CN116087227A - Method and system for detecting metal foreign matters and metal coating damage on surface of product - Google Patents

Abstract

The invention discloses a method and a system for detecting metal foreign matters and metal coating damage on the surface of a product.

Description

Method and system for detecting metal foreign matters and metal coating damage on surface of product

Technical Field

The invention relates to the technical field of industrial vision detection and defect detection, in particular to a method and a system for detecting metal foreign matters and metal coating damage on the surface of a product.

Background

At present, products directly facing consumers or industrial products facing manufacturers in the market have clear requirements on foreign matters on the surfaces of the products and the state of surface metal plating layers. In particular, in some products having an electrical function, it is generally required that metallic foreign substances cannot be present on the surface or that the metallic plating layer on the surface cannot be damaged.

At present, the common product delivery inspection mostly adopts a manual visual inspection mode, but the imaging means of the method is single, the detection is performed by mistake in a higher proportion corresponding to the distinction of metal foreign matters and the judgment of coating damage, the judgment of the damage to the metal and the coating is basically dependent on the experience of workers, and large uncertainty often exists. And is also difficult to trace back and improve after the problem has occurred.

In recent years, with the rapid development of artificial intelligence technology, particularly computer vision technology, some companies have made their trial use in the field of quality inspection of product appearance. However, from the current development situation of the industry, all the retrieved schemes (including marketed products, documents, patents, etc.) adopt image acquisition under visible light, and then use a traditional algorithm or a deep learning method for detection. However, the metal can be distinguished only from color and morphology under visible light, and is influenced by the structural characteristics of the metal foreign matter, the metal foreign matter can reflect light strongly only at a specific angle, and the imaging effect of the metal object under similar colors can be greatly reduced under the interference of the background texture of the detected object. Thus, the discrimination result of the detection algorithm is caused to have higher proportion of false omission. The identification of metal coating damage, especially in the identification of whether a deep and thin scratch damages the coating itself, is a difficulty in industry that images under visible light are extremely difficult to show high contrast.

Disclosure of Invention

The invention aims to solve the problems and provide a method and a system for detecting metal foreign matters and metal coating damages on the surface of a product.

The invention realizes the above purpose through the following technical scheme:

a method for detecting metal foreign matters and metal coating damage on the surface of a product comprises the following steps:

s1, acquiring an image: different imaging schemes are designed according to different detection objects, and a plurality of images containing product defects at the same position are acquired;

s2, main diagram detection: extracting the interested areas by using an interested area algorithm through the positions of suspected metal and coating defects on the main graph, and reserving the coordinate position of each interested area on the image;

s3, detecting from a graph: cutting out the region at the corresponding position in the slave graph according to the region of interest coordinates extracted from the master graph, and detecting the sub-blocks;

s4, comprehensively deciding the defect result: and comprehensively deciding the judging results of the corresponding areas in the master graph and the slave graph, and finally giving out the judging results.

Further, the acquiring a plurality of images including a defect in step S1 includes:

a) Images taken with different visible light, including coaxial light, ring light, tunnel light;

b) Images shot by using an ultraviolet camera, including images shot by using light sources of different wavelengths, with different metals reflecting different properties for different wavelengths;

c) According to the difference of the metal to be detected and the difference of the light absorption degree, a filter sheet and an image shot under a matched wavelength light source are developed; shooting a master graph and a slave graph, wherein the master graph is subjected to overdetection by default; the image is acquired more accurately by using filters with different frequencies and different bandwidths in the figure.

Further, the main map detection in step S2 includes the steps of:

d) Detecting the two main images through a traditional machine vision algorithm or a deep learning algorithm respectively and acquiring an imaged region of interest;

e) After the region of interest is subjected to unilateral expansion (such as 10 pixels), the sub-blocks are cut out, and the coordinate information of the sub-blocks on the image is recorded, so that the robustness of an algorithm is improved;

f) And summarizing the sub-blocks cut by the two main graphs, and deleting the repeated items.

Further, the detecting from the map in step S3 includes the steps of:

g) Cutting corresponding sub-blocks in the slave graph by using a deep learning algorithm or a traditional algorithm according to the sub-block coordinate position information given in the master graph;

h) And detecting the cut sub-blocks, and obtaining a detection result.

Further, the step S4 is a comprehensive decision of the defect result, and the final decision result is obtained according to the comprehensive decision of the main graph and the three decision results of the sub-blocks at the same position in the sub-graph, and the decision result of the sub-graph occupies a larger weight in the comprehensive decision.

The invention also provides a detection system for the damage of the metal foreign matters and the metal coating on the surface of the product, which comprises the following modules:

an image acquisition module: acquiring a plurality of images which comprise product defects and are applied with different imaging schemes at the same position;

a main diagram preprocessing module: acquiring a region of interest comprising a defect on a main map;

the main diagram sub-block defect training reasoning module: performing defect reasoning operation on the region of interest obtained on the main graph, only retaining sub-blocks suspected to contain metal foreign matters or coating defects, and obtaining the coordinate positions of the sub-blocks in the graph;

a slave graph preprocessing module: cutting out corresponding areas in the main graph according to coordinates of the sub-blocks suspected to contain metal obtained from the main graph;

training and reasoning module from graph sub-block defects: performing defect reasoning operation on the sub-blocks cut from the graph, and calculating whether the sub-blocks contain metal foreign matters or coating defects;

and the defect result comprehensive decision module: and carrying out comprehensive decision analysis on the calculation result of the main graph sub-block and the calculation result of the auxiliary graph sub-block to obtain a final defect discrimination result.

The invention has the beneficial effects that:

1) According to the invention, a means of non-visible light wave band is introduced for the first time in the visual detection field of the product appearance, and different imaging schemes are specially designed according to different detected objects, so that no detection omission of metal to be detected is ensured;

2) According to the invention, different imaging schemes are specially designed according to the difference of the wavelength of reflected light absorbed by the detected object, namely the technology to be detected, so that the metal foreign matters and defects are ensured to be detected completely;

3) The invention adopts a comprehensive decision mechanism, firstly ensures that all metal foreign matters and defects are detected on the main graph, allows certain overdetection, and then correspondingly cuts the region of interest detected on the main graph, and comprehensively compares the decision after the region of interest is cut from the main graph, thereby realizing the detection requirement under the condition of more accurate and lower overdetection.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following description will briefly explain the practical drawings required in the embodiments or the prior art description, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a diagram of a method for detecting damage of metal foreign matters and metal plating layers on the surface of a product in example 1;

FIG. 2 is a graph of absorbance at different wavelengths for several common metals of example 1;

FIG. 3 is a diagram of a system for detecting damage to metallic foreign matters and metallic plating on the surface of a product in example 2.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, based on the examples herein, which are within the scope of the invention as defined by the claims, will be within the scope of the invention as defined by the claims.

Example 1:

in the actual investigation process of the connector production side, it is known that most of the metallic foreign matters existing on the surface of the connector are copper, and the plating damage also takes whether the copper base material is exposed as a judging standard, so that an implementation case of carrying out the copper metallic foreign matters and the plating damage of the connector by utilizing the detection method is described.

As shown in fig. 1, a method for detecting metal foreign matters and plating damage on a connector surface is applied, and specifically includes the following steps:

s1, acquiring an image: different imaging schemes are designed according to different detection objects, and a plurality of images including product defects at the same position are acquired;

in this embodiment, the implementation is specifically as follows: firstly, a parallel light source with better metal distinguishing property under a visible light wave band is utilized, a coaxial light source is selected to be matched with a color camera with 500 ten thousand pixels to be matched with a telecentric lens, a first main image under visible light is collected, the possible existence of metal on the image is in a highlight, white, yellow and the like, the metal is basically confirmed to be metal foreign matters under the highlight, but the excessive detection risk exists especially for detecting copper on the metal under the white and yellow forms;

then, a 500-ten-thousand-pixel ultraviolet camera is used for matching a lens consistent with the resolution and the visual field, and a common 350-nm-wave-band ultraviolet light source is matched for collecting a second main image at the same shooting position, wherein the metal, especially copper, on the image appears to be dark near the metal, and the contrast is greatly improved compared with a visible light image; the contrast of the metal foreign matter image on the nonmetallic surface is basically consistent with that of the image under visible light, and the nonmetallic surface can be used as the supplement of the visible light image;

finally, by utilizing the characteristic that copper has an ultraviolet absorption peak near 320nm, and by utilizing imaging equipment for shooting a second main image, a 320nm narrow-band filter plate and a mercury lamp with 280-320 nm wavelength are matched, a subgraph is acquired, and according to an optical principle, the absorption rate of copper near 320nm is obviously lower than that of other metals, so that a high-contrast state can be displayed on an image, and particularly, in the detection of coating defects, the effect is quite obvious;

s2, main diagram detection: extracting the interested areas by using an interested area algorithm through the positions of suspected metal and coating defects on the main graph, and reserving the coordinate position of each interested area on the image;

in the embodiment, the method is specifically realized by adopting means such as deep learning and the like, respectively calculating two main images, expanding a single side of the two main images by 10 pixels after detecting the part carrying suspected metal and coating damage, cutting out sub-blocks, and recording coordinate information of the sub-blocks on an image; summarizing the sub-blocks obtained from the two main graphs, and deleting repeated items;

s3, detecting from a graph: cutting out the region at the corresponding position in the slave graph according to the region of interest coordinates extracted from the master graph, and detecting the sub-blocks;

cutting corresponding sub-blocks from the main graph according to the sub-block coordinate information obtained from the main graph, and then carrying out problems by utilizing an algorithm to judge whether the sub-blocks contain metal foreign matters or plating damages;

s4, comprehensively deciding the defect result: comprehensively deciding the judging results of the corresponding areas in the main graph and the auxiliary graph, and finally giving out the judging results;

according to the three judging results of the main graph and the auxiliary graph aiming at each sub-block, the final result is given after the comprehensive judgment, and in the embodiment, the judging result of the auxiliary graph is particularly emphasized to occupy larger weight in the comprehensive decision mechanism;

the embodiment provides an application case of detecting copper metal foreign matters and coating damage to a copper substrate by using a detection method of metal foreign matters and metal coating damage on the surface of a product, wherein a sub-block where a defect is possibly located is firstly cut by using a main graph (visible light+ultraviolet), then a strong ultraviolet absorption peak is formed near 320nm by using copper, and a narrow-band filter plate with 320nm wavelength is selected for accurate discrimination, so that full coverage of detection capability of the metal foreign matters and the coating damage is realized. Secondly, a detection mode of a master graph and a slave graph is adopted, so that the advantages of good visible light imaging effect and high sensitivity to metal under ultraviolet imaging are combined, and the problems of large noise and obvious background interference of pure ultraviolet imaging are avoided;

example 2:

as shown in fig. 3, a system for detecting metal foreign matters and metal coating damage on the surface of a product specifically comprises the following modules:

an image acquisition module: acquiring a plurality of images which comprise product defects and are applied with different imaging schemes at the same position;

in the specific implementation, a common color industrial camera, an ultraviolet camera and a filter plate matched with different wavelengths and a light source with different wavelengths are used to obtain 2 main images and 1 auxiliary image information;

a main diagram preprocessing module: acquiring a region of interest comprising a defect on a main map;

in specific implementation, algorithms such as target detection, semantic segmentation and the like are adopted, and ROI areas are respectively obtained and intercepted on 2 main graphs;

the main diagram sub-block defect training reasoning module: performing defect reasoning operation on the region of interest obtained on the main graph, only retaining sub-blocks suspected to contain metal foreign matters or coating defects, and obtaining the coordinate positions of the sub-blocks in the graph;

in the specific implementation, adopting methods such as a target detection algorithm and the like and carrying out reasoning calculation on the segmented sub-blocks according to a specific distinguishing threshold value of the metal foreign matters or the coating damage on the image, only retaining the sub-blocks suspected to contain the metal foreign matters or the coating damage, and recording the coordinate positions of the sub-blocks in the whole image;

a slave graph preprocessing module: cutting out corresponding areas in the main graph according to coordinates of the sub-blocks suspected to contain metal obtained from the main graph;

in the implementation, the sub-blocks used for pairs in the graph are cut according to the sub-block coordinate information obtained in the previous step;

training and reasoning module from graph sub-block defects: performing defect reasoning operation on the sub-blocks cut from the graph, and calculating whether the sub-blocks contain metal foreign matters or coating defects;

in the specific implementation, the cut sub-block data is subjected to reasoning calculation by applying algorithms such as target detection and the like;

and the defect result comprehensive decision module: carrying out comprehensive decision analysis on the calculation result of the main graph sub-block and the calculation result of the auxiliary graph sub-block to obtain a final defect discrimination result;

in a specific implementation, the comprehensive decision is made on 2 judgment results obtained from the main graph and 1 judgment result obtained from the graph, and the final decision can be considered to be made by adopting a weight integration mode, and it is emphasized that the weight of the graph is relatively large, for example, greater than 0.5;

in this embodiment, an ultraviolet imaging mode is introduced in the field of visual detection of surface defects of products, and a superposition fusion of multiple imaging schemes is used, so that the detection mode of main image and auxiliary image is used to change the detection of the metal foreign matters and the damage of the metal coating from the initial image to the detection of certainty, and the detection of the metal foreign matters and the damage of the metal coating on the surface of the products can be realized under the conditions of high efficiency and low over-detection.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims. In addition, the specific features described in the above embodiments may be combined in any suitable manner, and in order to avoid unnecessary repetition, various possible combinations are not described further. Moreover, any combination of the various embodiments of the invention can be made without departing from the spirit of the invention, which should also be considered as disclosed herein.

Claims (6)

1. The method for detecting the damage of the metal foreign matters and the metal plating layers on the surfaces of the products is characterized by comprising the following steps:

s1, acquiring an image: different imaging schemes are designed according to different detection objects, and a plurality of images containing product defects at the same position are acquired;

s2, main diagram detection: extracting the interested areas by using an interested area algorithm through the positions of suspected metal and coating defects on the main graph, and reserving the coordinate position of each interested area on the image;

s3, detecting from a graph: cutting out the region at the corresponding position in the slave graph according to the region of interest coordinates extracted from the master graph, and detecting the sub-blocks;

s4, comprehensively deciding the defect result: and comprehensively deciding the judging results of the corresponding areas in the master graph and the slave graph, and finally giving out the judging results.

2. The method for detecting metal foreign matter and metal plating damage on a product surface according to claim 1, wherein the step S1 of obtaining a plurality of images including defects includes:

a) Images taken with different visible light, including coaxial light, ring light, tunnel light;

b) Images shot by using an ultraviolet camera, including images shot by using light sources of different wavelengths, with different metals reflecting different properties for different wavelengths;

c) According to the difference of the metal to be detected and the difference of the light absorption degree, a filter sheet and an image shot under a matched wavelength light source are developed; shooting a master graph and a slave graph, wherein the master graph is subjected to overdetection by default; the image is acquired more accurately by using filters with different frequencies and different bandwidths in the figure.

3. The method for detecting metal foreign matter and metal coating damage on a product surface according to claim 1, wherein the main pattern detection in the step S2 comprises the following steps:

d) Detecting the two main images through a traditional machine vision algorithm or a deep learning algorithm respectively and acquiring an imaged region of interest;

e) After the region of interest is subjected to unilateral expansion, the sub-blocks are cut out, and the coordinate information of the sub-blocks on the image is recorded, so that the robustness of the algorithm is improved;

f) And summarizing the sub-blocks cut by the two main graphs, and deleting the repeated items.

4. The method for detecting metal foreign matter and metal plating damage on a product surface according to claim 1, wherein the step S3 of detecting from the map comprises the steps of:

g) Cutting corresponding sub-blocks in the slave graph by using a deep learning algorithm or a traditional algorithm according to the sub-block coordinate position information given in the master graph;

h) And detecting the cut sub-blocks, and obtaining a detection result.

5. The method for detecting metal foreign matter and metal coating damage on a product surface according to claim 1, wherein the defect result in step S4 is comprehensively decided, the final decision result is obtained according to three decision results of the sub-blocks at the same position in the main graph and the sub-graph, and the decision result of the sub-graph occupies a larger weight in the comprehensive decision.

6. The detection system for the damage of the metal foreign matters and the metal coating on the surface of the product is characterized by comprising the following modules:

an image acquisition module: acquiring a plurality of images which comprise product defects and are applied with different imaging schemes at the same position;

a main diagram preprocessing module: acquiring a region of interest comprising a defect on a main map;

the main diagram sub-block defect training reasoning module: performing defect reasoning operation on the region of interest obtained on the main graph, only retaining sub-blocks suspected to contain metal foreign matters or coating defects, and obtaining the coordinate positions of the sub-blocks in the graph;

a slave graph preprocessing module: cutting out corresponding areas in the main graph according to coordinates of the sub-blocks suspected to contain metal obtained from the main graph;

training and reasoning module from graph sub-block defects: performing defect reasoning operation on the sub-blocks cut from the graph, and calculating whether the sub-blocks contain metal foreign matters or coating defects;

and the defect result comprehensive decision module: and carrying out comprehensive decision analysis on the calculation result of the main graph sub-block and the calculation result of the auxiliary graph sub-block to obtain a final defect discrimination result.

Images