WO2018072333A1

WO2018072333A1 - Method for detecting wrong component and apparatus

Info

Publication number: WO2018072333A1
Application number: PCT/CN2016/113596
Authority: WO
Inventors: 李红匣
Original assignee: 广州视源电子科技股份有限公司
Priority date: 2016-10-18
Filing date: 2016-12-30
Publication date: 2018-04-26
Also published as: CN106485710A

Abstract

A method for detecting a wrong component and an apparatus, the method comprising: inputting an image of a component to be detected, the image of the component to be detected comprising a printed text of the component to be detected (S1); acquiring a stroke width value corresponding to each pixel point on the image of the component to be detected (S2); acquiring a printed text image on the image of the component to be detected according to the stroke width value corresponding to each pixel point (S3); comparing the printed text image to a preset text template image, and calculating a similarity level between the printed text image and the text template image (S4); and determining that the component to be detected is a wrong component when the similarity level is less than a preset threshold (S5). The provided method for detecting a wrong component and apparatus may effectively avoid interference of illumination when detecting component text, and increase accuracy of a wrong component detection result.

Description

Method and device for detecting component wrong parts

Technical field

The present invention relates to the field of detection, and in particular, to a method and device for detecting a component wrong component.

Background technique

Usually, in order to meet the needs of production, there are many different types of components, such as capacitors. In order to facilitate the differentiation of components of different types, it is necessary to print relevant parameter information on the surface of the component. In order to ensure the quality of the board, the detection of printed characters on the components is a common component error detection in automatic optical inspection (AOI).

The system for detecting the wrong parts of the printed characters on the components mainly includes two parts: text positioning and text comparison. Among them, text positioning plays a crucial role. At present, the text localization algorithm mainly adopts the template matching method, that is, the text template map is used to find the position closest to the detected image. However, the template matching algorithm is easily affected by the illumination. When the illumination changes greatly, the text area obtained by the template matching algorithm will be greatly offset, which is not conducive to the subsequent text comparison or recognition algorithm.

Summary of the invention

The object of the embodiments of the present invention is to provide a method and device for detecting a component wrong component, which can effectively avoid the interference of illumination on the detection component text, and improve the accuracy of the component fault detection result.

To achieve the above object, an embodiment of the present invention provides a method for detecting a component error component, including:

Inputting an image of the component to be detected, wherein the image of the component to be detected includes a printed character of the component to be detected;

Obtaining a stroke width value corresponding to each pixel on the image of the component to be detected;

Obtaining a printed character image on the image of the component to be detected according to a stroke width value corresponding to each pixel point;

Comparing the printed text image with a preset text template image, and calculating a similarity between the printed text image and the text template image;

When the similarity is less than a preset threshold, it is determined that the component to be detected is a wrong component.

Compared with the prior art, the method for detecting a component wrong component according to the present invention realizes the extraction of the printed character image on the component to be detected based on the characteristics of the width of the text stroke, thereby obtaining the printed character image; and the printed text image and the pre-printed image The text template image is compared to determine that the component to be detected with a lower similarity is a wrong component than the template image. Base of the invention The feature of the width of the stroke of the text realizes high positioning accuracy of the component text information, is not easy to be interfered by the illumination, solves the interference problem of the illumination on the detection component in the prior art, and obtains the beneficial effect of greatly improving the accuracy of the component detection. .

Further, the obtaining a stroke width value corresponding to each pixel on the image of the component to be detected includes:

Performing edge detection on the image of the input component to be detected, thereby acquiring a plurality of edge pixel points;

Determining, according to the plurality of edge pixel points, a plurality of stroke point pairs on the edge image; wherein the stroke point pair includes two edge pixel points, and each stroke is determined by the line segment of each stroke point pair being an endpoint;

Calculating a distance value of each of the stroke point pairs;

The distance value of each of the stroke point pairs is marked as a stroke width value corresponding to the pixel point through which the stroke passes.

As a further aspect of the present invention, by acquiring edge pixel points of the image to be detected, a stroke width value is calculated from a distance pair determined for each edge pixel point of the same stroke point pair, and the image of the component to be detected is The pixel points on the same stroke are marked with corresponding stroke width values; the technical solution only needs to extract the features of the pixel group, and is not limited to the horizontal direction text image, and the calculation is simple, and high precision can be ensured.

Further, the determining, according to the plurality of edge pixel points, the plurality of stroke point pairs on the edge image comprises:

Calculating a gradient direction of each of the edge pixel points, and searching for another edge pixel point along a gradient direction of each of the edge pixel points;

Determining whether a sum of a gradient direction of the edge pixel point and a gradient direction of the other edge pixel to be found is smaller than a preset angle threshold, and if yes, determining the edge pixel point and another edge pixel The point is the point pair of the stroke.

Further, the obtaining the stroke width value corresponding to each pixel on the image of the component to be detected further includes:

Obtaining a set of stroke width values of pixels on each of the strokes, calculating a median value in the set of stroke width values, and marking the median value as strokes of all the pixels passing through the corresponding strokes Width value.

Further, the acquiring the printed character image on the image of the to-be-detected component according to the stroke width value corresponding to each pixel point includes: detecting the stroke width value between the adjacent pixel points Whether the difference is within a preset threshold range, and if so, the adjacent pixels are clustered into the same connected area, thereby acquiring a plurality of the connected areas;

Performing filtering processing on the obtained connected regions;

Obtaining the printed character image on the image of the component to be detected according to the connected regions of the plurality of filtering processes.

Further, the filtering processing on the plurality of connected areas includes:

Calculating a variance of a stroke width value of each of the connected regions, if a variance of the stroke width of the connected region is greater than a pre- If the threshold is judged by the variance, the connected area is discarded.

As a further aspect of the present invention, since the stroke width characteristic of the character is relatively stable and the fluctuation of the noise is large, the communication region excluding the variance of the stroke width value can eliminate noise interference.

Further, the filtering processing of the obtained connected areas includes:

Calculating a width and a height of each of the connected regions, if the width of the connected region exceeds a preset width range of the connected region, and/or the height of the connected region exceeds a preset height range of the connected region, discarding the Connected area.

As a further aspect of the present invention, the printed characters on the element to be detected have a certain aspect ratio, and the width and height of the connected region for acquiring the printed character image can be defined, and the connected regions which are too small and too large can be excluded.

Further, comparing the printed text image with the preset text template image includes:

Comparing the image brightness with the image template image, contrasting the image contrast, and comparing the image structure, and respectively obtaining the brightness contrast value and the contrast contrast value of the printed text image and the text template image. And structural comparison values.

Further, the similarity between the printed character image and the text template image is a structural similarity between the printed text image and the text template image, and the structural similarity is calculated by the following formula:

SSIM(X,Y)=[l(X,Y)] ^α ·[c(X,Y)] ^β ·[s(X,Y)] ^γ

Wherein SSIM(X, Y) represents the structural similarity; l(X, Y) represents the brightness contrast value, c(X, Y) represents the contrast contrast value; s(X, Y) represents the The structural contrast value; α, β, and γ are parameters for adjusting the relative importance of the brightness contrast value, the contrast contrast value, and the structural contrast value, respectively, and α>0, β>0, γ>0.

Further, the obtaining the brightness contrast value of the printed text image and the text template image comprises:

Obtaining a brightness contrast value l(X, Y) of the printed text image and the text template image by calculating the following formula:

Wherein, μ _X represents the mean value of the printed character image; μ _Y represents the mean value of the text template image; C ₁ is a constant, C ₁ ≠ 0;

The obtaining contrast contrast values of the printed text image and the text template image includes:

Obtaining a contrast contrast value c(X, Y) of the printed text image and the text template image by calculating the following formula:

Wherein σ _X represents the standard deviation of the printed character image; σ _Y represents the standard deviation of the text template image; C ₂ is a constant, C ₂ ≠ 0;

The obtaining a structural comparison value between the printed text image and the text template image includes:

Obtaining a structural contrast value s(X, Y) of the printed text image and the text template image by calculating the following formula:

Where σ _XY represents the covariance of the standard deviation of the printed character image and the text template image; C ₃ is a constant, C ₃ ≠ 0.

In order to achieve the object of the present invention, the present invention further provides a component error detecting device, comprising:

Inputting an image unit to be detected for inputting an image of the component to be detected, wherein the image of the component to be detected includes a printed character of the component to be detected;

a stroke width value obtaining unit, configured to acquire a stroke width value corresponding to each pixel point on the image of the component to be detected;

a printed character image acquiring unit, configured to acquire a printed character image on the image of the component to be detected according to a stroke width value corresponding to each pixel point;

a similarity acquiring unit, configured to compare the printed text image with a preset text template image, and calculate a similarity between the printed text image and the text template image;

The wrong component determining unit is configured to determine that the component to be detected is a wrong component when the similarity is less than a preset threshold. Compared with the prior art, the present invention provides a component error detecting device that inputs an image of a component to be detected, and realizes extraction of a printed character image on a component to be detected based on a characteristic of a width of a text stroke, thereby acquiring a printed character image; The printed character image is compared with the preset text template image, and it is determined that the component to be detected having a lower similarity than the template image is a wrong component. The device realizes high positioning accuracy of component text information based on the characteristics of the width of the stroke of the text, and is not susceptible to illumination interference, and solves the problem of interference of the illumination on the detection component in the prior art, and obtains the accuracy of detecting the wrong component detection greatly. Beneficial effect.

DRAWINGS

1 is a schematic flow chart of an embodiment of a method for detecting a component wrong component according to the present invention;

2 is a partial view showing a printed text on an image of an element to be detected of an embodiment of the method for detecting a defective component of the present invention;

3 is a schematic flow chart of step S2 of an embodiment of a method for detecting a component wrong component according to the present invention,

4 is a partial, fragmentary view showing a plurality of edge pixel points of an image of a component to be detected in an embodiment of a component error detecting method according to the present invention;

FIG. 5 is a view showing an example of repeatedly marking a stroke width value of a pixel of an image of a to-be-detected element according to an embodiment of the present invention;

6 is a schematic flow chart of step S22 of an embodiment provided by a component error detecting method according to the present invention;

Fig. 7 is a view showing an example of a stroke structure in which printed characters on an image of a member to be detected are present in an embodiment of the method for detecting a defective component of the present invention:

FIG. 8 is a schematic flow chart of step S3 of an embodiment of a component error detecting method according to the present invention; FIG.

9 is a schematic flow chart of step S4 of an embodiment of a method for detecting a component wrong component according to the present invention;

Fig. 10 is a view showing the configuration of an embodiment of a component misdetecting device of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

Referring to FIG. 1, FIG. 1 is a schematic flowchart diagram of an embodiment of a method for detecting a component error component according to the present invention. The implementation specifically includes the following steps:

S1, input an image of the component to be detected, wherein the image of the component to be detected includes a printed text of the component to be detected;

S2. Obtain a stroke width value corresponding to each pixel on the image of the component to be detected;

S3. Obtain a printed character image on the image of the component to be detected according to a stroke width value corresponding to each pixel point;

S4, comparing the printed text image with the preset text template image, and calculating the similarity between the printed text image and the text template image;

S5. When the similarity is less than a preset threshold, determine that the component to be detected is a wrong component.

Referring to step S3 of the embodiment of FIG. 2, it is further explained that the stroke width of the characters in the partial area is generally consistent; similarly, FIG. 2 is a partial example view of the printed characters on the image of the element to be detected; Printing The stroke width value of the text is usually consistent. The present embodiment uses the feature to detect a text area on the image of the component to be detected, thereby distinguishing the printed text on the image of the component to be detected from other scene elements.

Referring to FIG. 3, FIG. 3 is a schematic flowchart of step S2 of the embodiment. The step S2 of obtaining the stroke width value corresponding to each pixel point specifically includes the following steps:

S21: Perform edge detection on the image of the input component to be detected, thereby acquiring a plurality of edge pixel points;

Referring to FIG. 4, FIG. 4 is a partial example view of inputting an image of a plurality of edge pixels to be detected by an image of a component to be detected. Preferably, in the embodiment, detecting an edge of the image of the input component to be detected uses a Canny edge detection operator to detect an image of the component to be detected. The edges to get a few edge pixels. The edge of an image refers to a dramatic change in the local characteristics of an image in a certain area, such as a sudden change in gray scale, a sudden change in color, and a sudden change in texture structure. The edge of the image is the most basic feature of the image and contains most of the information about the image. Since the Canny edge detection operator achieves a relatively good balance between image denoising and edge detail retention, and has the characteristics of simple implementation and fast processing speed, the embodiment preferably uses the edge detection algorithm to obtain the component to be detected. Edge pixel points.

S22. Determine, according to the plurality of edge pixel points, a plurality of stroke point pairs on the edge image; wherein the stroke point pair includes two edge pixel points, and each stroke point is determined by each stroke point pair as a line segment of the endpoint;

Referring to FIG. 4 as an example, an edge pixel p of the example shown in FIG. 4 can find another edge pixel q, and the edge pixel point p and the edge pixel point q are end point segments for representing a stroke, edge pixels. Point p and edge pixel point q are a single point pair;

S23. Calculate a distance value of each pair of stroke points;

Referring to FIG. 4 as an example, the calculation formula of the distance value of the edge pixel point p and the edge pixel point q as a one-point pair:

Where ||pq|| represents the distance value of the edge pixel point p and the edge pixel point q, x _p , y _p represents the coordinate of the edge point p in the image of the element to be detected, x _q , y _q represents the edge pixel point q The coordinates in the image of the component to be detected.

S24. Mark the distance value of each stroke point pair as the stroke width value of the pixel point corresponding to the stroke.

With reference to FIG. 5 as an example, it is assumed that the set of pixel points that pass from the edge pixel point m to the edge pixel point n is {m+1, m+2, . . . , n-2, n-1}; Any pixel in the set may be located on the stroke of the two stroke points. The stroke width of the two strokes is different, so that it is assigned twice in the labeling process. Taking pixel point k as an example, pixel point k is located at two The strokes are preferably applied to the strokes that are passed, and the solution is:

If the pixel point k is never assigned, the stroke width of the pixel point k is labeled as the distance value of the corresponding pair of stroke points, that is, the distance between the edge pixel point m and the edge pixel point n||mn||, where

If the pixel k has been assigned, the smaller value between the edge pixel m and the edge pixel n and the currently assigned stroke width value min(||mn||, curVal) is marked as The stroke width of pixel k, where curVal represents the stroke width value that the current pixel k has assigned.

4 and FIG. 6, FIG. 6 is a schematic flowchart of step S22 of the embodiment, and step S22 specifically includes the following steps:

S221: Calculate a gradient direction of each edge pixel, and search another edge pixel along a gradient direction of each edge pixel.

S222. Determine whether a sum of a gradient direction of the edge pixel point and a gradient direction of the other edge pixel that is found is less than a preset angle threshold. If yes, determine that the edge pixel point and the other edge pixel point are stroke point pairs.

The implementation process of determining the stroke point in the embodiment is described by taking the edge pixel point p and the edge pixel point q of FIG. 4 as an example: firstly calculating the gradient direction d _{p of the} edge pixel point _p ; the gradient direction d _p direction is substantially perpendicular to the direction of the stroke Then looking along the ray r=p+n*d _p , n>0 direction, if another edge pixel q can be found, the gradient direction of the edge pixel q is represented as d _q ; and the edge pixel p gradient direction d gradient direction _p and edge pixels q substantially opposite d _q direction, satisfies the gradient direction of edge pixels p d of the gradient direction _p and edge pixels q, d _q, and less than a preset angle threshold value, preferably, Edge pixel point p and edge pixel point q satisfy the formula

Then, the edge pixel point p and the edge pixel point q are the pair of stroke points sought. If the matching edge pixel q is not found, or if the sum of the gradient direction _{dp of the} edge pixel _p and the gradient direction _dq of the edge pixel _q is less than the preset angle threshold, the ray r is discarded.

In this embodiment, referring to FIG. 3, step S2 further includes step S25, specifically:

S25. Acquire a set of stroke width values of the pixels on each stroke, calculate a median value in the stroke width value set, and mark the median value as a stroke width value of all the pixels passing through the corresponding stroke.

Referring to FIG. 7, in general, there are still a large number of stroke structures similar to those shown in FIG. 7 in the character: taking the pixel point j in the figure as an example, after processing in step S24, the pixel point j stores the stroke width of the two strokes in which the stroke is located. The smaller value, but it is not the correct stroke width of the pixel point j, it needs to be adjusted: for each stroke where the pixel point j is located, the stroke width saved by all the pixels passing through the traversal is obtained by the set {w ₁ , w ₂ ,...,w _n }, calculates the median value w _mid , mid < n of the set {w ₁ , w ₂ , . . . , w _n }, and assigns the stroke width of these pixels to w _mid .

Referring to FIG. 8, FIG. 8 is a schematic flowchart of step S3, and step S3 specifically includes the following steps:

S31: Detecting whether a difference between the stroke width values of adjacent pixel points is within a preset threshold range, and if yes, merging adjacent pixel points into the same connected area, thereby acquiring a plurality of connected areas;

It can be seen from the stroke width of the text that the stroke width of the characters on the component to be detected is substantially the same. If the stroke widths of two adjacent pixels in the image of the component to be detected are similar, it is considered that the two pixels may belong to the same text. Should belong to the same connected area.

S32. Perform filtering processing on the obtained connected areas.

S33. Acquire a printed character image on the image of the component to be detected according to the connected region of the plurality of filtering processes.

The filtering processing method of step S32 preferably adopts the following methods:

Preferably, the variance of the stroke width value of each connected region is calculated, and if the variance of the stroke width value of the connected region is greater than the preset variance determination threshold, the connected region is discarded.

Since the stroke width feature of the text is relatively stable, the variance of the corresponding stroke width value is large; and the fluctuation of the noise is large, and the variance of the corresponding stroke width value is large. The connected region of noise is determined by calculating the variance of the stroke width of each connected region, and is discarded as a connected region of noise.

Preferably, the width and height of each connected area are calculated, and if the width of the connected area exceeds a preset width range of the connected area, and/or the height of the connected area exceeds a preset height range of the connected area, the connected area is discarded.

Since the printed characters on the element to be detected have a certain aspect ratio, it is possible to eliminate the too small and too large connected areas by defining the width and height of the connected areas for determining the text area.

After being processed by the above preferred filtering processing method, the obtained connected region is the text region of the component to be detected.

FIG. 9 and FIG. 9 are schematic diagrams of the process of step S4, and step S4 specifically includes the following steps:

S41: comparing the brightness of the printed image with the text template image, comparing the contrast of the image, and comparing the image structure, respectively obtaining brightness contrast values, contrast contrast values, and structural contrast values of the printed text image and the text template image.

S42. The similarity between the printed text image and the text template image is the structural similarity between the printed text image and the text template image, and the structural similarity is calculated by the following formula:

SSIM(X,Y)=[l(X,Y)] ^α ·[c(X,Y)] ^β ·[s(X,Y)] ^γ

Among them, SSIM (X, Y) represents structural similarity; l (X, Y) represents brightness contrast value, c (X, Y) represents contrast contrast value; s (X, Y) represents structural contrast value; α, β and γ is a parameter for adjusting the relative importance of the brightness contrast value, the contrast contrast value, and the structure contrast value, respectively, and α>0, β>0, γ>0.

In step S41, specifically, the brightness contrast value l(X, Y) of the printed text image and the text template image is obtained by calculating the following formula:

Contrast contrast value c(X, Y) between the word image and the text template image:

Where σ _X represents the standard deviation of the printed text image; σ _Y represents the standard deviation of the text template image; C ₂ is a constant, C ₂ ≠ 0;

In step S41, specifically, the structural contrast value s(X, Y) of the printed text image and the text template image is obtained by calculating the following formula:

Wherein C ₁ , C ₂ , C _{3 are} used to avoid the case where the denominator is 0; preferably, C ₁ = (K ₁ * L) ² , C ₂ = (K ₂ * L) ² , C ₃ = C ₂ /2, where K ₁ =0.01, K ₂ =0.03, L=255.

Generally, the two image similarities can be calculated by calculating the structural similarity of the two images. The structural similarity includes measuring the similarity of images from three aspects: brightness, contrast and structure. Therefore, the present embodiment determines the similarity by comprehensively comparing the contrast values of the brightness, contrast, and structure of the printed character image and the text template image.

In a specific implementation, the image of the component to be detected is input first, and then the stroke width value corresponding to each pixel on the image of the component to be detected is obtained; based on the approximate value of the stroke width value of the printed text on the component to be detected, corresponding to each pixel point The stroke width value of the stroke width is similar to the adjacent pixel points of the stroke width value, thereby obtaining a plurality of connected regions, and filtering the connected region to obtain a printed text image; finally, using the acquired printed text image and the preset Text template Comparing the three aspects of brightness, contrast and structure, and calculating the structural similarity between the printed text image and the text template image as three similar structures, the similarity is determined, and the component to be detected whose similarity is less than the preset threshold is determined as a wrong component.

In this embodiment, based on the characteristics of the width of the text stroke, the extraction of the printed character image on the object to be detected is performed, thereby obtaining the printed text image; and the printed text image is compared with the preset text template image to determine the similarity with the template image. The lower component to be tested is a wrong piece. The invention realizes the positioning accuracy of the component text information is high based on the characteristics of the width of the text stroke, and is not easily interfered by the illumination, and solves the interference problem of the illumination on the detection component in the prior art, and the accuracy of detecting the wrong component of the component is greatly improved. Beneficial effect.

Referring to FIG. 10, FIG. 10 is a diagram showing an embodiment of a component error detecting device according to the present invention, including

Inputting an image unit 1 to be detected for inputting an image of the component to be detected, wherein the image of the component to be detected includes a printed character of the component to be detected;

a stroke width value obtaining unit 2, configured to acquire a stroke width value corresponding to each pixel point on the image of the component to be detected;

The printed character image acquiring unit 3 is configured to obtain a printed character image on the image of the component to be detected according to the stroke width value corresponding to each pixel point;

The similarity obtaining unit 4 is configured to compare the printed text image with the preset text template image, and calculate the similarity between the printed text image and the text template image;

The wrong component determining unit 5 is configured to determine that the component to be detected is a wrong component when the similarity is less than a preset threshold.

In a specific implementation, the image of the component to be detected is input by inputting the image unit 1 to be detected, and the image includes the printed text on the component to be detected; and then, the image of the image to be detected is acquired by the stroke width value acquiring unit 2 a stroke width value corresponding to a pixel, and determining, by the printed character image acquisition unit 3, a printed character image on the image of the component to be detected based on the acquired stroke width value; and then, by the similarity unit 4, the acquired printed character image and The preset text template images are compared to calculate the similarity; finally, the wrong component judging unit 5 judges whether the detecting component is a wrong component based on the similarity.

The embodiment provided by the component wrong component detecting device of the present invention realizes high positioning accuracy of the component text information based on the characteristics of the character stroke width, is less susceptible to illumination interference, and greatly improves the accuracy of component misdetection detection.

The above is a preferred embodiment of the present invention, and it should be noted that those skilled in the art can also make several improvements without departing from the principles of the present invention. protected range.

Claims

A method for detecting a component wrong component, comprising:

Inputting an image of the component to be detected, wherein the image of the component to be detected includes a printed character of the component to be detected;

Obtaining a stroke width value corresponding to each pixel on the image of the component to be detected;

Obtaining a printed character image on the image of the component to be detected according to a stroke width value corresponding to each pixel point;

Comparing the printed text image with a preset text template image, and calculating a similarity between the printed text image and the text template image;

When the similarity is less than a preset threshold, it is determined that the component to be detected is a wrong component.
The method for detecting a component error component according to claim 1, wherein the obtaining a stroke width value corresponding to each pixel on the image of the component to be detected comprises:

Performing edge detection on the image of the input component to be detected, thereby acquiring a plurality of edge pixel points;

Determining, according to the plurality of edge pixel points, a plurality of stroke point pairs on the edge image; wherein the stroke point pair includes two edge pixel points, and each stroke is determined by the line segment of each stroke point pair being an endpoint;

Calculating a distance value of each of the stroke point pairs;

The distance value of each of the stroke point pairs is marked as a stroke width value corresponding to the pixel point through which the stroke passes.
The method of detecting a component error component according to claim 2, wherein the determining, according to the plurality of edge pixel points, a plurality of stroke point pairs on the edge image comprises:

Calculating a gradient direction of each of the edge pixel points, and searching for another edge pixel point along a gradient direction of each of the edge pixel points;

Determining whether a sum of a gradient direction of the edge pixel point and a gradient direction of the other edge pixel to be found is smaller than a preset angle threshold, and if yes, determining the edge pixel point and another edge pixel The point is the point pair of the stroke.
The method for detecting a component of a component according to claim 2, wherein the obtaining a stroke width value corresponding to each pixel on the image of the component to be detected further comprises:

Obtaining a set of stroke width values of pixels on each of the strokes, calculating a median value in the set of stroke width values, and marking the median value as strokes of all the pixels passing through the corresponding strokes Width value.
The method of detecting a component error component according to claim 1, wherein the image of the printed character on the image of the component to be detected is acquired according to a stroke width value corresponding to each pixel; Whether the difference of the stroke width values between the adjacent pixels is within a preset threshold range, and if so, the adjacent pixels are clustered into the same connected area, thereby acquiring a plurality of the connected areas ;

Performing filtering processing on the obtained connected regions;

Obtaining the printed character image on the image of the component to be detected according to the connected regions of the plurality of filtering processes.
The method of detecting a component error component according to claim 5, wherein the filtering processing on the plurality of connected regions comprises:

Calculating a variance of a stroke width value of each of the connected regions, and discarding the connected region if a variance of a stroke width of the connected region is greater than a preset variance determination threshold.
The method for detecting a component error component according to claim 5, wherein the filtering processing of the plurality of acquired connected regions comprises:

Calculating a width and a height of each of the connected regions, if the width of the connected region exceeds a preset width range of the connected region, and/or the height of the connected region exceeds a preset height range of the connected region, discarding the Connected area.
The method for detecting a component error component according to claim 1, wherein the comparing the printed text image with the preset text template image comprises:

Comparing the image brightness with the image template image, contrasting the image contrast, and comparing the image structure, and respectively obtaining the brightness contrast value and the contrast contrast value of the printed text image and the text template image. And structural comparison values.
The method of detecting a component wrong component according to claim 8, wherein the similarity between the printed character image and the text template image is a structural similarity between the printed character image and the text template image. Structural phase The similarity is calculated by the following formula:

SSIM(X,Y)=[l(X,Y)] α ·[c(X,Y)] β ·[s(X,Y)] γ

Wherein SSIM(X, Y) represents the structural similarity; l(X, Y) represents the brightness contrast value, c(X, Y) represents the contrast contrast value; s(X, Y) represents the The structural contrast value; α, β, and γ are parameters for adjusting the relative importance of the brightness contrast value, the contrast contrast value, and the structural contrast value, respectively, and α>0, β>0, γ>0.
The method for detecting a component error component according to claim 9, wherein the obtaining a brightness contrast value between the printed character image and the text template image comprises:

Obtaining a brightness contrast value l(X, Y) of the printed text image and the text template image by calculating the following formula:

Wherein, μ X represents the mean value of the printed character image; μ Y represents the mean value of the text template image; C 1 is a constant, C 1 ≠ 0;

The obtaining contrast contrast values of the printed text image and the text template image includes:

Obtaining a contrast contrast value c(X, Y) of the printed text image and the text template image by calculating the following formula:

Wherein σ X represents the standard deviation of the printed character image; σ Y represents the standard deviation of the text template image; C 2 is a constant, C 2 ≠ 0;

The obtaining a structural comparison value between the printed text image and the text template image includes:

Obtaining a structural contrast value s(X, Y) of the printed text image and the text template image by calculating the following formula:

Where σ XY represents the covariance of the standard deviation of the printed character image and the text template image; C 3 is a constant, C 3 ≠ 0.
A component wrong component detecting device, characterized in that

Inputting an image unit to be detected for inputting an image of the component to be detected, wherein the image of the component to be detected includes a printed character of the component to be detected;

a stroke width value obtaining unit, configured to acquire a stroke width value corresponding to each pixel point on the image of the component to be detected;

a printed character image acquiring unit, configured to acquire a printed character image on the image of the component to be detected according to a stroke width value corresponding to each pixel point;

a similarity acquiring unit, configured to compare the printed text image with a preset text template image, and calculate a similarity between the printed text image and the text template image;

The wrong component determining unit is configured to determine that the component to be detected is a wrong component when the similarity is less than a preset threshold.