的阈值包括：
    在每一子区域的差分图像上进行像素统计分析以得到每一子区域的统计特性，基于所述统计特性确定与对应子区域的像素相适配的阈值。
  4. 根据权利要求3所述的用于车道检测的图像处理方法，其特征在于，所述基于所述统计特性确定与对应子区域的像素相适配的阈值包括：
    通过统计分析获得对应子区域的差分图像的累积分布，再将设定比例的累积值作为所要确定的阈值；或者
    通过统计分析获取水平线中三元图像中的边缘数，并使对应要确定的阈值点能够保证水平线中三元图像或二值图像中的边缘数为所获取的边缘数。
  5. 根据权利要求1至4中的任意一项所述的用于车道检测的图像处理方法，其特征在于，在确定所述道路图像中覆盖有车道的矩形区域之后，所述图像处理方法还包括：
    确定所述道路图像的分析区域和统计区域，其中所述分析区域是用于执行获取所述三元图像或二值图像的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述分析区域被所述矩形区域覆盖。
  6. 根据权利要求5所述的用于车道检测的图像处理方法，其特 征在于，所述统计区域被所述矩形区域覆盖，且所述统计区域的面积小于所述矩形区域。
  7. 一种用于车道检测的图像处理装置，其特征在于，所述用于车道检测的图像处理装置包括：
    图像获取模块，用于获取车辆上的摄像头捕获的道路图像；
    区域划分模块，用于确定所述道路图像中覆盖有车道的矩形区域；
    差分处理模块，用于对所述矩形区域进行差分处理以获取对应的差分图像；
    子区域确定模块，用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域；
    阈值确定模块，用于确定与每一子区域的像素相适配的阈值；以及
    阈值处理模块，用于基于所确定的阈值来处理对应子区域的差分图像以获取相对应的三元图像或二值图像。
  8. 根据权利要求7所述的用于车道检测的图像处理装置，其特征在于，所述子区域确定模块用于将所述差分图像划分为单独包括车道的实线或点线的多个子区域包括：
    基于所述矩形区域的中心线，将所述差分图像划分为左子区域和右子区域，并使所述左子区域和所述右子区域分别包括车道的点线或实线。
  9. 根据权利要求7所述的用于车道检测的图像处理装置，其特征在于，所述阈值确定模块用于确定与每一子区域的像素相适配的阈值包括：
    在每一子区域的差分图像上进行像素统计分析以得到每一子区域的统计特性，基于所述统计特性确定与对应子区域的像素相适配的阈值。
  10. 根据权利要求9所述的用于车道检测的图像处理装置，其特征在于，所述阈值确定模块用于基于所述统计特性确定与对应子区域的像素相适配的阈值包括：
    通过统计分析获得对应子区域的差分图像的累积分布，再将设定比例的累积值作为所要确定的阈值；或者
    通过统计分析获取水平线中三元图像中的边缘数，并使对应要确定的阈值点能够保证水平线中三元图像或二值图像中的边缘数为所获取的边缘数。
  11. 根据权利要求7至10中的任意一项所述的用于车道检测的图像处理装置，其特征在于，所述区域划分模块还用于：
    确定所述道路图像的分析区域和统计区域，其中所述分析区域是用于执行获取所述三元图像或二值图像的区域的操作的区域，所述统计区域是用于执行通过统计分析确定阈值的操作的区域，且所述分析 区域被所述矩形区域覆盖。
  12. 根据权利要求11所述的用于车道检测的图像处理装置，其特征在于，所述统计区域被所述矩形区域覆盖，且所述统计区域的面积小于所述矩形区域。
  13. 一种机器可读存储介质，该机器可读存储介质上存储有指令，该指令用于使得控制器执行权利要求1至6中任意一项所述的用于车道检测的图像处理方法。
PCT/CN2019/114014 2018-10-29 2019-10-29 用于车道检测的图像处理方法及装置 WO2020088453A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP19879551.0A EP3872690A4 (en) 2018-10-29 2019-10-29 IMAGE PROCESSING METHOD AND APPARATUS USED FOR LANE DETECTION

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811271573.2 2018-10-29
CN201811271573.2A CN110490033B (zh) 2018-10-29 2018-10-29 用于车道检测的图像处理方法及装置

Publications (1)

Publication Number Publication Date
WO2020088453A1 true WO2020088453A1 (zh) 2020-05-07

Family

ID=68545597

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/114014 WO2020088453A1 (zh) 2018-10-29 2019-10-29 用于车道检测的图像处理方法及装置

Country Status (3)

Country Link
EP (1) EP3872690A4 (zh)
CN (1) CN110490033B (zh)
WO (1) WO2020088453A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104217427A (zh) * 2014-08-22 2014-12-17 南京邮电大学 一种交通监控视频中车道线定位方法
EP2879110A1 (en) * 2012-07-27 2015-06-03 Nissan Motor Co., Ltd. Three-dimensional object detection device, and three-dimensional object detection method
CN105261020A (zh) * 2015-10-16 2016-01-20 桂林电子科技大学 一种快速车道线检测方法
CN107301776A (zh) * 2016-10-09 2017-10-27 上海炬宏信息技术有限公司 基于视频检测技术的车道路况处理及发布方法
CN107590470A (zh) * 2017-09-18 2018-01-16 浙江大华技术股份有限公司 一种车道线检测方法及装置

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100403332C (zh) * 2006-11-02 2008-07-16 东南大学 用于车道偏离报警的车道线鲁棒识别方法
CN104102905B (zh) * 2014-07-16 2018-03-16 中电海康集团有限公司 一种车道线自适应的检测方法
EP3098753A1 (en) * 2015-05-28 2016-11-30 Tata Consultancy Services Limited Lane detection
CN106682646B (zh) * 2017-01-16 2020-12-22 北京新能源汽车股份有限公司 一种车道线的识别方法及装置
CN107462223B (zh) * 2017-08-08 2020-06-16 长安大学 一种公路转弯前行车视距自动测量装置及测量方法
CN107832674B (zh) * 2017-10-16 2021-07-09 西安电子科技大学 一种车道线检测方法
CN107862290B (zh) * 2017-11-10 2021-09-24 智车优行科技（北京）有限公司 车道线检测方法及系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2879110A1 (en) * 2012-07-27 2015-06-03 Nissan Motor Co., Ltd. Three-dimensional object detection device, and three-dimensional object detection method
CN104217427A (zh) * 2014-08-22 2014-12-17 南京邮电大学 一种交通监控视频中车道线定位方法
CN105261020A (zh) * 2015-10-16 2016-01-20 桂林电子科技大学 一种快速车道线检测方法
CN107301776A (zh) * 2016-10-09 2017-10-27 上海炬宏信息技术有限公司 基于视频检测技术的车道路况处理及发布方法
CN107590470A (zh) * 2017-09-18 2018-01-16 浙江大华技术股份有限公司 一种车道线检测方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3872690A4 *

Also Published As

Publication number Publication date
CN110490033A (zh) 2019-11-22
CN110490033B (zh) 2022-08-23
EP3872690A4 (en) 2022-08-10
EP3872690A1 (en) 2021-09-01

Similar Documents

Publication Publication Date Title
CN108549874B (zh) 一种目标检测方法、设备及计算机可读存储介质
CN108280450B (zh) 一种基于车道线的高速公路路面检测方法
US10037597B2 (en) Image inpainting system and method for using the same
WO2020107866A1 (zh) 一种文字区域获取方法、装置、存储介质及终端设备
WO2018145470A1 (zh) 一种图像检测方法和装置
EP3979196A1 (en) Image processing method and apparatus for target detection
CN104778669A (zh) 快速图像去噪方法及装置
CN108198208B (zh) 一种基于目标跟踪的移动侦测方法
CN113221861B (zh) 一种多车道线检测方法、装置及检测设备
CN106022268A (zh) 一种限速标识的识别方法和装置
CN111382722B (zh) 车牌图像优选方法、图像处理装置及具有存储功能的装置
WO2018068300A1 (zh) 图像处理方法和装置
CN116091421A (zh) 一种体外受精胚胎卵裂球图像自动分割及面积计算的方法
CN114127784A (zh) 用于生成相机流的掩模的方法、计算机程序产品和计算机可读介质
CN111833367A (zh) 一种图像处理方法、装置、车辆及存储介质
CN109800641B (zh) 基于阈值自适应二值化和连通域分析的车道线检测方法
CN111652033A (zh) 基于OpenCV的车道线检测方法
CN107977608B (zh) 一种应用于公路视频图像道路区域提取的方法
WO2019149213A1 (zh) 基于影像识别路锥的方法、装置、存储介质以及车辆
WO2019085929A1 (zh) 图像处理方法及其装置、安全驾驶方法
CN108090425B (zh) 一种车道线检测方法、装置及终端
CN113792583A (zh) 基于可行驶区域的障碍物检测方法、系统和智能终端
WO2020088453A1 (zh) 用于车道检测的图像处理方法及装置
CN115223031B (zh) 一种单目边框测距方法、装置、介质及幕墙机器人
US20200219280A1 (en) Attachable matter detection apparatus

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19879551

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019879551

Country of ref document: EP

Effective date: 20210528