WO2021169996A1 - 一种基于磨痕灰度相似性的磨痕角自动检测方法 - Google Patents

一种基于磨痕灰度相似性的磨痕角自动检测方法 Download PDF

Info

Publication number
WO2021169996A1
WO2021169996A1 PCT/CN2021/077659 CN2021077659W WO2021169996A1 WO 2021169996 A1 WO2021169996 A1 WO 2021169996A1 CN 2021077659 W CN2021077659 W CN 2021077659W WO 2021169996 A1 WO2021169996 A1 WO 2021169996A1
Authority
WO
WIPO (PCT)
Prior art keywords
angle
gray
wear scar
approach
wear
Prior art date
Application number
PCT/CN2021/077659
Other languages
English (en)
French (fr)
Inventor
肖梅
张雷
杨冰
王海明
徐婷
Original Assignee
长安大学
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
Application filed by 长安大学 filed Critical 长安大学
Publication of WO2021169996A1 publication Critical patent/WO2021169996A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/26—Measuring arrangements characterised by the use of optical techniques for measuring angles or tapers; for testing the alignment of axes
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N3/00—Investigating strength properties of solid materials by application of mechanical stress
    • G01N3/56—Investigating resistance to wear or abrasion
    • 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

Definitions

  • the invention relates to an extended application of a four-ball friction tester to determine the lubricating performance of lubricating oil, in particular to a method for automatically detecting wear scars by using the similarity of pixel gray levels in wear scar directions.
  • the lubricating oil with good lubricity can protect the machine and prolong the working life. It is often measured by the four-ball wear tester. According to China's petrochemical industry standards (GB-T 12583-1998 and HT 0762-2005), the determination method of lubricant anti-wear performance is: clamp three steel balls with a diameter of 12.7 mm in an oil box, and test After oil immersion, place a steel ball with a diameter on the top of the three balls.
  • the top ball rotates at a certain speed for 60min, and then take out the three steel balls at the bottom , Measure the wear scar diameters of three steel balls under a microscope with a measurement accuracy of 0.01mm. A total of six sets of measurement data are obtained from the three steel balls. The arithmetic mean of the spot diameter is evaluated. The determination of the direction of the wear scar can facilitate the measurement of the diameter of the wear scar, straighten the shooting angle of the image of the wear scar, and facilitate subsequent analysis and processing such as the strength and density of the wear scar. Based on this, we propose an automatic detection method for the wear scar angle based on the gray similarity of the wear scar.
  • the purpose of the present invention is to provide an automatic wear scar angle detection method based on the gray scale similarity of the wear scar, which solves the defect of inaccurate measurement accuracy caused by the perceptual error of the tester in the existing wear scar angle measurement method. .
  • the invention provides an automatic detection method of wear scar angle based on gray scale similarity of wear scar, including the following steps:
  • Step 1 Perform gray-scale processing on the collected color wear-spot image to obtain a gray-scale wear-spot image
  • Step 2 Select a pixel arbitrarily on the gray-scale wear scar image obtained in Step 1, and construct a neighborhood above the pixel;
  • Step 3 Calculate the gray level difference in the tilt angle direction according to the neighborhood obtained in Step 2;
  • Step 4 Calculate the line gray level change according to the gray level difference in the oblique angle direction obtained in step 3;
  • Step 5 Calculate the line approach angle based on the line gray scale change obtained in step 4.
  • Step 6 Calculate the frequency of occurrence of the line approach angle based on the line approach angle obtained in step 5;
  • Step 7 The frequency of occurrence of the row approach angle calculated in Step 6 selects multiple approach angles that appear at high frequencies;
  • Step 8 Calculate the wear scar angle based on the multiple approach angles with high frequency obtained in step 7.
  • the size of the constructed neighborhood is W ⁇ 2W;
  • the pixel set of the constructed neighborhood is a plurality of pixels on the boundary of the neighborhood, and each pixel satisfies any of the following equations A formula:
  • step 3 the gray level difference in the tilt angle direction is calculated according to the neighborhood obtained in step 2, and the specific method is:
  • [] is the rounding operation of rounding; the value range of ⁇ is 0° ⁇ 180°.
  • the inclination angle ⁇ is calculated by the following formula:
  • the line gray level change is calculated according to the gray level difference in the tilt angle direction obtained in step 3.
  • the specific method is to calculate the i-th line gray level change with the tilt angle ⁇ by the following formula:
  • step 5 the line approach angle is calculated according to the line gray scale change obtained in step 4.
  • the specific method is:
  • ⁇ (i) is the approach angle of the i-th row.
  • step 6 the frequency of occurrence of the row approach angle is calculated according to the row approach angle obtained in step 5.
  • the specific method is:
  • n( ⁇ (i)) represents the total number of rows with the row approach angle ⁇ (i), which is less than or equal to the total number of observation rows ([0.9M]-[0.1M]+1).
  • multiple row approach angles appearing at high frequencies are selected from the row approach angles calculated in step 6, and the multiple row approach angles are selected as follows: the appearance of the row approach angles obtained in step 6 The frequencies are added in order from largest to smallest, and t takes the minimum value and satisfies the following formula, then the t row approach angles are multiple row approach angles that occur with high frequency:
  • T is the frequency selection threshold, 0 ⁇ T ⁇ 1.
  • step 8 the wear scar angle is calculated according to the multiple row approach angles with high frequency obtained in step 7, and the specific method is:
  • the invention provides an automatic detection method for the wear scar angle based on the gray scale similarity of the wear scar, which establishes the gray scale difference of the pixel and its neighboring pixels in different tilt angle directions; on this basis, calculates the line under different tilt angles.
  • Grayscale change based on the similarity of the grayscale of the pixels in the direction of the wear scar, the approach angle of the line can be obtained from the inclination angle at which the grayscale change of the line reaches the minimum; finally, the frequency of occurrence of different approach angles is counted, and the frequency of occurrence is most often An approach angle is calculated to obtain the wear scar angle.
  • the invention converts the grayscale characteristics of the wear scar into the neighborhood grayscale difference of the pixel, overcomes the error generated in the rotation interpolation processing and the extra time-consuming, and improves the detection accuracy.
  • Figure 1 is a schematic diagram of the direction angle ⁇ of the wear scar
  • Fig. 2 is a schematic diagram of a worn spot image F
  • Figure 3 is a schematic diagram of a gray-scale wear pattern f
  • Figure 4 is a W ⁇ 2W neighborhood and neighborhood pixel map of pixel (i, j);
  • Figure 5 is a line chart of H(300, ⁇ );
  • Figure 6 is a broken line graph of the approach angle and its frequency of occurrence.
  • the wear scar angle is used to characterize the direction of the wear scar.
  • the wear scar angle is defined as the smallest positive angle that the wear scar rotates from the clockwise direction to the positive direction of the row axis, which is represented by the symbol ⁇ .
  • the schematic diagram of the wear scar angle ⁇ is shown in Figure 1.
  • the present invention proposes an automatic detection method for the wear scar angle based on the gray scale similarity of the wear scar, which includes the following steps:
  • i and j represent the row and column of the pixel respectively, and i and j are integers, satisfying: 1 ⁇ i ⁇ M And 1 ⁇ j ⁇ N holds.
  • the wear spot image F is shown in FIG. 2.
  • Step 2 Perform gray-scale processing on the collected color wear scar image.
  • the color wear scar image collected in step 1 does not have significant color information, so it is advisable to perform gray-scale processing on the wear scar image first, which can greatly speed up the processing speed.
  • gray-scale wear-spot image F is processed based on the weighted average method to obtain a gray-scale wear-spot image (shown in Figure 3), denoted by f.
  • the gray value of the pixel point (i, j) in the gray-scale wear scar image f is f(i, j), and its calculation formula is shown in formula (1):
  • f(i,j) represents the gray value of the pixel (i,j) in the gray-scale wear pattern f; R(i,j), G(i,j) and B(i,j) represent the wear The red component value, the green component value and the blue component value of the pixel (i, j) in the spot image F.
  • the gray scale wear pattern is shown in FIG. 3.
  • Step 3 Establish the neighborhood of pixels.
  • the gray values of the pixels in the direction of the wear scar are similar, so it is necessary to use the gray difference between the pixel and its neighboring pixels to characterize the characteristics of the wear scar.
  • a neighborhood of size W ⁇ 2W is constructed above the pixel (i,j), and the boundary of the W ⁇ 2W neighborhood is defined, and its coordinates satisfy any of formulas (2)-(4). All pixels in the formula (shown by the white dots in Figure 4) are neighborhood pixels. If (k,l) represents any pixel in the W ⁇ 2W neighborhood of pixel (i,j), it is:
  • W is the size of the neighborhood, and the value is 8-24 pixels.
  • the value of W is 10.
  • Step 4 Calculate the gray difference in the direction of the tilt angle.
  • the angle formed by the straight line (as shown in Figure 4) between the pixel (i, j) and its neighboring pixel (k, l) and the positive direction of the row axis is called the pixel (i, j).
  • the tilt angle ⁇ of its neighboring pixels (referred to as tilt angle).
  • (k,l) is the neighborhood pixel with the tilt angle ⁇ in the W ⁇ 2W neighborhood of (i,j); [] is the rounding operation; ⁇ is the tilt angle, 0° ⁇ 180° , The tilt angle ⁇ is related to the position coordinates of the pixels (i, j) and (k, l), and its calculation formula is shown in (6):
  • the size and number of the tilt angle ⁇ are determined by the size of the neighborhood size W, specifically: 0° ⁇ 180° is divided into 4W sampling points, the larger the neighborhood W size, the more the tilt angle sampling points, the tilt angle The smaller the change step; the smaller the neighborhood size W, the fewer the tilt angle sampling points, and the larger the tilt angle change step.
  • W 10, so the gray level difference h(i, j, ⁇ ) in 40 different tilt angle directions can be observed.
  • Step 5 Calculate the line gray change.
  • the gray level difference of the entire row can reflect the closeness between the tilt angle and the wear scar angle.
  • the gray level change of a row is used to represent the sum of gray levels of all pixels in the row along the same tilt angle. With different tilt angles, the amount of line gray change is usually different.
  • H(i, ⁇ ) represents the grayscale change of the i-th row with the inclination angle ⁇ , as shown in equation (7).
  • Step 6 Calculate the line approach angle. Since the gray levels of the pixels in the direction of the wear scar are similar, when the gray change of the entire row is the smallest, the tilt angle ⁇ and the wear scar angle are the closest. At this time, the tilt angle ⁇ is called the line approach angle.
  • the approach angle ⁇ (i) of the i-th row is equation (8):
  • Step 7 Calculate the frequency of occurrence of the line approach angle.
  • the approach angles of different rows are different, and the frequency of occurrence can be used to characterize the ratio of the number of occurrences of the row approach angles to the total number of rows.
  • n( ⁇ (i)) represents the total number of rows with the row approach angle ⁇ (i), which is less than or equal to the total number of observation rows ([0.9M]-[0.1M]+1).
  • Step 8 Select multiple line approach angles with high frequency: the wear scar angle is determined by the line approach angles with high frequency, and the appearance frequencies of the line approach angles obtained in step 7 are sorted and added in descending order , The maximum t row approach angles ⁇ 1 , ⁇ 2 ,..., ⁇ t that make t take the minimum value and satisfy the following formula (10) are considered to be multiple row approach angles that occur at high frequencies. Select the t row approach angle ⁇ 1 , ⁇ 2 ,..., ⁇ t to calculate the wear scar angle:
  • T is the frequency selection threshold, 0 ⁇ T ⁇ 1.
  • Step 9 Calculate the wear scar angle.
  • the high frequency line approach angle is calculated as the wear scar angle ⁇ :
  • T 0.6.
  • the calculated wear scar angle ⁇ is:
  • the simulation processing platform of the present invention is: an Intel I3 M350 processor, a computer with 2GB memory, and 200 samples are simulated and tested under the MATLAB platform.
  • the algorithm takes 1.313373 seconds. After the hardware implements the algorithm, the running time of the algorithm is still low. Will be greatly reduced.
  • the detection accuracy is characterized by the absolute error ⁇ , which is defined as the absolute value of the difference between the detected wear scar angle and the actual wear scar angle.
  • the actual wear scar angle is manually calibrated. Verification of 200 sample data: the maximum absolute error is 5.3099°, the smallest absolute error is 0°, and the average absolute error is 2.1238°.
  • the algorithm has high detection accuracy and good robustness.

Abstract

一种基于磨痕灰度相似性的磨痕角自动检测方法，包括：建立像素及其邻域像素在不同倾斜角方向的灰度差；在此基础上，计算不同倾斜角下的行灰度变化；基于磨痕方向像素的灰度具有相似性，由行灰度变化达到最小值的倾斜角可得到该行的接近角；最后，统计不同接近角的出现频率，出现频率最大的多个接近角计算得到磨痕角。将磨痕的灰度特性转化为像素的邻域灰度差，克服了旋转插值处理中产生的误差和额外的耗时，提高了检测精度。

Description

一种基于磨痕灰度相似性的磨痕角自动检测方法 技术领域
本发明涉及一种四球摩擦试验机测定润滑油润滑性能的扩展应用，具体涉及一种利用磨痕方向像素灰度相似性的磨痕自动检测的方法。
背景技术
润滑性好的润滑油可以保护机械、延长工作寿命，常常通过四球磨损试验机来测量。根据我国石油化工行业标准(GB-T 12583-1998和H-T 0762-2005)，润滑剂抗摩损性能测定方法为：将三个直径为12.7mm的钢球夹紧在一油盒中，并用试油浸没，在三球顶部放置一个直径的钢球，在试油温度达到75℃±2℃后，施加147N或392N作用力，顶球在一定转速下旋转60min，随后取出底部的三个钢球，在测量精度为0.01mm的显微镜下测量三个钢球的磨斑直径，三个钢球共得到六组测量数据，润滑油或润滑脂的抗摩性能通过三个球的六次测量的摩斑直径的算术平均值来评价。磨痕方向的确定可以便于磨斑直径的测量、摆正磨斑图像的拍摄角度，以及便于后续的磨痕强度、密度等后续的分析处理等等。基于此，我们提出了一种基于磨痕灰度相似性的磨痕角自动检测方法。
发明内容
本发明的目的在于提供的一种基于磨痕灰度相似性的磨痕角自动检测方法，解决了现有的磨痕角测定方法中由于试验人员的感知误差，导致的测量精度不准确的缺陷。
为了达到上述目的，本发明采用的技术方案是：
本发明提供的一种基于磨痕灰度相似性的磨痕角自动检测方法，包括以下步骤：
步骤1，对采集得到的彩色磨斑图像进行灰度化处理，得到灰度磨斑图像；
步骤2，在步骤1中得到的灰度磨斑图像上任意选取一像素，并在该像素的上方构建邻域；
步骤3，根据步骤2中得到的邻域计算倾斜角方向的灰度差；
步骤4，根据步骤3中得到的倾斜角方向的灰度差计算行灰度变化；
步骤5，根据步骤4中得到的行灰度变化计算行接近角；
步骤6，根据步骤5中得到的行接近角计算行接近角的出现频率；
步骤7，在步骤6中计算得到的行接近角的出现频率选取高频出现的多个接近角；
步骤8，根据步骤7中得到的高频出现的多个接近角计算磨痕角。
优选地，步骤2中，构建的邻域的大小为W×2W；构建得到的邻域的像素集为所述近邻域边界上的多个像素点，且每个像素点满足下式中的任一式：
{(k,l)|i-W≤k≤i and l＝j+W}
{(k,l)|i-W≤k<i and l＝j-W}
{(k,l)|k＝i-W and j-W≤l≤j+W}。
优选地，步骤3中，根据步骤2中得到的邻域计算倾斜角方向的灰度差，具体方法是：
将步骤2中任意选取的像素(i,j)和所构建的邻域中的某一像素(k,l)之间的直线与行轴正向之间所成的角定义为像素(i,j)与其邻域像素的倾斜角α；
通过下式计算像素(i,j)与其邻域上倾斜角α方向的像素(k,l)之间的灰度差：
Figure PCTCN2021077659-appb-000001
其中，[]为四舍五入的取整运算；α的取值范围为0°≤α<180°。
优选地，通过下式计算倾斜角α：
Figure PCTCN2021077659-appb-000002
优选地，步骤4中，根据步骤3中得到的倾斜角方向的灰度差计算行灰度变化，具体方法是，通过下式计算倾斜角为α的第i行灰度变化：
Figure PCTCN2021077659-appb-000003
优选地，步骤5中，根据步骤4中得到的行灰度变化计算行接近角，具体方法是：
Figure PCTCN2021077659-appb-000004
其中，β(i)为第i行的接近角。
优选地，步骤6中，根据步骤5中得到的行接近角计算行接近角的出现频率，具体方法是：
Figure PCTCN2021077659-appb-000005
其中，n(β(i))表示行接近角为β(i)的总行数，小于等于观测行的总行数([0.9M]-[0.1M]+1)。
优选地，步骤7中，在步骤6中计算得到的行接近角中选取高频出现的多个行接近角，该多个行接近角的选取为：将步骤6中得到的行接近角的出现频率按照从大到小的顺序进行依次相加，t取最小值并满足下式，则该t个行接近角为高频出现的多个行接近角：
Figure PCTCN2021077659-appb-000006
其中，T为频率选择阈值，0<T≤1。
优选地，步骤8中，根据步骤7中得到的高频出现的多个行接近角计算磨痕角，具体方法 是：
Figure PCTCN2021077659-appb-000007
与现有技术相比，本发明的有益效果是：
本发明提供的一种基于磨痕灰度相似性的磨痕角自动检测方法，建立像素及其邻域像素在不同倾斜角方向的灰度差；在此基础上，计算不同倾斜角下的行灰度变化；基于磨痕方向像素的灰度具有相似性，由行灰度变化达到最小值的倾斜角可得到该行的接近角；最后，统计不同接近角的出现频率，出现频率最大的多个接近角计算得到磨痕角。本发明将磨痕的灰度特性转化为像素的邻域灰度差，克服了旋转插值处理中产生的误差和额外的耗时；提高了检测精度。
附图说明
图1是磨痕方向角θ示意图；
图2是磨斑图像F示意图；
图3是灰度磨斑图f示意图；
图4是像素(i,j)的W×2W邻域及邻域像素图；
图5是H(300,α)折线图；
图6是接近角及其出现频率折线图。
具体实施方式
下面结合附图，对本发明进一步详细说明。
本发明用磨痕角来表征磨痕方向，磨痕角定义为磨痕沿顺时针方向至行轴正向所旋转的最小正角，用符号θ表示。磨痕角θ的示意图如图1所示。
本发明提出了一种基于磨痕灰度相似性的磨痕角自动检测方法，包括如下步骤：
步骤1：通过扫面电子显微镜采集试验钢球的磨斑图像。具体是：在四球摩擦试验结束后，分别将试验所使用的三个底部钢球取出并置于扫面电镜中，并调节扫面电镜的光照和放大倍数 等参数，以便清晰地采集到磨斑图像，所采集的磨斑图像用F表示。同时，得到的所述磨斑图像F的像素大小为M×N，例如磨斑图像的大小为768×1024，即M＝768，N＝1024。同时，利用(i,j)表示磨斑图像F的任一像素点的坐标，则i和j分别表示该像素点的行和列，且i和j均为整数，满足：1≤i≤M和1≤j≤N成立。
本实施例中，磨斑图像用F如图2所示。
步骤2：将采集到的彩色磨斑图像进行灰度化处理。步骤1中采集到的彩色磨斑图像不具有显著的颜色信息，故宜先对磨斑图像进行灰度化处理，这样可以大大加快处理速度。考虑到人眼对颜色的不同敏感性，基于加权平均法对磨斑图像F进行灰度化处理，得到灰度磨斑图(如图3所示)，用f表示。灰度磨斑图像f中像素点(i,j)的灰度值为f(i,j)，其计算式如式(1)所示：
f(i,j)＝0.3·R(i,j)+0.59·G(i,j)+0.11·B(i,j)  (1)
其中，f(i,j)表示灰度磨斑图f中像素(i,j)的灰度值；R(i,j)、G(i,j)和B(i,j)分别表示磨斑图像F中像素(i,j)的红色分量值、绿色分量值和蓝色分量值。
本实施例中，灰度磨斑图如图3所示。
步骤3：建立像素点的邻域。磨痕方向上的像素的灰度值相似，故需要利用像素和其邻域像素的灰度差异来表征磨痕的该特性。以像素为计量单位，在像素(i,j)的上方构建尺寸为W×2W的邻域，并定义W×2W邻域边界上，且其坐标满足式(2)-(4)中的任一式的所有像素点(如图4中白点所示)为邻域像素。若(k,l)表示像素(i,j)的W×2W邻域上的任一像素点，其：
{(k,l)|i-W≤k≤i and l＝j+W}   (2)
{(k,l)|i-W≤k<i and l＝j-W}    (3)
{(k,l)|k＝i-W and j-W≤l≤j+W}   (4)
其中，W为邻域的尺寸，取值为8-24像素。
本实施例中，W取值为10。
步骤4：计算倾斜角方向的灰度差。在像素坐标系中，经过像素(i,j)和及其邻域像素(k,l)之间的直线(如图4所示)与行轴正向之间所成的角叫做像素(i,j)与其邻域像素的倾斜角α(简称倾斜角)。考虑到磨斑图像四周的磨痕信息不丰富，对后续处理的结果会造成干扰，故对图像的四周进行直接赋值处理，像素(i,j)与其W×2W邻域上倾斜角α方向的像素(k,l)的灰度差，用来表征像素间的灰度差异，其计算式为：
Figure PCTCN2021077659-appb-000008
其中，(k,l)为(i,j)的W×2W邻域上倾斜角为α的邻域像素；[]为四舍五入的取整运算；α为倾斜角，0°≤α<180°，倾斜角α与像素(i,j)、(k,l)的位置坐标相关，其计算式如(6)所示：
Figure PCTCN2021077659-appb-000009
倾斜角α的大小和数量由邻域尺寸W的尺寸决定，具体地：将0°～180°不等分为4W个采样点，邻域W尺寸越大，倾斜角采样点越多，倾斜角变化步长越小；邻域尺寸W越小，倾斜角采样点越少，倾斜角变化步长越大。例如：当W＝1时，能观测像素(i,j)分别在0°,45°,90°和135°的灰度差；而当W＝10时，能观测像素(i,j)从0°到180°共40个不同的倾斜角方向 的灰度差。
本实施例中，W＝10，故能观察40个不同的倾斜角方向的灰度差h(i,j,α)。
步骤5：计算行灰度变化。整行的灰度差的大小可以反映出倾斜角和磨痕角间的接近程度。行灰度变化用来表示该行所有像素沿着相同倾斜角的灰度差和。不同倾斜角，行灰度变化量通常不一样。H(i,α)表示倾斜角为α的第i行灰度变化，如式(7)所示。
Figure PCTCN2021077659-appb-000010
本实施例中，当i＝300时，H(300,α)如图5所示。
步骤6：计算行接近角。由于磨痕方向像素的灰度相似，当整行的灰度变化最小时，倾斜角α和磨痕角最为接近，此时倾斜角α称为行接近角。第i行的接近角β(i)为式(8)：
Figure PCTCN2021077659-appb-000011
本实施例中，i＝300时，的接近角β(300)＝48.01279。
步骤7：计算行接近角的出现频率。不同行的接近角不尽相同，用出现频率可以表征行接近角的出现次数占总行数的比值。
Figure PCTCN2021077659-appb-000012
其中，n(β(i))表示行接近角为β(i)的总行数，小于等于观测行的总行数([0.9M]-[0.1M]+1)。
步骤8：选择高频出现的多个行接近角：磨痕角由高频出现的行接近角确定，将步骤7得到的行接近角的出现频率按照从大到小的顺序进行依次排序相加，使得t取最小值并满足下式(10)的最大的t个行接近角β 1,β 2,…,β t，认为是高频出现的多个行接近角。选取该t个行接近角β 1,β 2,…,β t用于计算磨痕角：
Figure PCTCN2021077659-appb-000013
(10)其中，T为频率选择阈值，0<T≤1。
步骤9：计算磨痕角。高出现频率的行接近角计算磨痕角θ为：
Figure PCTCN2021077659-appb-000014
本实施例中，T＝0.6。
出现频率最大的两个(t＝2)行接近角为：β 1＝48.01°(P(β 1)＝0.556)和β 2＝45°(P(β 2)＝0.249)，满足：P(β 1)+P(β 2)＝0.556+0.249＝0.805>T＝0.6。
计算得到的磨痕角θ为：
Figure PCTCN2021077659-appb-000015
按照以上本发明的技术方案，从运行时间和检测精度分析两方面，分析本发明方案的优缺点。
1.运行时间。本发明的仿真处理平台为：Intel I3 M350处理器，2GB内存的计算机，在MATLAB平台下对200个样本进行仿真测试，算法所用的时间为1.313373s秒，硬件实现本算法后，算法运行时间还会大大减少。
2.检测精度分析。为了验证算法的有效性，检测精度用绝对误差ξ来表征，定义为检测磨痕角和实际磨痕角的差值的绝对值，实际磨痕角由人工标定。对200样本数据验证：最大的绝对误差为5.3099°，最小的绝对误差为0°，平均的绝对误差为2.1238°，算法的检测检测精度高，鲁棒性好。

Claims (9)

  1. 一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，包括以下步骤：
    步骤1，对采集得到的彩色磨斑图像进行灰度化处理，得到灰度磨斑图像；
    步骤2，在步骤1中得到的灰度磨斑图像上任意选取一像素，并在该像素的上方构建邻域；
    步骤3，根据步骤2中得到的邻域计算倾斜角方向的灰度差；
    步骤4，根据步骤3中得到的倾斜角方向的灰度差计算行灰度变化；
    步骤5，根据步骤4中得到的行灰度变化计算行接近角；
    步骤6，根据步骤5中得到的行接近角计算行接近角的出现频率；
    步骤7，在步骤6中计算得到的行接近角的出现频率选取高频出现的多个接近角；
    步骤8，根据步骤7中得到的高频出现的多个接近角计算磨痕角。
  2. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤2中，构建的邻域的大小为W×2W；构建得到的邻域的像素集为所述近邻域边界上的多个像素点，且每个像素点满足下式中的任一式：
    {(k,l)|i-W≤k≤i and l＝j+W}
    {(k,l)|i-W≤k<i and l＝j-W}
    {(k,l)|k＝i-W and j-W≤l≤j+W}。
  3. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤3中，根据步骤2中得到的邻域计算倾斜角方向的灰度差，具体方法是：
    将步骤2中任意选取的像素(i,j)和所构建的邻域中的某一像素(k,l)之间的直线与行轴正向之间所成的角定义为像素(i,j)与其邻域像素的倾斜角α；
    通过下式计算像素(i,j)与其邻域上倾斜角α方向的像素(k,l)之间的灰度差：
    Figure PCTCN2021077659-appb-100001
    其中，[]为四舍五入的取整运算；α的取值范围为0°≤α<180°。
  4. 根据权利要求3所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，通过下式计算倾斜角α：
    Figure PCTCN2021077659-appb-100002
  5. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤4中，根据步骤3中得到的倾斜角方向的灰度差计算行灰度变化，具体方法是，通过下式计算倾斜角为α的第i行灰度变化：
    Figure PCTCN2021077659-appb-100003
  6. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤5中，根据步骤4中得到的行灰度变化计算行接近角，具体方法是：
    Figure PCTCN2021077659-appb-100004
    其中，β(i)为第i行的接近角。
  7. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤6中，根据步骤5中得到的行接近角计算行接近角的出现频率，具体方法是：
    Figure PCTCN2021077659-appb-100005
    其中，n(β(i))表示行接近角为β(i)的总行数，小于等于观测行的总行数([0.9M]- [0.1M]+1)。
  8. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤7中，在步骤6中计算得到的行接近角中选取高频出现的多个行接近角，该多个行接近角的选取为：将步骤6中得到的行接近角的出现频率按照从大到小的顺序进行依次相加，t取最小值并满足下式，则该t个行接近角为高频出现的多个行接近角：
    Figure PCTCN2021077659-appb-100006
    其中，T为频率选择阈值，0<T≤1。
  9. 根据权利要求1所述的一种基于磨痕灰度相似性的磨痕角自动检测方法，其特征在于，步骤8中，根据步骤7中得到的高频出现的多个行接近角计算磨痕角，具体方法是：
    Figure PCTCN2021077659-appb-100007
PCT/CN2021/077659 2020-02-26 2021-02-24 一种基于磨痕灰度相似性的磨痕角自动检测方法 WO2021169996A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010121050.0 2020-02-26
CN202010121050.0A CN111412864B (zh) 2020-02-26 2020-02-26 一种基于磨痕灰度相似性的磨痕角自动检测方法

Publications (1)

Publication Number Publication Date
WO2021169996A1 true WO2021169996A1 (zh) 2021-09-02

Family

ID=71491004

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/077659 WO2021169996A1 (zh) 2020-02-26 2021-02-24 一种基于磨痕灰度相似性的磨痕角自动检测方法

Country Status (2)

Country Link
CN (1) CN111412864B (zh)
WO (1) WO2021169996A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116071363A (zh) * 2023-03-27 2023-05-05 山东中济鲁源机械有限公司 一种自动化型钢智能生产监测系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111412864B (zh) * 2020-02-26 2021-12-07 长安大学 一种基于磨痕灰度相似性的磨痕角自动检测方法

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02132568A (ja) * 1988-11-14 1990-05-22 Fuji Electric Co Ltd ２値化回路
CN103761729A (zh) * 2013-12-31 2014-04-30 长安大学 一种基于邻域灰度相似性的钢球磨痕检测方法
CN106770296A (zh) * 2017-01-11 2017-05-31 长安大学 一种四球摩擦试验磨斑图像磨痕方向角自动测定方法
CN106841575A (zh) * 2017-01-11 2017-06-13 长安大学 一种四球摩擦试验磨斑图像磨痕方向自动定位方法
CN107392899A (zh) * 2017-07-20 2017-11-24 长安大学 一种钢球磨斑图像的磨痕水平角的自动检测方法
CN109816644A (zh) * 2019-01-16 2019-05-28 大连理工大学 一种基于多角度光源影像的轴承缺陷自动检测系统
CN111412864A (zh) * 2020-02-26 2020-07-14 长安大学 一种基于磨痕灰度相似性的磨痕角自动检测方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0677818B1 (en) * 1994-04-15 2000-05-10 Canon Kabushiki Kaisha Image pre-processor for character recognition system
US20040147830A1 (en) * 2003-01-29 2004-07-29 Virtualscopics Method and system for use of biomarkers in diagnostic imaging
CN104535004B (zh) * 2015-01-29 2017-05-31 江南大学 一种基于图像处理的纱线直径检测方法
CN109191494B (zh) * 2018-07-13 2021-03-26 深圳大学 润滑油磨粒在线监测方法、终端及存储介质
CN110095468A (zh) * 2019-01-15 2019-08-06 云南安视智能设备有限公司 一种消炎药粒表面缺陷检测的方法
CN110276750A (zh) * 2019-06-17 2019-09-24 浙江大学 一种任意倾斜角晶圆直线边长的提取与晶粒区隔离方法
CN110363756A (zh) * 2019-07-18 2019-10-22 佛山市高明金石建材有限公司 一种用于磨头的磨损检测系统及检测方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02132568A (ja) * 1988-11-14 1990-05-22 Fuji Electric Co Ltd ２値化回路
CN103761729A (zh) * 2013-12-31 2014-04-30 长安大学 一种基于邻域灰度相似性的钢球磨痕检测方法
CN106770296A (zh) * 2017-01-11 2017-05-31 长安大学 一种四球摩擦试验磨斑图像磨痕方向角自动测定方法
CN106841575A (zh) * 2017-01-11 2017-06-13 长安大学 一种四球摩擦试验磨斑图像磨痕方向自动定位方法
CN107392899A (zh) * 2017-07-20 2017-11-24 长安大学 一种钢球磨斑图像的磨痕水平角的自动检测方法
CN109816644A (zh) * 2019-01-16 2019-05-28 大连理工大学 一种基于多角度光源影像的轴承缺陷自动检测系统
CN111412864A (zh) * 2020-02-26 2020-07-14 长安大学 一种基于磨痕灰度相似性的磨痕角自动检测方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
MEI XIAO ET AL.: "Direction of Groove Detection for Wear Image of Four-Ball Friction Test Based on Gray Difference of Inclination Angle.", ADVANCES IN MATERIALS SCIENCE AND ENGINEERING., 1 September 2020 (2020-09-01), ISSN: 1687-8434 *
XIAO MEI, ZHANG LEI;HAN GUANG;YANG JING-SHUAI;LIU LONG: "Antiwear Property Measuring Method of Lubricant based on Wear Scar Detection", JOURNAL OF TRAFFIC AND TRANSPORTATION ENGINEERING, vol. 14, no. 3, 1 June 2014 (2014-06-01), pages 73 - 78, XP055841674, ISSN: 1671-1637 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116071363A (zh) * 2023-03-27 2023-05-05 山东中济鲁源机械有限公司 一种自动化型钢智能生产监测系统

Also Published As

Publication number Publication date
CN111412864A (zh) 2020-07-14
CN111412864B (zh) 2021-12-07

Similar Documents

Publication Publication Date Title
WO2021169996A1 (zh) 一种基于磨痕灰度相似性的磨痕角自动检测方法
CN108764257B (zh) 一种多视角的指针式仪表识别方法
CN115294113B (zh) 一种木饰面板质量检测方法
CN115294120B (zh) 基于图像识别的阀门表面质量检测方法
CN111242123B (zh) 一种基于红外图像的电力设备故障诊断方法
WO2021169638A1 (zh) 一种双邻域的磨痕角自动检测方法
CN116309537A (zh) 一种极耳模具表面油污的缺陷检测方法
CN116137036B (zh) 基于机器学习的基因检测数据智能处理系统
CN110136196B (zh) 一种桥梁裂缝宽度自动测量方法
CN106770296A (zh) 一种四球摩擦试验磨斑图像磨痕方向角自动测定方法
WO2021185035A1 (zh) 一种基于外形特征的异常磨斑图像的检测方法
CN113109240A (zh) 一种计算机实施的粮食不完善粒测定方法和系统
CN103712565B (zh) 一种基于钢球磨痕梯度的磨斑直径测量方法
CN106841575B (zh) 一种四球摩擦试验磨斑图像磨痕方向自动定位方法
CN116645364B (zh) 一种基于图像数据的合金钢铸件气孔缺陷检测方法
CN113724315B (zh) 眼底视网膜血管宽度测量方法、电子设备及计算机可读存储介质
CN111539953A (zh) 钢材脱碳层深度的检测方法
CN107358604B (zh) 一种四球摩擦试验磨斑图像异常检测方法
CN107403445B (zh) 一种用于润滑剂抗磨性的形貌对比分析方法
CN109614758A (zh) 具有空间相关性的圆形形状误差的监控方法
CN115343279A (zh) 一种基于试纸的皮肤油脂评价方法
CN111473944B (zh) 观测流场中存在复杂壁面的piv数据修正方法、装置
CN109726429B (zh) 一种零件小圆弧采样数据拟合处理优化方法
CN113723441B (zh) 一种唇腺病理智能分析系统及方法
CN116977331B (zh) 基于机器视觉的3d模型表面检测方法

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: 21760812

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21760812

Country of ref document: EP

Kind code of ref document: A1