CN118096751A

CN118096751A - Injection molding appearance defect detection method and device, electronic equipment and storage medium

Info

Publication number: CN118096751A
Application number: CN202410504142.5A
Authority: CN
Inventors: 王水亮; 陈浩林; 李国林; 孙通; 吴晓飞; 郭思谊; 颜润明; 吕汉平
Original assignee: Midea Group Co Ltd; GD Midea Air Conditioning Equipment Co Ltd
Current assignee: Midea Group Co Ltd; GD Midea Air Conditioning Equipment Co Ltd
Priority date: 2024-04-25
Filing date: 2024-04-25
Publication date: 2024-05-28
Anticipated expiration: 2044-04-25
Also published as: CN118096751B

Abstract

The invention provides a method, a device, electronic equipment and a storage medium for detecting appearance defects of injection molding parts, which belong to the technical field of intelligent detection, wherein the method comprises the following steps: determining a plurality of preset detection points for searching gray value mutation points in steps on an appearance binarized image of an injection molding to be detected, selecting at least two effective points from the gray value mutation points to determine a foreground separation boundary line, and inputting an image with a background removed into an appearance defect detection model. Before the injection molding image is input into the detection model, the pixel points with the gray value mutation in the image are searched in a stepping way by utilizing the preset detection points, the gray value mutation points influenced by the water gap or the appearance defects are eliminated, and the foreground separation boundary line in the image is determined according to at least two effective gray value mutation points, so that the appearance binarized image is subjected to the pretreatment of removing the background, the image input into the model is more effective, the interference factors are fewer, and the accuracy of the detection of the appearance defects of the injection molding can be improved.

Description

Injection molding appearance defect detection method and device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of intelligent detection technologies, and in particular, to a method and an apparatus for detecting an appearance defect of an injection molding part, an electronic device, and a storage medium.

Background

The detection of defects on the appearance of injection molding parts is an indispensable link in the production process of products, and the detection of the appearance by adopting a machine vision detection technology is also widely applied to the fields of precision manufacturing production lines, on-line automatic detection of industrial product quality and the like.

At present, when the appearance of an injection molding part is detected by adopting a machine vision detection technology, a photographed original image is often directly input into a detection model, and an appearance defect detection result output by the detection model is obtained.

However, since the photographed original image is not properly preprocessed, the method of directly inputting the original image into the detection model is only suitable for a simple and variable-controllable environment such as a laboratory, and once the detection background in the original image becomes complex, the performance of the detection model is seriously affected.

Disclosure of Invention

The invention provides a method, a device, electronic equipment and a storage medium for detecting appearance defects of an injection molding part, which are used for solving the defect that original images are not properly preprocessed in the prior art.

In a first aspect, the present invention provides a method for detecting an appearance defect of an injection molded part, including:

performing binarization processing on a shot image of the injection molding to be detected to obtain an appearance binarized image;

determining a plurality of preset detection points on the appearance binarized image, and searching gray value mutation points in a stepping way by utilizing the preset detection points;

Selecting at least two effective gray value mutation points from all gray value mutation points to determine a foreground separation boundary line based on the effective gray value mutation points;

Removing the background in the appearance binarization image according to the foreground separation boundary line to obtain a detection binarization image;

And inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model.

According to the method for detecting the appearance defects of the injection molding, a plurality of preset detection points are determined on the appearance binarized image, and the method comprises the following steps:

determining the number of the preset detection points according to the specification parameters of the injection molding to be detected;

And determining the distribution of the preset detection points according to the position of the injection molding to be detected on the appearance binarized image.

According to the method for detecting the appearance defects of the injection molding piece, the plurality of preset detection points at least comprise a first search point set and a second search point set;

the first search point set and the second search point set both comprise at least two preset detection points;

When the gray value abrupt change points are searched in a stepping mode by utilizing the preset detection points, the moving directions of all the preset detection points in the first search point set are first directions, the moving directions of all the preset detection points in the second search point set are second directions, and the first directions are opposite to the second directions.

According to the method for detecting the appearance defects of the injection molding, the arrangement directions of the first search point set and the second search point set are perpendicular to the width direction or the length direction of the appearance binarized image;

When the arrangement direction is perpendicular to the width direction of the appearance binarized image, the first direction and the second direction are respectively towards two side edges of the appearance binarized image in the width direction;

When the arrangement direction is perpendicular to the longitudinal direction of the appearance binarized image, the first direction and the second direction are respectively towards two side edges of the appearance binarized image in the longitudinal direction.

According to the method for detecting the appearance defects of the injection molding piece, provided by the invention, a searching point set to which any preset detecting point belongs is determined aiming at any preset detecting point so as to determine a stepping direction when the gray value mutation points are searched in a stepping way, and the step searching of the gray value mutation points by using the preset detecting point comprises the following steps:

When the gray value of the pixel point where any preset detection point is located is 1, controlling the any preset detection point to perform step search by taking one pixel as a step length in the step direction;

judging whether the gray value of the pixel point where any preset detection point is located after each step search is suddenly changed or not;

If no mutation occurs, controlling any preset detection point to continue step searching until the gray value of the pixel point where the preset detection point is located is suddenly changed;

and taking the pixel points with the gray values suddenly changed after the step search of any preset detection point as the gray value suddenly changed points.

According to the method for detecting the appearance defects of the injection molding, provided by the invention, at least two effective gray value mutation points are selected from all gray value mutation points, and the method comprises the following steps:

calculating a first slope and a second slope between each gray value abrupt point and two adjacent gray value abrupt points respectively;

If the directions of the first slope and the second slope are inconsistent, the gray value abrupt change point is used as an invalid gray value abrupt change point;

and removing the invalid gray value mutation points in all gray value mutation points to obtain the valid gray value mutation points.

According to the method for detecting the appearance defects of the injection molding, the foreground separation boundary line is determined based on the effective gray value mutation points, and the method comprises the following steps:

performing straight line fitting on all the effective gray value mutation points based on a least square method to obtain a fitting straight line;

And taking the straight line where the two effective gray value abrupt change points with the shortest perpendicular distance from the fitting straight line are located as the foreground separation boundary line.

According to the method for detecting the appearance defects of the injection molding provided by the invention, the detected binarized image is input into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model, and the method comprises the following steps:

dividing the detected binarized image into a plurality of sub-images using a sliding window;

inputting each sub-image to the appearance defect detection model in sequence to obtain an appearance defect detection result output by the appearance defect detection model; the appearance defect detection result comprises detection results corresponding to each sub-image.

According to the method for detecting the appearance defects of the injection molding, the sliding window is utilized to divide the detected binary image into a plurality of sub-images, and the method comprises the following steps:

controlling the sliding window to move step by step on the detected binarized image according to a preset step length so as to preliminarily define dividing lines of a plurality of sub-images;

and dividing the detection binarized image according to the dividing line to obtain a plurality of sub-images.

According to the method for detecting the appearance defects of the injection molding, after the dividing lines of a plurality of sub-images are primarily defined, the method further comprises the following steps:

If the abnormal pixel points exist on the dividing line, the abnormal pixel points are the pixel points with the gray value of 0;

Step-by-step adjusting the position of the dividing line along the direction perpendicular to the dividing line until the abnormal pixel point does not exist on the dividing line;

and dividing the position of the dividing line after stepping adjustment to obtain a new sub-image, and preprocessing the new sub-image so that the size of the new sub-image is the same as the size of the sub-image obtained by dividing before stepping adjustment of the position of the dividing line.

According to the method for detecting the appearance defects of the injection molding, the preset step length is smaller than the size of the sliding window.

According to the method for detecting the appearance defects of the injection molding, the appearance defects are detected by using the target frame to mark the appearance defects in the appearance binarized image;

correspondingly, after the appearance defect detection result output by the appearance defect detection model is obtained, the method further comprises the following steps:

if the area of the appearance defect is larger than a first preset area, determining that the injection molding piece to be detected is unqualified;

If the area of the appearance defect is smaller than or equal to the first preset area but larger than the second preset area, and other appearance defects exist in the preset range of the appearance defect, determining that the injection molding to be detected is unqualified;

and if the area of the appearance defect is smaller than or equal to a second preset area, confirming that the injection molding to be detected is qualified.

In a second aspect, the present invention also provides an apparatus for detecting an appearance defect of an injection molded part, including:

The binarization processing unit is used for performing binarization processing on the shot image of the injection molding piece to be detected to obtain an appearance binarized image;

The abrupt point searching unit is used for determining a plurality of preset detection points on the appearance binarized image so as to search gray value abrupt points in a stepping way by utilizing the preset detection points;

A boundary line determining unit for selecting at least two effective gray value abrupt points from all gray value abrupt points to determine a foreground separation boundary line based on the effective gray value abrupt points;

the background removing unit is used for removing the background in the appearance binarized image according to the foreground separation boundary line to obtain a detection binarized image;

And the defect detection unit is used for inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model.

In a third aspect, the present invention provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method for detecting an appearance defect of an injection molded part as described in any one of the above when executing the program.

In a fourth aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of detecting an appearance defect of an injection molded part as described in any of the above.

According to the injection molding appearance defect detection method, the injection molding appearance defect detection device, the electronic equipment and the storage medium, before the injection molding image is input into the detection model, the pixel points with the gray value mutation in the image are searched in a stepping mode through the preset detection points, the gray value mutation points influenced by the water gap or the appearance defect are eliminated, and the foreground separation boundary line in the image is determined according to at least two effective gray value mutation points, so that the appearance binarized image is subjected to background elimination pretreatment, the image input into the model is more effective, interference factors are fewer, and the accuracy of the injection molding appearance defect detection can be improved.

Drawings

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

FIG. 1 is a schematic flow chart of an injection molding appearance defect detection method provided by the invention;

FIG. 2 is a schematic diagram of a preset detection point step-by-step searching gray value mutation points according to the present invention;

FIG. 3 is a schematic diagram of a step-by-step searching of gray-scale value mutation points at a preset detection point according to the present invention;

FIG. 4 is a third schematic diagram of a step-by-step searching of gray-scale value mutation points at a preset detection point according to the present invention;

FIG. 5 is a schematic diagram of a step-by-step searching of gray-scale value mutation points at a preset detection point according to the present invention;

FIG. 6 is a second flow chart of the method for detecting defects in an injection molding according to the present invention;

fig. 7 is a schematic flow chart of an air conditioner panel appearance detection method provided by the invention;

FIG. 8 is a schematic diagram of an apparatus for detecting appearance defects of an injection molded part according to the present invention;

Fig. 9 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It should be noted that in the description of the present invention, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.

The terms "first," "second," and the like in this specification are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged, as appropriate, such that embodiments of the present invention may be implemented in sequences other than those illustrated or described herein, and that the objects identified by "first," "second," etc. are generally of a type, and are not limited to the number of objects, such as the first object may be one or more.

The method, the device, the electronic equipment and the storage medium for detecting the appearance defects of the injection molding part provided by the invention are described below with reference to fig. 1 to 9.

It should be noted that, the execution main body of the injection molding appearance defect detection method provided by the invention is a corresponding injection molding appearance detection device, and the injection molding appearance detection device comprises or is connected with a server and computer equipment, and specific forms comprise, but are not limited to, any one of a mobile phone, a tablet Personal computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an Ultra-Mobile Personal Computer (UMPC), a netbook or a Personal digital assistant (Personal DIGITAL ASSISTANT, PDA) and the like.

FIG. 1 is a schematic flow chart of the method for detecting defects in appearance of an injection molded part according to the present invention, as shown in FIG. 1, including but not limited to the following steps:

Step 110: and carrying out binarization processing on the shot image of the injection molding to be detected to obtain an appearance binarized image.

The binarization process refers to a process of converting a photographed image of an injection molding to be detected into a binary image containing only two pixel values.

For example, a photographed image of an injection molding to be inspected is converted into a binary image containing only two pixel values of a gray minimum value (0, representing black) and a gray maximum value (1 or 255, representing white).

Before binarizing the shot image of the injection molding to be detected, the shot image of the injection molding to be detected needs to be acquired.

For example, the shooting device is arranged right above the conveyor belt, one surface of the injection molding to be detected faces to the right above the conveyor belt, and in the process that the injection molding to be detected is transmitted through the conveyor belt, the shooting device is used for shooting the injection molding to be detected, so that a shooting image of the injection molding to be detected is obtained. The shooting equipment can also be arranged right in front of the conveyor belt, one surface of the injection molding piece to be detected faces to the right in front of the conveyor belt, and shooting images of the injection molding piece to be detected can be obtained by using the shooting equipment.

After the shooting image of the injection molding to be detected is obtained, the shooting original image is not directly input into the appearance defect detection model, but the shooting image of the injection molding to be detected is subjected to binarization processing, information in the shooting image is simplified, the target contour and the characteristics of the injection molding to be detected are highlighted, and an appearance binarization image which is convenient for subsequent processing and analysis is obtained.

Optionally, shooting equipment for shooting the injection molding to be detected is an 8K line scanning camera and a large depth of field lens.

Optionally, the injection molding to be detected is photographed using a tunnel-ray photo imaging technique.

Step 120: and determining a plurality of preset detection points on the appearance binarized image so as to search gray value mutation points in a stepping way by utilizing the preset detection points.

Specifically, after the captured image is subjected to binarization processing to obtain an appearance binarized image, a plurality of preset detection points are selected on the appearance binarized image, and the selected preset detection points respectively move in a stepping manner on the appearance binarized image, so that gray value mutation points on the appearance binarized image are searched.

The gray value abrupt change point means that the gray value of the pixel where the preset detection point is located is abrupt in the stepping process of any preset detection point, and generally the gray value of the pixel where the preset detection point is located is changed from the gray maximum value (1 or 255) to the gray minimum value (0), so that the pixel where the preset detection point is located before the gray value abrupt change or the pixel where the preset detection point is located after the gray value abrupt change can be determined as the gray value abrupt change point.

In an embodiment, the determining a plurality of preset detection points on the appearance binarized image includes: determining the number of the preset detection points according to the specification parameters of the injection molding to be detected; and determining the distribution of the preset detection points according to the position of the injection molding to be detected on the appearance binarized image.

For example, when the specification parameters of the injection molding to be detected are smaller, 5-8 preset detection points can be selected on the appearance binarized image; when the specification parameters of the injection molding piece to be detected are large, 8-16 preset detection points can be selected on the appearance binarization image.

Further, after the number of preset detection points is determined, the distribution of the preset detection points is determined according to the positions of the injection molding to be detected on the appearance binarized image.

For example, when the upper and/or lower boundary of the injection molding to be detected is respectively overlapped with the upper and/or lower boundary of the appearance binarized image, the preset detection points are longitudinally arranged and distributed at equal intervals or unequal intervals in the direction perpendicular to the upper and/or lower boundary of the appearance binarized image; when the left and/or right boundary of the injection molding piece to be detected is coincident with the left and/or right boundary of the appearance binarized image, the preset detection points are distributed in a transverse arrangement mode at equal intervals or unequal intervals in the direction perpendicular to the left and/or right boundary of the appearance binarized image. It is understood that the upper, lower, left and right are used herein for clearly describing the distribution of the preset detection points on the appearance binarized image, and should not be construed as limiting the present invention.

Step 130: at least two effective gray value abrupt points are selected from all gray value abrupt points to determine a foreground separation boundary line based on the effective gray value abrupt points.

The foreground separation boundary line refers to a boundary line between an injection molding part (foreground) to be detected in the appearance binarized image and environments (background) such as a conveyor belt, a production workshop and the like.

Specifically, after the gray value mutation point on the appearance binarized image is searched by using the preset detection point step by step, the searched gray value mutation point may be any boundary point on the boundary line of the foreground and the background in the appearance binarized image, or any point on the boundary of the appearance defect, the water gap and the like in the injection molding to be detected in the appearance binarized image.

Therefore, in order to better determine the boundary line between the injection molding to be detected and the environmental background in the appearance binarized image, it is necessary to exclude the interference of gray value mutation points located on the boundary of appearance defects, water gaps and the like, select at least two effective gray value mutation points located on the boundary line between the foreground and the background from gray value mutation points searched stepwise from a plurality of preset detection points, and determine the foreground separation boundary line according to the selected effective gray value mutation points.

Step 140: and removing the background in the appearance binarized image according to the foreground separation boundary line to obtain a detection binarized image.

Specifically, after a foreground separation boundary line is determined through an effective gray value mutation point searched by a preset detection point step, cutting out the background in the appearance binarized image and the environment background such as a conveyor belt, a production workshop and the like according to the foreground separation boundary line, removing the background in the appearance binarized image, and reserving the foreground formed by injection molding to be detected in the appearance binarized image to obtain a detection binarized image.

Step 150: and inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model.

The appearance defect detection result refers to a result of judging whether any one of appearance defects such as variegated, material patterns, clamping lines, black spots, scratches, top white, air lines, bubbles, diffusion lines, greasy dirt, mold dust, bulges and the like is detected on a detected binary image corresponding to the injection molding to be detected.

Specifically, the detected binarized image after the background is removed is input into a trained appearance defect detection model, and an appearance defect detection result which is output by the appearance defect detection model and is whether an appearance defect exists on the injection molding to be detected is obtained.

It should be noted that the appearance defect detection model may be obtained by inputting a plurality of history detection binarized images and a plurality of labels corresponding to appearance defect detection results in a production process into any machine vision algorithm such as convolutional neural network (Convolutional Neural Networks, CNN), generating countermeasure network (GENERATIVE ADVERSARIAL Networks, GAN), multi-Scale convolutional neural network (MS-CNN), or support vector machine (Support Vector Machine, SVM), and the like, which is not limited in the present invention.

No matter the shot image or the appearance binarized image of the injection molding to be detected, the environmental background in the shot image or the appearance binarized image of the injection molding can not have appearance defects of the injection molding. Due to limitations of shooting imaging, there is often an environmental background in the shot image of the injection molding to be detected. After the shot image is subjected to binarization processing, if the appearance binarized image is not subjected to pretreatment of removing the background and retaining the foreground, the appearance binarized image containing the background is used for training an appearance defect detection model or actually detecting the appearance defect of the injection molding piece, and the accuracy of the appearance defect detection model can be greatly influenced.

According to the injection molding appearance defect detection method, before the injection molding image is input into the detection model, the pixel points with the gray value mutation in the image are searched in a stepping mode by utilizing the preset detection points, the gray value mutation points influenced by the water gap or the appearance defect are eliminated, and the foreground separation boundary line in the image is determined according to at least two effective gray value mutation points, so that the appearance binarized image is subjected to the pretreatment of removing the background, the image input into the model is more effective, interference factors are fewer, and the accuracy of the injection molding appearance defect detection can be improved.

Based on the foregoing embodiments, as an optional embodiment, at least a first search point set and a second search point set are included in the plurality of preset detection points;

Specifically, the plurality of preset detection points determined on the appearance binarized image at least include a first search point set and a second search point set, each of which includes at least two preset detection points. And the moving direction of all preset detection points in the first search point set when the gray value mutation points are searched in a stepping way is a first direction, and the moving direction of all preset detection points in the second search point set when the gray value mutation points are searched in a stepping way is a second direction opposite to the first direction.

Therefore, when the gray value abrupt change point is searched stepwise by using the preset detection points, that is, two groups of preset detection points perform step search on the pixel value with the gray value abrupt change, and the moving search directions of the two groups of preset detection points are opposite. Along with the continuous stepping movement of the two groups of preset detection points, at least two opposite boundaries of the injection molding piece to be detected in the appearance binarized image are searched.

According to the method for detecting the appearance defects of the injection molding piece, provided by the invention, the plurality of preset detection points are divided into the first search point set and the second search point set, so that step search can be carried out on gray value mutation points in the appearance binarized image from two directions, and at least two opposite boundaries of the injection molding piece to be detected in the appearance binarized image are searched.

Based on the above embodiments, as an optional embodiment, the arrangement directions of the first search point set and the second search point set are both perpendicular to the width direction or the length direction of the appearance binarized image;

Fig. 2 is a schematic diagram of a step-by-step searching gray value mutation point at a preset detection point, as shown in fig. 2, where the preset detection points include a first search point set and a second search point set, the first search point set is composed of five preset detection points arranged in a direction perpendicular to the width direction of the appearance binarized image on the left, and the second search point set is composed of five preset detection points arranged in a direction perpendicular to the width direction of the appearance binarized image on the right.

Correspondingly, the directions of the five preset detection points in the first search point set for step-by-step searching gray value abrupt points are the left directions parallel to the width direction of the appearance binarized image, and the directions of the five preset detection points in the second search point set for step-by-step searching gray value abrupt points are the right directions parallel to the width direction of the appearance binarized image.

Optionally, the arrangement directions of the first search point set and the second search point set are perpendicular to the length direction of the appearance binarized image, and correspondingly, the direction of step-by-step searching gray value mutation points of all preset detection points in the first search point set is an upward direction parallel to the length direction of the appearance binarized image; the direction of step-by-step searching gray value mutation points of all preset detection points in the second search point set is the downward direction parallel to the length direction of the appearance binarized image.

It will be appreciated that the width and length directions of the appearance binarized image are relative.

According to the method for detecting the appearance defects of the injection molding piece, the plurality of preset detection points in the first search point set or the second search point set are respectively arranged side by side, and the arrangement directions are perpendicular to the width/length direction of the appearance binarized image, so that when the preset detection points search gray value mutation points in a stepping mode, gray value mutation points of the foreground and the background in the width/length direction in the appearance binarized image can be searched to the greatest extent, and therefore foreground separation boundary lines are accurately determined, and more accurate background removal is carried out on the appearance binarized image.

Based on the foregoing embodiments, as an optional embodiment, for any preset detection point, determining a search point set to which the any preset detection point belongs to determine a step direction when searching for a gray value abrupt change point in a step manner, where searching for the gray value abrupt change point in a step manner by using the preset detection point includes:

Specifically, as shown in fig. 2, taking a first preset detection point a in the first search point set as an example.

When the preset detection point A is utilized to search the gray value mutation points in a stepping way, firstly, a search point set to which the preset detection point A belongs is judged, and the preset detection point A belongs to a first search point set, so that the stepping direction of the preset detection point A when the gray value mutation points are searched in a stepping way is a first direction, namely, the left direction parallel to the width direction of the appearance binarized image.

After the preset detection point A is determined on the appearance binarized image, firstly acquiring the gray value of the pixel point where the preset detection point A is located, if the gray value of the pixel point where the preset detection point A is located is 1, controlling the preset detection point A to perform step search by taking one pixel as a step length in the left step direction, and judging whether the gray value of the pixel point where the preset detection point A is located is suddenly changed or not.

After the gray value of the pixel point where the preset detection point A is located in the last step is set to be 1, and one pixel is moved step by step, if the gray value of the pixel point where the preset detection point A is located is set to be 0, the representative gray value is suddenly changed, and if the gray value is kept to be 1, the representative gray value is not suddenly changed.

If the gray value of the pixel point where the preset detection point A is located is not suddenly changed, the preset detection point A is controlled to continue to search in a left stepping mode until the gray value of the pixel point where the preset detection point A is located is suddenly changed.

If the gray value of the pixel point where the preset detection point A is located is suddenly changed, the preset detection point A is controlled to finish the step search, and the pixel point where the gray value is suddenly changed after the step search is finished is used as a gray value suddenly changed point.

It can be understood that for each preset detection point, step search can be performed on the appearance binarized image according to the above steps to search for a plurality of gray value abrupt points.

According to the method for detecting the appearance defects of the injection molding piece, when the gray value of the pixel point where the preset detection point is located is 1, the preset detection point is controlled to conduct fine search by taking 1 pixel as a step length, and the gray value of the pixel point where the preset detection point is controlled to conduct continuous search until the pixel point where the gray value is suddenly changed is searched, so that a foreground separation boundary line is accurately determined, and more accurate background removal is conducted on an appearance binary image.

Based on the foregoing embodiments, as an optional embodiment, the selecting at least two valid gray value abrupt points from all gray value abrupt points includes:

And if the gray value mutation point searched by using the step search of the preset detection point is an effective gray value mutation point, the gray value mutation point is represented to be just positioned at the boundary between the foreground and the background in the appearance binary image. Correspondingly, gray value abrupt points positioned on boundaries such as appearance defects and water gaps on the injection molding piece are invalid gray value abrupt points, and the gray value abrupt points need to be eliminated in the process of determining the foreground separation boundary line.

Specifically, fig. 3 is a schematic diagram of a step-by-step searching for a gray value mutation point at a preset detection point, as shown in fig. 2 and 3, where the first search point set includes five preset detection points A, B, C, D, E, after the step-by-step searching is performed at the five preset detection points, the obtained gray value mutation point a, c, d, e is relatively located on the same straight line, and the gray value mutation point b is located at a larger position difference from other gray value mutation points.

Further, a first slope between the gray value abrupt point b and the gray value abrupt point a is calculated, a second slope between the gray value abrupt point b and the gray value abrupt point c is also calculated, and if the directions of the first slope and the second slope are inconsistent, the gray value abrupt point b is used as an invalid gray value abrupt point.

The same method is used for calculating a first slope between the gray value abrupt point c and the gray value abrupt point b, and also calculating a second slope between the gray value abrupt point c and the gray value abrupt point d, wherein the direction of the first slope is consistent with that of the second slope, and the gray value abrupt point c is not used as an invalid gray value abrupt point. Similarly, after the gray value mutation point d is calculated, the first slope is the same as the second slope, and the gray value mutation point d is not an invalid gray value mutation point.

After all the invalid gray value mutation points in the gray value mutation points are determined, the invalid gray value mutation points are removed, and the rest gray value mutation points are the valid gray value mutation points.

In this embodiment, when a border line between the foreground and the background on the appearance binary image is searched in a step mode, 5 preset detection points are selected in total, and in gray value abrupt change points obtained after the step search, the gray value abrupt change point b is not in the same straight line with other gray value abrupt change points due to the influence of a water gap or appearance defect on an injection molding piece to be detected.

According to the method for detecting the appearance defects of the injection molding piece, firstly, the slope between the adjacent gray value mutation points is calculated, and whether the preset detection point is affected by a water gap or the appearance defects in the step search process or not is judged by judging whether the slope between any gray value mutation point and the adjacent two gray value mutation points is consistent, and the affected gray value mutation points are removed, so that the interference and adverse effects of the water gap or the appearance defects on accurately determining a foreground separation boundary line are effectively removed.

Based on the foregoing embodiment, as an optional embodiment, the determining a foreground separation boundary line based on the valid gray value mutation point includes:

The least squares method (The Least Square Method) refers to a method of finding the best matching function of the data by minimizing the sum of squares of the errors.

Specifically, after at least two effective gray value abrupt points are selected from all gray value abrupt points, straight line fitting is carried out on all effective gray value abrupt points by using a least square method to obtain a fitted straight line, and the straight line where the two effective gray value abrupt points with the shortest vertical distance and the shortest next-shortest vertical distance to the fitted straight line are located is used as a foreground separation boundary line.

Fig. 4 is a third schematic diagram of a preset detection point step-by-step searching gray value mutation point provided by the present invention, fig. 5 is a fourth schematic diagram of a preset detection point step-by-step searching gray value mutation point provided by the present invention, as shown in fig. 4 and fig. 5, gray value mutation points except for the gray value mutation point b are all effective gray value mutation points, and then a straight line fitting is performed on the effective gray value mutation points a, c, d, e based on a least square method to obtain a fitting straight line i.

Further, the vertical distance between the effective gray value abrupt points a and c and the fitting straight line is shortest, so that the straight line where the effective gray value abrupt points a and c are located is used as a foreground separation boundary line l, and the background in the appearance binarized image is removed according to the foreground separation boundary line to obtain the detection binarized image.

Alternatively, when the effective gray value abrupt points selected from all the gray value abrupt points are two, a straight line is fitted, that is, a straight line in which the two effective gray value abrupt points are located, that is, a foreground separation boundary line.

In another embodiment, performing straight line fitting on all the gray value abrupt points based on a least square method to obtain a fitted straight line; and taking the straight line where the two effective gray value abrupt change points with the shortest perpendicular distance from the fitting straight line are located as the foreground separation boundary line.

In general, there are two methods for separating foreground from background in an image: (1) a threshold segmentation method: the pixels in the image are classified according to the gray value and a preset threshold value, so that separation of front background and background is realized, and the method has the advantages of simplicity and easy understanding, but has poor effect on complex image scenes; (2) method of image processing technique: the front background is separated mainly by image processing technologies such as filtering and morphological operation, and the like, so that the method has the advantage of good effect, but needs targeted processing for complex image scenes.

In combination, the two foreground-background separation methods can be applied to laboratory scenes, but cannot exert good performance in complex environments.

Firstly, the injection molding piece to be inspected is provided with a water gap after exiting the injection molding machine, which is determined by the injection molding process. Meanwhile, the water gap does not belong to the range for detecting the appearance defects, and the front surface of the injection molding piece is generally only detected when the injection molding piece is detected. However, the two front background separation methods all adopt the thought of removing the edges of the object, the certain part of the object cannot be subtracted independently, the injection molding parts are various in variety, and the water gap positions of each model are not fixed, so that fixed-point positioning subtraction cannot be performed, the accuracy of front background separation is affected, the effectiveness of images input into the appearance defect detection model is affected, and the accuracy of the appearance defect detection model is greatly affected.

Secondly, many injection molding pieces (such as air conditioner panels) to be detected are different from normal material pieces, certain bending exists at the edges of the injection molding pieces, the shot images obtained after shooting are different in imaging, and the gray scale of the edges can be changed along with the increase of radian. Moreover, the slight angle can lead to larger gray level fluctuation, so that the traditional mode such as threshold segmentation and the like has poor effect in distinguishing the panel from the conveyor belt, and the proper threshold cannot be selected for segmentation due to larger gray level change at the edge position. Either too much segmentation results in the background being framed or too little segmentation results in a portion of the panel image being rejected as background, i.e. the accuracy of the front-background separation is lower.

According to the method for detecting the appearance defects of the injection molding piece, an N-point stepping foreground-background separation method is adopted in the whole, each pixel point on a binary image is searched in a stepping mode, so that gray value mutation points are determined at the pixel point precision, whether slopes between each gray value mutation point and adjacent gray value mutation points are consistent or not is judged, invalid gray value mutation points influenced by water gaps, appearance defects and the like on the injection molding piece to be detected are removed, effective gray value mutation points are utilized for determining foreground separation, the background can be accurately removed in a real shooting environment which is much more complex than that of a laboratory, a more effective binary image is obtained, and therefore the accuracy of detecting the appearance defects of the injection molding piece is improved.

Based on the foregoing embodiments, as an optional embodiment, the inputting the detected binarized image to an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model includes:

The sliding window is to slide in a horizontal and vertical direction from the upper left corner on the detected binary image by a fixed step length (stride) by defining a window with a fixed size, and sequentially traverse the whole detected binary image.

Specifically, the detection binarized image obtained after the background is removed is traversed by using a sliding window, so that the detection binarized image is divided into a plurality of sub-images with the same size as the sliding window. And inputting each sub-image into the appearance defect detection model in turn to obtain an appearance defect detection result output by the appearance defect detection model, wherein the appearance defect detection result comprises a detection result corresponding to each sub-image.

It can be understood that the appearance defect detection model for detecting the sub-images can be obtained by inputting a plurality of history sub-images and a plurality of labels corresponding to the detection results in the production process into any machine vision algorithm such as a CNN convolutional neural network, a GAN generation countermeasure network, an MS-CNN multi-scale convolutional neural network or an SVM support vector machine, and the like, and the invention is not limited to this.

In one embodiment, a resizing operation is performed on the detected binary image before the detected binary image is segmented into a plurality of sub-images using a sliding window.

For example, assume that the size of the sliding window is 1000px×1000px, and a binarized image is detected after removing the backgroundSize/>2500Px x 1500px. Then, the binarized image/>, will be detected using a sliding windowBefore being divided into a plurality of sub-images, the sub-images are divided into a plurality of sub-images by bilinear interpolation function/>The size of the detected binarized image was reset to 8000px x 4000px. And then cutting the detection binarized image with the reset size into 32 sub-images by utilizing a sliding window with the step length of 1000 px.

For appearance defect detection of injection molding parts, the problems that appearance defects are difficult to detect due to the fact that the defect area is too small and samples are unbalanced often exist.

Taking an air conditioner panel as an example of an injection molding to be detected, a main stream detection method in the market is shooting detection and splicing through an area array camera, and the method is not fully applicable to scenes of the air conditioner panel, because the types of defects of the air conditioner panel are quite various, the minimum defects are 0.2mm-0.5mm, and the size of the air conditioner panel is quite large.

If the detected binarized image of the air conditioner panel is directly input in whole image, the appearance defect is difficult to be detected by the appearance defect detection model because the area of the appearance defect is too small and the detected binarized image is too large.

According to the method for detecting the appearance defects of the injection molding piece, the sliding window is utilized to cut the detected binarized image into a plurality of sub-images, so that the size ratio of the appearance defects to the image input into the detection model is remarkably improved, effective information in the image is amplified, and the problem that the accuracy of the appearance defect detection model is reduced due to the fact that the area of the appearance defects is too small and the detected binarized image is too large is avoided from the image level.

Based on the above embodiment, as an optional embodiment, the dividing the detected binarized image into a plurality of sub-images using a sliding window includes:

The preset step length may be set to 900px, 1000px, 1500px, or the like, which is not limited in the present invention.

Specifically, when the detected binarized image is divided into a plurality of sub-images by using the sliding window, the sliding window is controlled to move stepwise on the detected binarized image according to a preset step length so as to preliminarily define dividing lines of the plurality of sub-images according to the sliding window. At this time, the detected binarized image may be directly divided according to the preliminarily defined dividing line, thereby obtaining a plurality of sub-images having the same size as the sliding window.

Based on the above embodiment, as an alternative embodiment, after preliminarily defining the dividing lines of the plurality of sub-images, further includes:

Specifically, after the sliding window is controlled to move step by step on the detected binarized image according to a preset step length so as to preliminarily define dividing lines of a plurality of sub-images according to the sliding window, whether abnormal pixel points with gray values of 0 exist on the preliminarily defined dividing lines is also judged.

When an abnormal pixel point with the gray value of 0 exists on the initially defined dividing line, if the detected binary image is divided according to the initially defined dividing line, the appearance defect in the detected binary image is very likely to be divided, so that erroneous judgment is caused when the area of the appearance defect is calculated, and even the unqualified injection molding is finally erroneous judged as the qualified injection molding.

Therefore, when it is determined that the initially defined dividing line has the abnormal pixel point with the gray value of 0, the position of the dividing line is adjusted in a stepping manner along the direction perpendicular to the initially defined dividing line until the dividing line has no abnormal pixel point, and then the detected binarized image is divided according to the adjusted dividing line.

Further, since the size of the new sub-image obtained by dividing after the position of the dividing line is adjusted according to the step is different from the size of the sub-image obtained by dividing before the position of the dividing line is adjusted according to the step, and the input requirement of the appearance defect detection model is not met, the new sub-image obtained by dividing after the position of the dividing line is preprocessed so that the size of the new sub-image is the same as the size of the sub-image obtained by dividing before the position of the dividing line is adjusted according to the step.

In an embodiment, the step-by-step adjustment of the position of the dividing line is followed by dividing to obtain a new sub-image for preprocessing, which includes: and filling the new sub-image into a filling sub-image with the same length and width by using a pixel point with the gray value of 1, and uniformly adjusting the size of the filling sub-image to be the size of the sub-image obtained by dividing before the position of the step-by-step adjustment dividing line if the size of the filling sub-image is different from the size of the sub-image obtained by dividing before the position of the step-by-step adjustment dividing line, so as to obtain the preprocessed sub-image.

For example, assuming that the size of the sliding window is 1000px×1000px, the size of the sub-image divided before the position of the dividing line is adjusted stepwise is 1000px×1000px. If the size of the new sub-image obtained by dividing after the position of the dividing line is adjusted in a stepping way is 1020px multiplied by 1000px, filling the new sub-image with pixel points with gray value of 1, filling the new sub-image into 1020px multiplied by 1020px, uniformly reducing the 1020px multiplied by 1020px filling sub-image into sub-images with the size of 1000px multiplied by 1000px, obtaining the preprocessed sub-image, and taking the preprocessed sub-image as the input of the appearance defect detection model.

If the size of the new sub-image obtained by dividing after the position of the dividing line is adjusted in a stepping way is 980px multiplied by 1000px, filling the new sub-image by using the pixel point with the gray value of 1, filling the new sub-image into a filling sub-image with the gray value of 1000px multiplied by 1000px, obtaining the pre-processed sub-image, and taking the pre-processed sub-image as the input of the appearance defect detection model.

In an embodiment, when the position of the dividing line is adjusted in a stepping manner, the position of the dividing line is adjusted in an outward stepping manner based on the principle of enlarging the size of the sub-image, so as to avoid that the cut sub-image omits the appearance defect feature on the image.

According to the method for detecting the appearance defects of the injection molding piece, a self-adaptive fine adjustment mode is adopted on the whole, and aiming at the special situation that when a sub-image is segmented by a sliding window, certain appearance defects are possibly segmented into two halves so as to influence the qualification judgment of a final product, after a segmentation line is primarily defined by the sliding window, whether the boundary frame of the sliding window has gray transition on the segmentation line is judged, if so, the size of the window is self-adaptively adjusted, the adjustment is stopped until the gray transition is not existing on the segmentation line, and the segmentation line without the gray transition is utilized for segmenting the detected binary image, so that the misjudgment situation caused by the fact that the long-strip-shaped defect is cut into two halves can be effectively avoided.

Based on the above embodiment, as an optional embodiment, the preset step size is smaller than the size of the sliding window.

Specifically, since the preset step length of the sliding window is smaller than the size of the sliding window, after the sliding window is utilized to cut and divide the detected binary image into a plurality of sub-images, the obtained plurality of sub-images all contain repeated areas, and the situation that the cut sub-images miss the appearance defect features on the detected binary image can be effectively avoided to a certain extent.

According to the method for detecting the appearance defects of the injection molding piece, disclosed by the invention, for some appearance defects (such as black points) which are small in area and easy to ignore in the process of dividing and detecting the binarized image, an overlapping step length sliding window dividing method is introduced, so that no image information loss condition can occur no matter what size of image is cut, more production and appearance defect detection scenes can be adapted, and the accuracy of appearance defect detection is improved.

The N-point type stepping front background separation method and the sliding window segmentation method both belong to the data enhancement algorithm for enhancing the image input to the appearance defect detection model, irrelevant image information of the appearance defect detection model in the training process and the actual detection process can be removed, and the area occupation ratio of an appearance defect area in the detection image is improved, so that the accuracy of appearance defect detection of an injection molding part is improved.

Based on the foregoing embodiment, as an optional embodiment, the appearance defect detection result is a result image obtained by labeling an appearance defect existing in the appearance binarized image with a target frame;

The appearance defect detection result at least comprises one target frame for marking the existing appearance defects, and if the appearance defects exist in a plurality of ways, the appearance defect detection result correspondingly comprises a plurality of target frames for marking the appearance defects.

Specifically, after the detected binarized image or the plurality of images after the segmentation of the detected binarized image are input into the appearance defect detection model, the appearance defect detection model detects appearance defects existing in the image, the detected appearance defects are marked by utilizing the target frame, and a result image marked by the target frame for the appearance defects existing in the external binarized image is output.

After the appearance defect detection result is obtained, the appearance defects detected by the appearance defect detection model are judged and screened one by one, and then whether the appearance detection result of the injection molding to be detected is qualified or unqualified is determined.

If the area of any appearance defect in the appearance defect detection result is larger than the preset area, determining that the injection molding piece to be detected is unqualified.

If any appearance defect exists in the appearance defect detection result, the area of the appearance defect is smaller than or equal to the first preset area but larger than the second preset area, and other appearance defects with unlimited areas exist in the preset range of the appearance defect, determining that the injection molding to be detected is unqualified.

And if the areas of all the appearance defects in the appearance defect detection result are smaller than or equal to the second preset area, confirming that the injection molding to be detected is qualified.

It is understood that the second predetermined area is smaller than the first predetermined area.

For example, let the injection molding to be detected be an air conditioner panel, the first preset area be 1The second preset area is 0.5The preset range is 10/>。

For any detected black point, if the calculated area of the black point is larger than 1Directly judging the air conditioner panel as an unqualified product; if the calculated black spot area is less than or equal to 1/>But greater than 0.5/>Further judging that 10/>, taking the black point as the center of the circleWhether other appearance defects exist in the range or not, and if the other appearance defects exist, judging that the air conditioner panel is a disqualified product no matter how much the area of the other appearance defects is; if the area of each appearance defect obtained by calculation is less than or equal to 0.5/>And judging the air conditioner panel as a qualified product.

It should be noted that, for the specific settings of the first preset area, the second preset area and the preset range, the specific settings may be determined comprehensively according to factors such as the specification parameters of the injection molding to be detected, the manufacturing process of the injection molding to be detected, the accuracy requirements of appearance detection, and the like, which is not limited in the present invention.

It can be understood that if the detected binary image or the sub-image input to the appearance defect detection model is subjected to preprocessing such as size reset to affect the calculation of the real area of the appearance defect, when the real area of the appearance defect is calculated, a corresponding recovery operation is performed, so that the calculated area accords with the real area of the appearance defect on the injection molding.

In an embodiment, for calculating the area of the appearance defect in a certain target frame, by traversing each pixel point in the target frame, counting the number of the pixel points with the gray value of the minimum gray value in the target frame, and determining the area of the appearance defect according to the number.

Generally, due to the single laboratory environment, many manufacturers will not perform secondary filtration and selection, i.e. post-treatment, after selecting the currently mainstream detection method to obtain the detection result, so that a lot of errors will be introduced, resulting in poor detection effect.

In addition, in the production process of the injection molding, there are many injection molding with different types and specifications, and the acceptability of the injection molding with different types and specifications for the appearance defects is different, and if the appearance defect detection result is not screened for the second time, the production of the injection molding with different types cannot be adapted.

According to the method for detecting the appearance defects of the injection molding piece, after the appearance defect detection results output by the appearance defect detection model are obtained, the proper first preset area, second preset area and preset range are set according to different acceptability of the injection molding piece with different models and specifications for the appearance defects, so that secondary screening is carried out on the appearance defect detection results, the qualified results of final appearance defect detection of the injection molding piece are obtained, the method can adapt to complex non-laboratory environments, and the accuracy of appearance defect detection of the injection molding piece is improved.

In one embodiment, the appearance defect detection model, when pre-trained, first divides the data sample set into a training set and a verification set.

For the data in the data sample set, the foreground and background separation and the sliding window segmentation area are used for carrying out data enhancement on the image, the network overfitting can be prevented by means of random horizontal and vertical overturning, random brightness and brightness enhancement and the like, then the image after the data enhancement is output to an appearance defect detection model based on a BP algorithm for training, and model weight parameters are saved.

And then calculating the false detection rate, the omission factor and the accuracy of the appearance defect detection model, wherein the formula is as follows:

；

Wherein, Is to correctly identify the number of defective pictures,/>Is the number of wrongly identified defective pictures,/>Is to correctly identify the number of non-defective pictures,/>Is the number of erroneously identified non-defective pictures.

Optionally, the appearance defect detection model is deployed into an injection molding appearance defect detection device through C++ language, so that model reasoning is realized.

In order to better demonstrate the method for detecting the appearance defects of the injection molding provided by the invention, an embodiment using the injection molding as an air conditioner panel is provided below to illustrate the complete process of the invention when detecting the appearance defects of the injection molding. It should be noted that this embodiment is merely an illustration of the method for detecting an appearance defect of an injection molding provided by the present invention, and should not be construed as limiting the present invention.

Fig. 6 is a second flow chart of the method for detecting appearance defects of injection molding provided by the present invention, and fig. 7 is a flow chart of the method for detecting appearance defects of air-conditioning panels provided by the present invention, as shown in fig. 6 and 7, when the air-conditioning panels to be detected are transported by the conveyor belt to reach a preset detection position, the position sensor senses the incoming material, the shooting device is opened by the injection molding appearance detecting device, the air-conditioning panels are shot, and the shot images of the air-conditioning panels are acquired and transmitted to the detecting system. After the detection system receives the shot image, the shot image is stored in the Bitmap.

The detection system at the rear end pre-processes the image before detecting the appearance defect of the shot image, and comprises the following steps:

(1) Binarizing the shot image to obtain an appearance binarized image;

(2) Determining a plurality of preset detection points on the appearance binarization image, searching gray value mutation points of the air conditioner panel on the appearance binarization image by utilizing the prediction detection points in a stepping way, determining a foreground separation boundary line for separating a foreground from a background according to effective gray value mutation points in the gray value mutation points, cutting the appearance binarization image according to the foreground separation boundary line, and removing a background part in the appearance binarization image to obtain a detection binarization image;

(3) Moving on the detected binarized image by utilizing a sliding window with a preset step length smaller than the size of the sliding window so as to preliminarily define dividing lines of a plurality of sub-images; judging whether the dividing line has gray transition or not, if so, properly adjusting the position of the dividing line until the dividing line does not have gray transition; dividing the detected binarized image by using the adjusted dividing line to obtain a plurality of sub-images, and preprocessing the sub-images;

(4) And sequentially inputting the preprocessed sub-images into an appearance defect detection model to obtain an appearance defect detection result, wherein the appearance defect detection result is a result image obtained after the appearance defect is marked by using a target frame.

After the detection system at the rear end detects the appearance defect detection result, the display equipment at the front end is controlled to display the appearance defect detection result, the appearance defect detection result is screened, whether the air conditioner panel is qualified or not is displayed, and workers can monitor the appearance detection process of the air conditioner panel in real time in a detection workshop.

The display device at the front end displays the number of air-conditioning panels with qualified detection results, the number of air-conditioning panels with unqualified detection results, the total number of detected air-conditioning panels and the through rate in addition to the detection results of the appearance defects of the single air-conditioning panel. The through rate refers to the ratio of the number of qualified air-conditioning panels detected in unit time to the total number of air-conditioning panels produced.

Fig. 8 is a schematic structural diagram of an apparatus for detecting an appearance defect of an injection molding part according to the present invention, as shown in fig. 8, including but not limited to:

A binarization processing unit 801, configured to perform binarization processing on a shot image of an injection molding to be detected, to obtain an appearance binarized image;

A mutation point searching unit 802, configured to determine a plurality of preset detection points on the appearance binarized image, so as to step-search gray value mutation points by using the preset detection points;

a boundary line determining unit 803 for selecting at least two effective gray value abrupt points from all gray value abrupt points to determine a foreground separation boundary line based on the effective gray value abrupt points;

A background removing unit 804, configured to remove a background in the appearance binarized image according to the foreground separation boundary line, so as to obtain a detection binarized image;

and a defect detection unit 805, configured to input the detected binarized image to an appearance defect detection model, and obtain an appearance defect detection result output by the appearance defect detection model.

It should be noted that, when the device for detecting an appearance defect of an injection molding piece provided by the present invention is specifically operated, the method for detecting an appearance defect of an injection molding piece described in any one of the above embodiments may be executed, and this embodiment will not be described in detail.

According to the injection molding appearance defect detection device provided by the invention, before the injection molding image is input into the detection model, the pixel points with the gray value mutation in the image are searched in a stepping way by utilizing the preset detection points, the gray value mutation points influenced by the water gap or the appearance defect are eliminated, and the foreground separation boundary line in the image is determined according to at least two effective gray value mutation points, so that the appearance binarized image is subjected to the pretreatment of removing the background, the image input into the model is more effective, the interference factors are fewer, and the accuracy of the detection of the appearance defect of the injection molding can be improved.

Fig. 9 is a schematic structural diagram of an electronic device provided by the present invention, and as shown in fig. 9, the electronic device may include: processor (Processor) 910, communication interface (Communications Interface) 920, memory (Memory) 930, and communication bus 940, wherein Processor 910, communication interface 920, and Memory 930 perform communication with each other via communication bus 940. Processor 910 may invoke logic instructions in memory 930 to perform injection molding appearance defect detection methods including, but not limited to: performing binarization processing on a shot image of the injection molding to be detected to obtain an appearance binarized image; determining a plurality of preset detection points on the appearance binarized image, and searching gray value mutation points in a stepping way by utilizing the preset detection points; selecting at least two effective gray value mutation points from all gray value mutation points to determine a foreground separation boundary line based on the effective gray value mutation points; removing the background in the appearance binarization image according to the foreground separation boundary line to obtain a detection binarization image; and inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model.

Further, the logic instructions in the memory 930 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the method for detecting an appearance defect of an injection molded part provided in the above embodiments, the method including, but not limited to, the steps of: performing binarization processing on a shot image of the injection molding to be detected to obtain an appearance binarized image; determining a plurality of preset detection points on the appearance binarized image, and searching gray value mutation points in a stepping way by utilizing the preset detection points; selecting at least two effective gray value mutation points from all gray value mutation points to determine a foreground separation boundary line based on the effective gray value mutation points; removing the background in the appearance binarization image according to the foreground separation boundary line to obtain a detection binarization image; and inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model.

The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. The method for detecting the appearance defects of the injection molding piece is characterized by comprising the following steps of:

2. The method of claim 1, wherein determining a plurality of predetermined detection points on the appearance binarized image comprises:

3. The method for detecting the appearance defect of the injection molding according to claim 2, wherein at least a first search point set and a second search point set are included in the plurality of preset detection points;

4. The method for detecting an appearance defect of an injection molded article according to claim 3, wherein the arrangement direction of the first search point set and the second search point set is perpendicular to the width direction or the length direction of the appearance binarized image;

5. The method for detecting an appearance defect of an injection molded article according to claim 3, wherein, for any preset detection point, a search point set to which the any preset detection point belongs is determined to determine a step direction when a gray value mutation point is searched step by step, and the step searching for the gray value mutation point by using the preset detection point comprises:

6. The method of claim 5, wherein selecting at least two valid gray value discontinuities from all gray value discontinuities comprises:

7. The method of claim 6, wherein determining a foreground separation boundary line based on the valid gray value discontinuity point comprises:

8. The method for detecting an appearance defect of an injection molded article according to claim 1, wherein the inputting the detected binarized image into an appearance defect detection model to obtain an appearance defect detection result output by the appearance defect detection model comprises:

9. The method of claim 8, wherein the dividing the detected binarized image into a plurality of sub-images using a sliding window, comprises:

10. The method of detecting an appearance defect of an injection molded article according to claim 9, further comprising, after preliminarily defining the dividing lines of the plurality of sub-images:

11. The method for detecting an appearance defect of an injection molded article according to claim 9 or 10, wherein the preset step size is smaller than the size of the sliding window.

12. The method according to claim 1, wherein the appearance defect detection result is a result image obtained by labeling appearance defects existing in the appearance binarized image with a target frame;

13. An injection molding appearance defect detection device, characterized by comprising:

14. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of detecting an appearance defect of an injection molded part according to any one of claims 1 to 12 when the computer program is executed by the processor.

15. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the method of detecting an apparent defect of an injection molded article according to any one of claims 1 to 12.