WO2015024294A1 - 一种车辆遮阳板状态的检测方法 - Google Patents

一种车辆遮阳板状态的检测方法 Download PDF

Info

Publication number
WO2015024294A1
WO2015024294A1 PCT/CN2013/085951 CN2013085951W WO2015024294A1 WO 2015024294 A1 WO2015024294 A1 WO 2015024294A1 CN 2013085951 W CN2013085951 W CN 2013085951W WO 2015024294 A1 WO2015024294 A1 WO 2015024294A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
main connected
sun visor
edge
area
Prior art date
Application number
PCT/CN2013/085951
Other languages
English (en)
French (fr)
Inventor
李熙莹
陈玲
Original Assignee
中山大学
广东方纬科技有限公司
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 中山大学, 广东方纬科技有限公司 filed Critical 中山大学
Priority to US14/438,895 priority Critical patent/US9424476B2/en
Publication of WO2015024294A1 publication Critical patent/WO2015024294A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/0008—Industrial image inspection checking presence/absence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20036—Morphological image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30268—Vehicle interior

Definitions

  • the invention relates to the field of image processing, and in particular to a method for detecting a state of a vehicle sun visor.
  • the sun visor is mounted on the front windshield of the car to prevent glare from the sun and can be lowered by the driver or adjusted to a suitable down angle.
  • some lawless elements will put down the sun visor at night, using the visor of the sun visor to cover their facial features, preventing their facial features from being photographed by the bayonet system, thus avoiding legal sanctions.
  • timely feedback of the vehicle's suspiciousness, and effectively provide reference information for the public security it is necessary to quickly and effectively detect the state of the vehicle's sun visor.
  • the existing night sun visor state detection methods have the following problems: 1. Road texture has a certain impact on sun visor detection; Parts with similar shapes (such as rectangular objects placed in the car and labels attached to the car window) cause failure in detection as a sun visor; 3. Practicality is not high and it is difficult to apply to the actual bayonet system. These factors seriously affect the efficiency and detection accuracy of the visor state detection. Therefore, there is still no perfect and mature night visor state detection algorithm that can be applied in practical engineering.
  • the object of the present invention is to provide a method for detecting the state of a vehicle sun visor with high detection efficiency, high detection accuracy and high practicability, in order to meet the security warning requirements of the public security department.
  • the technical solution adopted by the present invention to solve the technical problem thereof is: a method for detecting a state of a vehicle sun visor, comprising:
  • step B includes:
  • the mathematical image method is used to remove the highlighted area of the stretched image, thereby obtaining a grayscale image.
  • step C includes:
  • C1 Thresholding the grayscale image by using a preset threshold to obtain a binary image
  • step C4 includes:
  • the geometric features of the main connected region include a centroid position, a center of gravity position, a width, a height, an aspect ratio, an area, an circumscribed rectangle, and an area of the circumscribed rectangle of the main connected region;
  • the geometric features of the equivalent rectangle include an equivalent The position of the rectangle and the area of the equivalent rectangle.
  • step of performing horizontal long edge extraction on the grayscale image in the step D includes:
  • the horizontal long edge is matched with the main connected area to obtain a region edge matching relationship, which is specifically:
  • Feature matching is performed on the horizontal long edge and the main connected area, and whether the upper edge and the lower edge matching the horizontal long edge exist in the main connected area, thereby obtaining the regional edge matching relationship.
  • step E includes:
  • step E1 determining whether there is a matching edge in the main connected area according to the area edge matching relationship, if not, executing step E2, otherwise, performing step E3;
  • step E2 according to the geometrical features of the main connected area and the horizontal long edge to determine whether the positional relationship of the main connected area is correct, and if so, step E3 is performed; otherwise, step E5 is performed;
  • step E3 determining whether the degree of similarity of the main connected area meets the preset threshold condition, and if so, executing step E4, otherwise, performing step E5;
  • E5. Determine the state of the vehicle sun visor so that the sun visor is not lowered.
  • the detected image is a window image of the vehicle.
  • the beneficial effects of the invention are: based on the mathematical morphology and the connected domain theory, comprehensively utilizing the geometric features of the main connected regions, the degree of similarity of the main connected regions and the matching relationship of the regions to detect the state of the sun visor, which can effectively reduce the road.
  • the influence of the texture and the interior environment on the detection of the shading state, the detection efficiency and the detection accuracy are high, and the detection accuracy rate can reach more than 90%; it can be applied in the actual bayonet system to effectively detect whether the night visor is in the down state. High practicality.
  • FIG. 1 is a flow chart showing the steps of a method for detecting a state of a sun visor of a vehicle according to the present invention
  • FIG. 2 is a flow chart of step B of the present invention.
  • FIG. 3 is a flow chart of step C of the present invention.
  • FIG. 4 is a flow chart of step C4 of the present invention.
  • step D of the present invention is a flow chart of performing horizontal long edge extraction on a grayscale image in step D of the present invention.
  • Figure 6 is a flow chart of step E of the present invention.
  • a method of vehicle sun visor’s state Detection a method for detecting the state of a vehicle sun visor
  • Grayscale stretching One of the basic image grayscale transformation methods.
  • the grayscale of the image is transformed by a simple transformation function to improve the dynamic range of the grayscale during image processing.
  • a method for detecting a state of a vehicle sun visor includes:
  • the detected image is a window image of the vehicle.
  • the geometrical features of the main connected area include the centroid position, center of gravity position, width, height, aspect ratio, area, circumscribed rectangle, and area of the circumscribed rectangle of the main connected area.
  • the area edge matching relationship indicates whether the main connected area has an edge that matches the horizontal long edge (ie, whether the position of the main connected area and the horizontal long edge are close).
  • the primary connected area is a candidate area for the vehicle sun visor.
  • the invention analyzes various image features of the night visor, first extracts the main connected area, horizontal long edge extraction and regional edge feature matching, and extracts effective features that can be used for visor state discrimination (including geometric features of the main connected area). , the degree of similarity of the rectangle and the edge matching relationship, etc., and then the visor state discrimination based on the extracted features is more efficient and accurate.
  • the step B includes:
  • the mathematical image method is used to remove the highlighted area of the stretched image, thereby obtaining a grayscale image.
  • r, g, and b are components of the red channel, components of the green channel, and components of the blue channel, respectively, and I is the gray level of the pixel.
  • the image brightness is distributed within a reasonable range.
  • the step C includes:
  • C1 Thresholding the grayscale image by using a preset threshold to obtain a binary image
  • the mathematical morphology operation processing on the binary image includes an open operation process and a vertical close operation process.
  • the open operation process refers to performing a mathematical morphology open operation on the binary image by using a structural element of appropriate length to remove the adhesion of the sun visor and the face in the binary image.
  • each area of the binary image is marked to obtain a mark image of the binary image; and then the mark image is subjected to a longitudinal closing operation to obtain a longitudinal closed operation image (ie, a morphological operation image). After the longitudinal closing operation, the longitudinal fracture portions in the same mark region are connected, and the problem of region merger does not occur.
  • the culling operation image processing means that the main connecting area is obtained by eliminating the area where the width, the height, the area, and the aspect ratio do not meet the requirements in the morphological operation image according to the prior knowledge of the shape of the visor.
  • the geometrical features of the main connected areas and the degree of similarity of the rectangles are used for subsequent visor state determination.
  • the step C4 includes:
  • the geometric features of the equivalent rectangle include the position of the equivalent rectangle and the area of the equivalent rectangle.
  • the circumscribed rectangle of the main connected area refers to the circumscribed rectangle of the outer contour of the main connected area.
  • the equivalent rectangle of the main connected area refers to a rectangle satisfying the following principles: 1) the center of gravity of the equivalent rectangle is the same as the center of gravity of the main connected area; 2) the difference between the area of the equivalent rectangle and the area of the main connected area is less than a preset threshold .
  • r 1 and r 2 There are two measures of the degree of similarity of the rectangles of the main connected areas: r 1 and r 2 .
  • the calculation formulas for r 1 and r 2 are: , In the formula, S is the area of the main connected region, S c represents the area of the circumscribed rectangle, S r represents the area remaining after removing the equivalent rectangle in the main connected region, and S e represents the area of the equivalent rectangle.
  • S is the area of the main connected region
  • S c represents the area of the circumscribed rectangle
  • S r represents the area remaining after removing the equivalent rectangle in the main connected region
  • S e represents the area of the equivalent rectangle.
  • the closer r 1 is to 1, the closer the main connected region is to the rectangle, and the smaller r 1 is , indicating that the main connected region contains noise points that cause the region to deviate significantly from the rectangle; the closer r 2 is to 0, the closer the main connected region is. In the rectangle.
  • the degree of similarity of the main connected regions can be obtained by calculating r 1 and r 2 according to the area of the main connected region, the area of the circumscribed rectangle, and the area of the equivalent rectangle.
  • the geometric features of the main connected region include a centroid position, a center of gravity position, a width, a height, an aspect ratio, an area, an circumscribed rectangle, and an area of the circumscribed rectangle of the main connected region; Geometric features include the position of the equivalent rectangle and the area of the equivalent rectangle.
  • the step of performing horizontal long edge extraction on the grayscale image in the step D includes:
  • the horizontal long edge is the longest horizontal edge among the horizontal edges.
  • the horizontal long edge is matched with the main connected area to obtain a region edge matching relationship, which is specifically:
  • Feature matching is performed on the horizontal long edge and the main connected area, and whether the upper edge and the lower edge matching the horizontal long edge exist in the main connected area, thereby obtaining the regional edge matching relationship.
  • the visor area is generally a rectangular area, there is generally a horizontal edge around the area, so this condition can be used to determine whether the area is a visor area.
  • the step E includes:
  • step E1 determining whether there is a matching edge in the main connected area according to the area edge matching relationship, if not, executing step E2, otherwise, performing step E3;
  • step E2 according to the geometrical features of the main connected area and the horizontal long edge to determine whether the positional relationship of the main connected area is correct, and if so, step E3 is performed; otherwise, step E5 is performed;
  • step E3 determining whether the degree of similarity of the main connected area meets the preset threshold condition, and if so, executing step E4, otherwise, performing step E5;
  • E5. Determine the state of the vehicle sun visor so that the sun visor is not lowered.
  • the invention considers the influence of factors such as the edge matching relationship of the region, the geometric features of the main connected region and the degree of similarity of the rectangle when performing the state determination of the sun visor at night, and the detection accuracy is higher than that of the prior art.
  • the detected image is a window image of the vehicle.
  • the present invention proposes an efficient and practical night sun visor state detection method based on mathematical morphology and connected domain theory, which can effectively reduce the influence of road texture and interior environment on the shading state detection.
  • the detection accuracy rate can reach more than 90%, and can be applied to the actual bayonet system to effectively detect whether the night visor is in the down state, thereby helping the public security department to investigate various violations or illegal acts.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Multimedia (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)

Abstract

本发明公开了一种车辆遮阳板状态的检测方法，包括：获取检测图像；对检测图像进行灰度预处理，从而得到灰度图像；对灰度图像进行主要连通区域提取，并计算主要连通区域的几何特征和矩形相似程度；对灰度图像进行水平长边缘提取，并将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系；根据区域边缘匹配关系、主要连通区域的几何特征和主要连通区域的矩形相似程度对遮阳板的状态进行判定。本发明基于数学形态学和连通域理论，综合利用主要连通区域的几何特征、主要连通区域的矩形相似程度和区域边缘匹配关系来对遮阳板的状态进行检测，检测效率和检测精度高，实用性高。本发明可广泛应用于图像处理领域。

Description

一种车辆遮阳板状态的检测方法
技术领域
本发明涉及图像处理领域，尤其是一种车辆遮阳板状态的检测方法。
背景技术
遮阳板安装在汽车前风挡玻璃上，能防止阳光刺眼，可由驾驶员自行放下或者调整合适的放下角度。当前一些不法分子会在夜间放下遮阳板，利用遮阳板的遮挡性来遮掩其面容特征，防止其面容特征被卡口系统拍摄到，从而逃避了法律的制裁。为了提前对此种情形进行预警，及时地反馈车辆的可疑性，有效地为公安提供参考信息，有必要对车辆的遮阳板状态进行快速、有效的检测。
但是，由于受道路复杂和车内环境复杂等客观因素的影响，现有的夜间车辆遮阳板状态检测方法，存在如下问题：1、道路纹理对遮阳板检测造成了一定的影响；2、容易把形状相似的部分（如车内摆放的矩形物体和车窗所粘贴的标签等）当成遮阳板而引起检测失败；3、实用性不高，难以应用到实际卡口系统中。这些因素严重影响了遮阳板状态检测的效率和检测精度，因此，目前仍没有完善、成熟的夜间遮阳板状态检测算法可应用于实际工程中。
发明内容
为了解决上述技术问题，本发明的目的是：提供一种检测效率高、检测精度高和实用性高的车辆遮阳板状态的检测方法，以满足公安部门的安全预警需求。
本发明解决其技术问题所采用的技术方案是：一种车辆遮阳板状态的检测方法，包括：
A、获取检测图像；
B、对检测图像进行灰度预处理，从而得到灰度图像；
C、对灰度图像进行主要连通区域提取，并计算主要连通区域的几何特征和矩形相似程度；
D、对灰度图像进行水平长边缘提取，并将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系；
E、根据区域边缘匹配关系、主要连通区域的几何特征和主要连通区域的矩形相似程度对遮阳板的状态进行判定。
进一步，所述步骤B，其包括：
B1、对检测图像进行彩色图像灰度化，从而得到灰度化图像；
B2、对灰度化图像进行灰度拉伸，从而得到拉伸图像；
B3、采用数学形态学方法去掉拉伸图像的高亮区域，从而得到灰度图像。
进一步，所述步骤C，其包括：
C1、采用预设的阈值对灰度图像进行阈值化，从而得到二值图像；
C2、对二值图像进行数学形态学处理，从而得到形态学运算图像；
C3、根据遮阳板形状的先验知识，对形态学运算图像进行剔除处理，从而得到主要连通区域；
C4、计算主要连通区域的几何特征和矩形相似程度。
进一步，所述步骤C4，其包括：
C41、计算主要连通区域的几何特征；
C42、提取主要连通区域的等效矩形，并计算等效矩形的几何特征；
C43、根据主要连通区域的几何特征和等效矩形的几何特征对主要连通区域的矩形相似程度进行计算。
进一步，所述主要连通区域的几何特征包括主要连通区域的质心位置、重心位置、宽度、高度、宽高比、面积、外接矩形和外接矩形的面积；所述等效矩形的几何特征包括等效矩形的位置以及等效矩形的面积。
进一步，所述步骤D中对灰度图像进行水平长边缘提取这一步骤，其包括：
D1、对灰度图像进行边缘提取，从而得到边缘图；
D2、对边缘图进行水平边缘点提取，从而得到水平边缘图；
D3、利用连通域原理从水平边缘图中提取出水平长边缘。
进一步，所述步骤D中将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系这一步骤，其具体为：
对水平长边缘和主要连通区域进行特征匹配，寻找主要连通区域中是否存在着与水平长边缘相匹配的上边缘和下边缘，从而得到区域边缘匹配关系。
进一步，所述步骤E，其包括：
E1、根据区域边缘匹配关系判断主要连通区域是否存在着匹配边缘，若不存在，则执行步骤E2，反之，则执行步骤E3；
E2、根据主要连通区域的几何特征和水平长边缘判断主要连通区域的位置关系是否正确，若是，则执行步骤E3，反之，则执行步骤E5；
E3、判断主要连通区域的矩形相似程度是否符合预设的阈值条件，若是，则执行步骤E4，反之，则执行步骤E5；
E4、判定车辆遮阳板的状态为遮阳板已被放下，并记录灰度图像中遮阳板的位置；
E5、判定车辆遮阳板的状态为遮阳板未被放下。
进一步，所述检测图像为车辆的车窗图像。
本发明的有益效果是：基于数学形态学和连通域理论，综合利用主要连通区域的几何特征、主要连通区域的矩形相似程度和区域边缘匹配关系来对遮阳板的状态进行检测，能有效降低道路纹理和车内环境等因素对遮阳状态检测的影响，检测效率和检测精度高，检测正确率可达到90%以上；可以应用在实际卡口系统中对夜间遮阳板是否处于放下状态进行有效检测，实用性高。
附图说明
下面结合附图和实施例对本发明作进一步说明。
图1为本发明一种车辆遮阳板状态的检测方法的步骤流程图；
图2为本发明步骤B的流程图；
图3为本发明步骤C的流程图；
图4为本发明步骤C4的流程图；
图5为本发明步骤D中对灰度图像进行水平长边缘提取的流程图；
图6为本发明步骤E的流程图。
具体实施方式
为了便于下文的描述，首先给出下列名词的定义或解释：
A method of vehicle sun visor’s state detection：一种车辆遮阳板状态的检测方法；
灰度拉伸：基本的图像灰度变换方式之一，通过简单的变换函数对图像的灰度进行变换，以提高图像处理时灰度级的动态范围。
参照图1，一种车辆遮阳板状态的检测方法，包括：
A、获取检测图像；
B、对检测图像进行灰度预处理，从而得到灰度图像；
C、对灰度图像进行主要连通区域提取，并计算主要连通区域的几何特征和矩形相似程度；
D、对灰度图像进行水平长边缘提取，并将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系；
E、根据区域边缘匹配关系、主要连通区域的几何特征和主要连通区域的矩形相似程度对遮阳板的状态进行判定。
其中，检测图像为车辆的车窗图像。
主要连通区域的几何特征，包括主要连通区域的质心位置、重心位置、宽度、高度、宽高比、面积、外接矩形和外接矩形的面积等。
区域边缘匹配关系表示主要连通区域是否存在与水平长边缘相匹配的边缘（即主要连通区域与水平长边缘的位置是否接近）。
主要连通区域是车辆遮阳板的候选区域。
本发明分析夜间遮阳板的各种图像特征，先通过进行主要连通区域提取、水平长边缘提取和区域边缘特征匹配，提取出能用于遮阳板状态判别的有效特征（包括主要连通区域的几何特征、矩形相似程度和边缘匹配关系等），然后再根据提取的特征进行遮阳板状态判别，较为高效和精确。
参照图2，进一步作为优选的实施方式，所述步骤B，其包括：
B1、对检测图像进行彩色图像灰度化，从而得到灰度化图像；
B2、对灰度化图像进行灰度拉伸，从而得到拉伸图像；
B3、采用数学形态学方法去掉拉伸图像的高亮区域，从而得到灰度图像。
本发明对检测图像进行灰度预处理，从而得到灰度图像的步骤如下：
首先，对检测图像进行彩色图像灰度化，进行彩色图像灰度化依据的公式如下：
Figure PCTCN2013085951-appb-M000001
，式中，r、g、b分别为红色通道的分量、绿色通道的分量、蓝色通道的分量，I为像素点的灰度。
然后，利用灰度拉伸方式，使图像亮度分布在合理范围内。
最后，采用数学形态学去掉图像中高亮区域。
参照图3，进一步作为优选的实施方式，所述步骤C，其包括：
C1、采用预设的阈值对灰度图像进行阈值化，从而得到二值图像；
C2、对二值图像进行数学形态学处理，从而得到形态学运算图像；
C3、根据遮阳板形状的先验知识，对形态学运算图像进行剔除处理，从而得到主要连通区域；
C4、计算主要连通区域的几何特征和矩形相似程度。
其中，对二值图像进行数学形态学运算处理包括开运算处理和纵向闭运算处理。
开运算处理是指采用一个长度适当的结构元对二值图像进行数学形态学开运算处理，以去除二值图像中遮阳板和人脸的粘连。
有些情况下，由于光照影响，二值化后的遮阳板区域可能存在纵向部分断裂现象。为了解决这种问题，需要对数学形态学开运算图像进行纵向闭运算处理。纵向闭运算处理的过程如下：首先对二值图像各区域进行标记，从而得到二值图像的标记图像；然后对标记图像进行纵向闭运算，从而得到纵向闭运算图像（即形态学运算图像）。经过纵向闭运算处理后，同一个标记区域内的纵向断裂部分被连接起来，且没有发生区域合并的问题。
对形态学运算图像进行剔除处理是指，根据遮阳板形状的先验知识，剔除形态学运算图像中宽度、高度、面积和宽高比不符合要求的区域，从而得到主要连通区域。
主要连通区域的几何特征和矩形相似程度均用于后续的遮阳板状态判定。
参照图4，进一步作为优选的实施方式，所述步骤C4，其包括：
C41、计算主要连通区域的几何特征；
C42、提取主要连通区域的等效矩形，并计算等效矩形的几何特征；
C43、根据主要连通区域的几何特征和等效矩形的几何特征对主要连通区域的矩形相似程度进行计算。
其中，等效矩形的几何特征包括等效矩形的位置以及等效矩形的面积等。主要连通区域的外接矩形是指主要连通区域外轮廓的外接矩形。主要连通区域的等效矩形是指满足以下原则的矩形：1）等效矩形的重心和主要连通区域的重心相同；2）等效矩形的面积和主要连通区域面积的差值小于预设的阈值。
主要连通区域的矩形相似程度的衡量参数有两个：r1和r2。r1和r2的计算公式分别为：
Figure PCTCN2013085951-appb-M000002
，
Figure PCTCN2013085951-appb-M000003
，式中，S为主要连通区域的面积，Sc表示外接矩形的面积，Sr表示主要连通区域中除去等效矩形后余下的面积，Se表示等效矩形的面积。r1越接近于1，则主要连通区域越接近于矩形，而r1越小，表示主要连通区域中含有致使区域严重偏离矩形的噪声点；r2越接近于0，表示主要连通区域越接近于矩形。这两个参数共同决定主要连通区域相似于矩形的程度。因此，根据主要连通区域的面积、外接矩形的面积与等效矩形的面积计算出r1和r2即可得到主要连通区域的矩形相似程度。
进一步作为优选的实施方式，所述主要连通区域的几何特征包括主要连通区域的质心位置、重心位置、宽度、高度、宽高比、面积、外接矩形和外接矩形的面积；所述等效矩形的几何特征包括等效矩形的位置以及等效矩形的面积。
参照图5，进一步作为优选的实施方式，所述步骤D中对灰度图像进行水平长边缘提取这一步骤，其包括：
D1、对灰度图像进行边缘提取，从而得到边缘图；
D2、对边缘图进行水平边缘点提取，从而得到水平边缘图；
D3、利用连通域原理从水平边缘图中提取出水平长边缘。
其中，水平长边缘为水平边缘中长度最长的水平边缘。
进一步作为优选的实施方式，所述步骤D中将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系这一步骤，其具体为：
对水平长边缘和主要连通区域进行特征匹配，寻找主要连通区域中是否存在着与水平长边缘相匹配的上边缘和下边缘，从而得到区域边缘匹配关系。
由于遮阳板区域一般为矩形区域，所以区域周围一般存在水平边缘，所以可以利用这个条件判断区域是否为遮阳板区域。
参照图6，进一步作为优选的实施方式，所述步骤E，其包括：
E1、根据区域边缘匹配关系判断主要连通区域是否存在着匹配边缘，若不存在，则执行步骤E2，反之，则执行步骤E3；
E2、根据主要连通区域的几何特征和水平长边缘判断主要连通区域的位置关系是否正确，若是，则执行步骤E3，反之，则执行步骤E5；
E3、判断主要连通区域的矩形相似程度是否符合预设的阈值条件，若是，则执行步骤E4，反之，则执行步骤E5；
E4、判定车辆遮阳板的状态为遮阳板已被放下，并记录灰度图像中遮阳板的位置；
E5、判定车辆遮阳板的状态为遮阳板未被放下。
本发明在进行夜间车辆遮阳板状态判定时，综合考虑了区域边缘匹配关系、主要连通区域的几何特征和矩形相似程度等因素的影响，与现有技术相比，检测精度更高。
进一步作为优选的实施方式，所述检测图像为车辆的车窗图像。
与现有技术相比，本发明基于数学形态学和连通域理论，提出了一种高效实用的夜间遮阳板状态检测方法，能有效降低道路纹理和车内环境等因素对遮阳状态检测的影响，检测正确率可达到90%以上，可以应用在实际卡口系统中对夜间遮阳板是否处于放下状态进行有效检测，从而帮助公安部门查处各种违规或违法行为。
以上是对本发明的较佳实施进行了具体说明，但本发明创造并不限于所述实施例，熟悉本领域的技术人员在不违背本发明精神的前提下还可做作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (9)

  1. 一种车辆遮阳板状态的检测方法，其特征在于，包括：
    A、获取检测图像；
    B、对检测图像进行灰度预处理，从而得到灰度图像；
    C、对灰度图像进行主要连通区域提取，并计算主要连通区域的几何特征和矩形相似程度；
    D、对灰度图像进行水平长边缘提取，并将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系；
    E、根据区域边缘匹配关系、主要连通区域的几何特征和主要连通区域的矩形相似程度对遮阳板的状态进行判定。
  2. 根据权利要求1所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤B，其包括：
    B1、对检测图像进行彩色图像灰度化，从而得到灰度化图像；
    B2、对灰度化图像进行灰度拉伸，从而得到拉伸图像；
    B3、采用数学形态学方法去掉拉伸图像的高亮区域，从而得到灰度图像。
  3. 根据权利要求1所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤C，其包括：
    C1、采用预设的阈值对灰度图像进行阈值化，从而得到二值图像；
    C2、对二值图像进行数学形态学处理，从而得到形态学运算图像；
    C3、根据遮阳板形状的先验知识，对形态学运算图像进行剔除处理，从而得到主要连通区域；
    C4、计算主要连通区域的几何特征和矩形相似程度。
  4. 根据权利要求3所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤C4，其包括：
    C41、计算主要连通区域的几何特征；
    C42、提取主要连通区域的等效矩形，并计算等效矩形的几何特征；
    C43、根据主要连通区域的几何特征和等效矩形的几何特征对主要连通区域的矩形相似程度进行计算。
  5. 根据权利要求4所述的一种车辆遮阳板状态的检测方法，其特征在于，所述主要连通区域的几何特征包括主要连通区域的质心位置、重心位置、宽度、高度、宽高比、面积、外接矩形和外接矩形的面积；所述等效矩形的几何特征包括等效矩形的位置以及等效矩形的面积。
  6. 根据权利要求1所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤D中对灰度图像进行水平长边缘提取这一步骤，其包括：
    D1、对灰度图像进行边缘提取，从而得到边缘图；
    D2、对边缘图进行水平边缘点提取，从而得到水平边缘图；
    D3、利用连通域原理从水平边缘图中提取出水平长边缘。
  7. 根据权利要求1所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤D中将水平长边缘与主要连通区域进行特征匹配，从而得到区域边缘匹配关系这一步骤，其具体为：
    对水平长边缘和主要连通区域进行特征匹配，寻找主要连通区域中是否存在着与水平长边缘相匹配的上边缘和下边缘，从而得到区域边缘匹配关系。
  8. 根据权利要求1所述的一种车辆遮阳板状态的检测方法，其特征在于，所述步骤E，其包括：
    E1、根据区域边缘匹配关系判断主要连通区域是否存在着匹配边缘，若不存在，则执行步骤E2，反之，则执行步骤E3；
    E2、根据主要连通区域的几何特征和水平长边缘判断主要连通区域的位置关系是否正确，若是，则执行步骤E3，反之，则执行步骤E5；
    E3、判断主要连通区域的矩形相似程度是否符合预设的阈值条件，若是，则执行步骤E4，反之，则执行步骤E5；
    E4、判定车辆遮阳板的状态为遮阳板已被放下，并记录灰度图像中遮阳板的位置；
    E5、判定车辆遮阳板的状态为遮阳板未被放下。
  9. 根据权利要求1-8任一项所述的一种车辆遮阳板状态的检测方法，其特征在于，所述检测图像为车辆的车窗图像。
PCT/CN2013/085951 2013-08-20 2013-10-25 一种车辆遮阳板状态的检测方法 WO2015024294A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/438,895 US9424476B2 (en) 2013-08-20 2013-10-25 Method for detecting a vehicle sunvisor's state

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201310365024.2 2013-08-20
CN201310365024.2A CN103440647B (zh) 2013-08-20 2013-08-20 一种车辆遮阳板状态的检测方法及系统

Publications (1)

Publication Number Publication Date
WO2015024294A1 true WO2015024294A1 (zh) 2015-02-26

Family

ID=49694340

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2013/085951 WO2015024294A1 (zh) 2013-08-20 2013-10-25 一种车辆遮阳板状态的检测方法

Country Status (3)

Country Link
US (1) US9424476B2 (zh)
CN (1) CN103440647B (zh)
WO (1) WO2015024294A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10846562B2 (en) * 2018-01-12 2020-11-24 Arizona Board Of Regents On Behalf Of Arizona State University Systems and methods for image matching
CN110826411B (zh) * 2019-10-10 2022-05-03 电子科技大学 一种基于无人机图像的车辆目标快速识别方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1710596A (zh) * 2005-06-16 2005-12-21 上海交通大学 基于机器视觉的计算机拼图自动识别方法
CN101179710A (zh) * 2007-11-30 2008-05-14 浙江工业大学 铁路道口智能视频监控装置
CN101183427A (zh) * 2007-12-05 2008-05-21 浙江工业大学 基于计算机视觉的违章停车检测装置
CN101859382A (zh) * 2010-06-03 2010-10-13 复旦大学 一种基于最大稳定极值区域的车牌检测与识别的方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101246551A (zh) * 2008-03-07 2008-08-20 北京航空航天大学 一种快速的车牌定位方法
CN102609702B (zh) * 2012-02-03 2014-04-02 中山大学 一种道路指路标志的快速定位方法以及系统
CN102663760B (zh) * 2012-04-23 2014-08-20 苏州大学 一种对图像中的车辆挡风镜区域进行定位分割的方法
JP5821833B2 (ja) * 2012-12-24 2015-11-24 株式会社デンソー 撮影装置、近赤外光照射装置、およびサンバイザー
CN103065140A (zh) * 2012-12-30 2013-04-24 信帧电子技术（北京）有限公司 车标的定位方法及装置
KR101438640B1 (ko) * 2013-09-02 2014-11-03 현대자동차 주식회사 선바이저 제어장치 및 선바이저 제어방법

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1710596A (zh) * 2005-06-16 2005-12-21 上海交通大学 基于机器视觉的计算机拼图自动识别方法
CN101179710A (zh) * 2007-11-30 2008-05-14 浙江工业大学 铁路道口智能视频监控装置
CN101183427A (zh) * 2007-12-05 2008-05-21 浙江工业大学 基于计算机视觉的违章停车检测装置
CN101859382A (zh) * 2010-06-03 2010-10-13 复旦大学 一种基于最大稳定极值区域的车牌检测与识别的方法

Also Published As

Publication number Publication date
CN103440647B (zh) 2016-12-07
US20160171318A1 (en) 2016-06-16
CN103440647A (zh) 2013-12-11
US9424476B2 (en) 2016-08-23

Similar Documents

Publication Publication Date Title
WO2016088960A1 (ko) 야간 환경에서의 운전자 보조 시스템을 위한 위험 보행자 검출 방법 및 시스템
WO2018223857A1 (zh) 文本行识别方法及系统
WO2013151266A1 (en) Method and system for lane departure warning based on image recognition
US10102614B2 (en) Fog removing device and image generating method
WO2018072354A1 (zh) 车道导向方法及装置
WO2020116680A1 (ko) 차량 번호판 검출 및 인식 시스템 및 방법
Qiao et al. Integrative binocular vision detection method based on infrared and visible light fusion for conveyor belts longitudinal tear
KR101546700B1 (ko) 영상을 활용한 도로 포트홀 영역 인식 장치 및 방법
WO2016110007A1 (zh) 基于三维点云的三维人脸识别装置和方法
WO2015188685A1 (zh) 基于深度相机的人体模型获取方法及网络虚拟试衣系统
WO2013172580A1 (ko) 동영상에 포함된 헤이즈 제거를 위한 영상 처리 장치 및 그 방법
WO2015024294A1 (zh) 一种车辆遮阳板状态的检测方法
US11282235B2 (en) Vehicle surroundings recognition apparatus
WO2016122042A9 (ko) 인공위성 영상과 랜덤포레스트 분류기 결합을 이용한 자동 하천 검출 시스템 및 방법
WO2018070844A1 (ko) 에지 모델링을 위한 에지 블러 설정 방법
WO2019054636A1 (ko) 차량용 카메라 캘리브레이션 장치 및 그 방법
WO2014017697A1 (ko) 가이디드 가버 필터를 이용한 지정맥 패턴 추출 방법 및 장치
WO2018076575A1 (zh) 非法行驶记录方法及装置
EP3649460A1 (en) Apparatus for optimizing inspection of exterior of target object and method thereof
WO2014079327A1 (zh) 信息推送方法和系统、数字电视接收终端及计算机存储介质
TWI394096B (zh) 影像追蹤處理的方法
WO2015199354A1 (ko) 인접 합산 이진화를 이용한 동공 추출 방법 및 이를 이용한 동공 추출 제어장치
WO2021145713A1 (ko) 가상 모델 생성 장치 및 방법
WO2018058857A1 (zh) 车辆行驶路段提示方法及装置
WO2018097677A1 (en) Image processing apparatus, image processing method, and computer-readable recording medium thereof

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: 13891909

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 14438895

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 13891909

Country of ref document: EP

Kind code of ref document: A1