CN109903287A

CN109903287A - Quality inspection method and device

Info

Publication number: CN109903287A
Application number: CN201910247360.4A
Authority: CN
Inventors: 文亚伟
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2019-03-29
Filing date: 2019-03-29
Publication date: 2019-06-18

Abstract

The present invention provides a quality detection method and device, wherein the method includes: acquiring a product image to be detected; inputting the product image into a preset defect detection model to acquire the category of each pixel in the product image; determining the product according to the category of each pixel The defect information corresponding to the image; the defect information includes: whether there is a defect, the defect type, and the defect location; according to the defect information, the product corresponding to the product image is processed, so that the product image can be automatically detected. The obtained defect information is easy for secondary utilization mining.

Description

Quality determining method and device

Technical field

The present invention relates to technical field of quality detection more particularly to a kind of quality determining methods and device.

Background technique

There are mainly two types of current quality determining methods.One is pure artificial quality inspection modes, observe production dependent on expert Product image in environment provides judgement；One is the artificial quality inspection modes of machine auxiliary, by the matter with certain judgement Check system, which filters out, does not have defective product image, and expert carries out detection judgement to the product image of doubtful existing defects.It is above-mentioned Two methods need expert to participate in, low efficiency, are easy erroneous judgement of failing to judge, and data are difficult to carry out secondary use excavation.

Summary of the invention

The present invention is directed to solve at least some of the technical problems in related technologies.

For this purpose, the first purpose of this invention is to propose a kind of quality determining method, examined in the prior art for solving Inefficient is surveyed, and data are difficult the problem of carrying out secondary use excavation.

Second object of the present invention is to propose a kind of quality detection device.

Third object of the present invention is to propose another quality detection device.

Fourth object of the present invention is to propose a kind of non-transitorycomputer readable storage medium.

5th purpose of the invention is to propose a kind of computer program product.

In order to achieve the above object, first aspect present invention embodiment proposes a kind of quality determining method, comprising:

Obtain product image to be detected；

The product image is inputted into preset defects detection model, obtains the class of each pixel in the product image Not；

According to the classification of each pixel, the corresponding defect information of the product image is determined；The defect information includes: to be No existing defects, defect classification, defective locations；

According to the defect information, the corresponding product of the product image is handled.

Further, the defects detection model includes: deep neural network module and up-sampling treatment module；

The defects detection model is to the treatment process of the product image,

The product image is inputted to the convolutional layer and pond layer of the deep neural network module, obtains the output of pond layer Characteristics of image；

Described image feature is inputted into up-sampling treatment module, obtains the classification of each pixel in the product image.

Further, the deep neural network module is SE-ResNet neural network.

Further, described that the product image is inputted into preset defects detection model, it obtains in the product image Before the classification of each pixel, further includes:

Training data is obtained, includes: greater than the product image of preset quantity and corresponding true in the training data Defect information；

For each product image in the training data, according to the corresponding real defect information of the product image, Determine the true classification of each pixel in the product image；The product image is inputted into defects detection model, described in acquisition The classification of each pixel in product image；

According to the classification and true classification of each pixel in each product image in the training data, lacked to described The coefficient for falling into detection model is adjusted.

Further, described according to the defect information, the corresponding product of the product image is handled, comprising:

According to the defect information, the processing operation to the corresponding product of the product image, the processing operation are determined Including in following operation any one or it is a variety of: control mechanical arm the corresponding product of the product image is handled, Prompt person is handled；

By the storage of the product image, corresponding defect information and processing operation into Production database.

Further, described according to the defect information, after handling the corresponding product of the product image, also Include:

Judge whether the quantity of product image in the Production database meets preset condition；

If the quantity of product image meets the preset condition in the Production database, prompt person is to the production The corresponding defect information of product image carries out manual examination and verification in database, obtains product image that audit passes through and corresponding lacks Fall into information；

According to product image and corresponding defect information that audit passes through, training data is updated, is updated Training data afterwards；

According to updated training data, the defects detection model is trained.

The quality determining method of the embodiment of the present invention, by obtaining product image to be detected；Product image input is pre- If defects detection model, obtain product image in each pixel classification；According to the classification of each pixel, product image is determined Corresponding defect information；Defect information includes: whether existing defects, defect classification, defective locations；According to defect information, to production The corresponding product of product image is handled, and so as to carry out defects detection automatically to product image, improves quality testing efficiency, And the defect information that detection obtains is easy to carry out secondary use excavation.

In order to achieve the above object, second aspect of the present invention embodiment proposes a kind of quality detection device, comprising:

Module is obtained, for obtaining product image to be detected；

Input module obtains in the product image for the product image to be inputted preset defects detection model The classification of each pixel；

Determining module determines the corresponding defect information of the product image for the classification according to each pixel；It is described to lack Sunken information includes: with the presence or absence of defect, defect classification, defective locations；

Processing module, for handling the corresponding product of the product image according to the defect information.

The defects detection model is to the treatment process of the product image,

Further, the deep neural network module is SE-ResNet neural network.

Further, the device further include: the first training module；

The acquisition module, is also used to obtain training data, includes: the product greater than preset quantity in the training data Image and corresponding real defect information；

The acquisition module is also used to for each product image in the training data, according to the product image Corresponding real defect information determines the true classification of each pixel in the product image；Product image input is lacked Detection model is fallen into, the classification of each pixel in the product image is obtained；

First training module, for the classification according to each pixel in each product image in the training data And true classification, the coefficient of the defects detection model is adjusted.

Further, the processing module is specifically used for,

Further, the device further include: judgment module, cue module, update module and the second training module；

The judgment module, for judging whether the quantity of product image in the Production database meets preset condition；

The cue module when quantity for the product image in the Production database meets preset condition, prompts Personnel carry out manual examination and verification to the corresponding defect information of product image in the Production database, obtain the product figure that audit passes through Picture and corresponding defect information；

The update module, product image and corresponding defect information for being passed through according to audit, to training data It is updated, obtains updated training data；

Second training module, for being trained to the defects detection model according to updated training data.

The quality detection device of the embodiment of the present invention, by obtaining product image to be detected；Product image input is pre- If defects detection model, obtain product image in each pixel classification；According to the classification of each pixel, product image is determined Corresponding defect information；Defect information includes: whether existing defects, defect classification, defective locations；According to defect information, to production The corresponding product of product image is handled, and so as to carry out defects detection automatically to product image, improves quality testing efficiency, And the defect information that detection obtains is easy to carry out secondary use excavation.

In order to achieve the above object, third aspect present invention embodiment proposes another quality detection device, comprising: storage Device, processor and storage are on a memory and the computer program that can run on a processor, which is characterized in that the processor Quality determining method as described above is realized when executing described program.

To achieve the goals above, fourth aspect present invention embodiment proposes a kind of computer readable storage medium, On be stored with computer program, which realizes quality determining method as described above when being executed by processor.

To achieve the goals above, fifth aspect present invention embodiment proposes a kind of computer program product, when described When instruction processing unit in computer program product executes, quality determining method as described above is realized.

The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description Obviously, or practice through the invention is recognized.

Detailed description of the invention

Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments Obviously and it is readily appreciated that, in which:

Fig. 1 is a kind of flow diagram of quality determining method provided in an embodiment of the present invention；

Fig. 2 is the flow diagram of another quality determining method provided in an embodiment of the present invention；

Fig. 3 is a kind of structural schematic diagram of quality detection device provided in an embodiment of the present invention；

Fig. 4 is the structural schematic diagram of another quality detection device provided in an embodiment of the present invention；

Fig. 5 is the structural schematic diagram of another quality detection device provided in an embodiment of the present invention；

Fig. 6 is the structural schematic diagram of another quality detection device provided in an embodiment of the present invention.

Specific embodiment

The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.

Below with reference to the accompanying drawings the quality determining method and device of the embodiment of the present invention are described.

Fig. 1 is a kind of flow diagram of quality determining method provided in an embodiment of the present invention.As shown in Figure 1, the quality Detection method includes the following steps:

S101, product image to be detected is obtained.

The executing subject of quality determining method provided by the invention is quality detection device, and quality detection device can be for eventually The hardware devices such as end equipment, server, or the software to be installed on hardware device.In the present embodiment, product figure to be detected As the image that can be shot at each position of production line to product for camera etc., such as handled by various processes The image of product afterwards.In addition, product image to be detected may be any one article that the shootings such as camera obtain Image.

S102, product image is inputted to preset defects detection model, obtains the classification of each pixel in product image.

In the present embodiment, defects detection model may include: deep neural network module and up-sampling treatment module.It is corresponding , defects detection model is to the treatment process of product image, by the convolutional layer of product image input deep neural network module With pond layer, the characteristics of image of pond layer output is obtained；Characteristics of image is inputted into up-sampling treatment module, is obtained in product image The classification of each pixel.

Wherein, deep neural network module can be SE-ResNet neural network.SE-ResNet neural network be On the basis of ResNet neural network, extruding (squeeze) operation, excitation (excitation) operation and reweight are increased Operation, can make full use of the relationship in characteristics of image between different characteristic channel.Wherein, extrusion operation is along space dimension Degree carries out Feature Compression to characteristics of image, each two-dimensional feature channel is become a real number, this real number has global Receptive field, and the feature port number of the characteristics of image of the dimension and input exported matches.Excitation operation, it is each for generating The weight in feature channel.Reweight operation, be by Weight into characteristics of image in the feature in individual features channel.

In the present embodiment, up-sampling treatment module is specifically as follows the treatment process of characteristics of image, to characteristics of image into Row up-sampling, obtains multiple characteristic images identical with product image size, is labeled with each pixel in characteristic image and belongs to respectively The probability of a classification；The classification of maximum probability in multiple characteristic images is determined as the class of the pixel for each pixel Not.

Further, in order to ensure the accuracy of defects detection model, before step 102, the method can be with It include: the process being trained to defects detection model.The training process of defects detection model is specifically as follows, training is obtained Data include: greater than the product image of preset quantity and corresponding real defect information in training data；For training data In each product image the true of each pixel in product image is determined according to the corresponding real defect information of product image Classification；Product image is inputted into defects detection model, obtains the classification of each pixel in product image；According in training data The classification of each pixel and true classification, are adjusted the coefficient of defects detection model in each product image.

It should be noted that defects detection model is by product figure when handling product image in the present embodiment Article of the defects of the picture as a classification, multiple defects correspond to the article of multiple classifications；Do not have in product image defective Region is as background.Therefore, after quality detection device gets training data, being can be according to product image in training data Corresponding real defect information determines the classification of each pixel in product image.Wherein, it is included: whether in real defect information Existing defects, real defect classification and real defect position.

In addition, in order to improve detection speed, the quantity of defects detection model can be it is multiple, be deployed in different clothes respectively It is engaged on device, it, can be in conjunction with the load of multiple defects detection models after quality detection device gets product image to be detected Situation selects one of defects detection model, and product image to be detected is sent to and is deployed with the defects detection model Server, so that the defects detection model handles product image to be detected.

S103, according to the classification of each pixel, determine the corresponding defect information of product image；Defect information includes: whether Existing defects, defect classification, defective locations.

In the present embodiment, using defect as in the case where the article of a classification, quality detection device executes step 103 Process be specifically as follows, by article determined by the classification of pixel each in product image, be determined as in product image lack Fall into classification；According to the pixel position of the same category, defective locations are determined.In addition, if in product image all pixels class It Wei background or be empty, then it represents that defect is not present in product image, indicates that defect is not present in the corresponding product of product image.

S104, according to defect information, the corresponding product of product image is handled.

In the present embodiment, the process that quality detection device executes step 104 is specifically as follows, and according to defect information, determines To the processing operation of the corresponding product of product image, processing operation include in following operation any one or it is a variety of: control Mechanical arm handles the corresponding product of product image, prompt person is handled；Product image, corresponding defect are believed Breath and processing operation storage are into Production database.

Specifically, when defect is not present in product image, quality detection device can be to the corresponding product of product image It does not deal with, directly by the storage of product image, corresponding defect information and processing operation into Production database；Work as product In image when existing defects, quality detection device can control mechanical arm etc. and pick out the corresponding product of product image, with Person who happens to be on hand for an errand person's subsequent processing；Or the corresponding product existing defects of the prompt person product image, so that personnel are according to defect size Situations such as determine whether to pick out corresponding product.

Fig. 2 is the flow diagram of another quality determining method provided in an embodiment of the present invention.As shown in Fig. 2, in Fig. 1 On the basis of illustrated embodiment, after step 104, the quality determining method can with the following steps are included:

S105, judge whether the quantity of product image in the Production database meets preset condition.

In the present embodiment, preset condition for example can be that the quantity of product image is greater than the first amount threshold, product image The increase ratio that increased quantity is greater than the second amount threshold or product image quantity is greater than preset ratio threshold value etc., can be with It is set according to actual needs.

If the quantity of product image meets the preset condition in S106, Production database, prompt person is to production number Manual examination and verification are carried out according to the corresponding defect information of product image in library, obtain product image and corresponding defect that audit passes through Information.

In the present embodiment, there may be partial errors for the corresponding defect information of product image in Production database, therefore, need Prompt person is wanted to carry out manual examination and verification to the corresponding defect information of product image in Production database, to improve Production database The accuracy of the corresponding defect information of middle product image.

S107, the product image and corresponding defect information passed through according to audit, are updated training data, obtain Updated training data.

In the present embodiment, renewal process refers to that the product image passed through and corresponding defect information supplement will be audited Into training data, updated training data is obtained.

S108, according to updated training data, defects detection model is trained.

In the present embodiment, step 105 to step 108 can be repeated.For example, when product image in Production database When quantity is greater than 1000, execute primary；When the quantity of product image in Production database is greater than 5000, execute primary.To Defects detection model can be trained, avoided based on the product image and corresponding defect information in Production database A large amount of training datas are manually obtained, reduce human cost, while also can be improved the acquisition speed of training data, and then improves and lacks Fall into training speed and the accuracy of detection model.

Fig. 3 is a kind of structural schematic diagram of quality detection device provided in an embodiment of the present invention.As shown in Figure 3, comprising: obtain Modulus block 31, input module 32, determining module 33 and processing module 34.

Wherein, module 31 is obtained, for obtaining product image to be detected；

Input module 32 obtains the product image for the product image to be inputted preset defects detection model In each pixel classification；

Determining module 33 determines the corresponding defect information of the product image for the classification according to each pixel；It is described Defect information includes: whether existing defects, defect classification, defective locations；

Processing module 34, for handling the corresponding product of the product image according to the defect information.

Quality detection device provided by the invention can be the hardware devices such as terminal device, server, or set for hardware The software of standby upper installation.In the present embodiment, product image to be detected can for camera etc. at each position of production line to production The image that product are shot, such as the image by various processes treated product.

Further, in conjunction with reference Fig. 4, on the basis of embodiment shown in Fig. 3, the device can also include: One training module 35；

Wherein, the acquisition module 31, is also used to obtain training data, includes: greater than present count in the training data The product image of amount and corresponding real defect information；

The acquisition module 31 is also used to for each product image in the training data, according to the product figure As corresponding real defect information, the true classification of each pixel in the product image is determined；The product image is inputted Defects detection model obtains the classification of each pixel in the product image；

First training module 35, for the class according to each pixel in each product image in the training data Other and true classification is adjusted the coefficient of the defects detection model.

In the present embodiment, using defect as in the case where the article of a classification, determining module 33 specifically can be used for, By article determined by the classification of pixel each in product image, it is determined as the defects of product image classification；According to mutually similar Other pixel position, determines defective locations.In addition, if the classification of all pixels is background or is sky in product image, It then indicates to indicate that defect is not present in the corresponding product of product image there is no defect in product image.

Further, on the basis of the above embodiments, processing module 34 specifically can be used for, according to defect information, really The fixed processing operation to the corresponding product of product image, processing operation include in following operation any one or it is a variety of: control Mechanical arm processed handles the corresponding product of product image, prompt person is handled；Product image, corresponding defect are believed Breath and processing operation storage are into Production database.

Fig. 5 is the structural schematic diagram of another quality detection device provided in an embodiment of the present invention.As shown in figure 5, in Fig. 3 On the basis of illustrated embodiment, the device can also include: judgment module 36, cue module 37, update module 38 and Two training modules 39；

Wherein, the judgment module 36, for judging it is pre- whether the quantity of product image in the Production database meets If condition；

The cue module 37 mentions when the quantity for the product image in the Production database meets preset condition The person of leting others have a look at carries out manual examination and verification to the corresponding defect information of product image in the Production database, obtains the product that audit passes through Image and corresponding defect information；

The update module 38, product image and corresponding defect information for being passed through according to audit, to training number According to being updated, updated training data is obtained；

Second training module 39, for being instructed to the defects detection model according to updated training data Practice.

In the present embodiment, judgment module 36, cue module 37, update module 38 and the second training module 39 can repeat to hold Row above procedure.For example, being executed primary when the quantity of product image in Production database is greater than 1000；Work as Production database When the quantity of middle product image is greater than 5000, execute primary.So as to based on product image in Production database and right The defect information answered is trained defects detection model, avoids manually obtaining a large amount of training datas, reduces human cost, together When also can be improved the acquisition speed of training data, and then improve training speed and the accuracy of defects detection model.

Fig. 6 is the structural schematic diagram of another quality detection device provided in an embodiment of the present invention.The quality detection device Include:

Memory 1001, processor 1002 and it is stored in the calculating that can be run on memory 1001 and on processor 1002 Machine program.

Processor 1002 realizes the quality determining method provided in above-described embodiment when executing described program.

Further, quality detection device further include:

Communication interface 1003, for the communication between memory 1001 and processor 1002.

Memory 1001, for storing the computer program that can be run on processor 1002.

Memory 1001 may include high speed RAM memory, it is also possible to further include nonvolatile memory (non- Volatile memory), a for example, at least magnetic disk storage.

Processor 1002 realizes quality determining method described in above-described embodiment when for executing described program.

If memory 1001, processor 1002 and the independent realization of communication interface 1003, communication interface 1003, memory 1001 and processor 1002 can be connected with each other by bus and complete mutual communication.The bus can be industrial standard Architecture (Industry Standard Architecture, referred to as ISA) bus, external equipment interconnection (Peripheral Component, referred to as PCI) bus or extended industry-standard architecture (Extended Industry Standard Architecture, referred to as EISA) bus etc..The bus can be divided into address bus, data/address bus, control Bus processed etc..Only to be indicated with a thick line in Fig. 6, it is not intended that an only bus or a type of convenient for indicating Bus.

Optionally, in specific implementation, if memory 1001, processor 1002 and communication interface 1003, are integrated in one It is realized on block chip, then memory 1001, processor 1002 and communication interface 1003 can be completed mutual by internal interface Communication.

Processor 1002 may be a central processing unit (Central Processing Unit, referred to as CPU), or Person is specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC) or quilt It is configured to implement one or more integrated circuits of the embodiment of the present invention.

The present invention also provides a kind of non-transitorycomputer readable storage mediums, are stored thereon with computer program, the journey Quality determining method as described above is realized when sequence is executed by processor.

The present invention also provides a kind of computer program products, when the instruction processing unit in the computer program product executes When, realize quality determining method as described above.

In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples It closes and combines.

In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three It is a etc., unless otherwise specifically defined.

Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes It is one or more for realizing custom logic function or process the step of executable instruction code module, segment or portion Point, and the range of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discussed suitable Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be of the invention Embodiment person of ordinary skill in the field understood.

Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for Instruction execution system, device or equipment (such as computer based system, including the system of processor or other can be held from instruction The instruction fetch of row system, device or equipment and the system executed instruction) it uses, or combine these instruction execution systems, device or set It is standby and use.For the purpose of this specification, " computer-readable medium ", which can be, any may include, stores, communicates, propagates or pass Defeated program is for instruction execution system, device or equipment or the dress used in conjunction with these instruction execution systems, device or equipment It sets.The more specific example (non-exhaustive list) of computer-readable medium include the following: there is the electricity of one or more wirings Interconnecting piece (electronic device), portable computer diskette box (magnetic device), random access memory (RAM), read-only memory (ROM), erasable edit read-only storage (EPROM or flash memory), fiber device and portable optic disk is read-only deposits Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can print described program on it or other are suitable Medium, because can then be edited, be interpreted or when necessary with it for example by carrying out optical scanner to paper or other media His suitable method is handled electronically to obtain described program, is then stored in computer storage.

It should be appreciated that each section of the invention can be realized with hardware, software, firmware or their combination.Above-mentioned In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage Or firmware is realized.Such as, if realized with hardware in another embodiment, following skill well known in the art can be used Any one of art or their combination are realized: have for data-signal is realized the logic gates of logic function from Logic circuit is dissipated, the specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile Journey gate array (FPGA) etc..

Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries It suddenly is that relevant hardware can be instructed to complete by program, the program can store in a kind of computer-readable storage medium In matter, which when being executed, includes the steps that one or a combination set of embodiment of the method.

It, can also be in addition, each functional unit in each embodiment of the present invention can integrate in a processing module It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer In read/write memory medium.

Storage medium mentioned above can be read-only memory, disk or CD etc..Although having been shown and retouching above The embodiment of the present invention is stated, it is to be understood that above-described embodiment is exemplary, and should not be understood as to limit of the invention System, those skilled in the art can be changed above-described embodiment, modify, replace and become within the scope of the invention Type.

Claims

1. a quality detection method, is characterized in that, comprises:

Obtain the image of the product to be detected;

Inputting the product image into a preset defect detection model to obtain the category of each pixel in the product image;

According to the category of each pixel, determine the defect information corresponding to the product image; the defect information includes: whether there is a defect, defect type, defect location;

According to the defect information, the product corresponding to the product image is processed.

2. The method according to claim 1, wherein the defect detection model comprises: a deep neural network module and an upsampling processing module;

The processing process of the defect detection model to the product image is:

Input the product image into the convolution layer and the pooling layer of the deep neural network module, and obtain the image features output by the pooling layer;

The image features are input into the upsampling processing module to obtain the category of each pixel in the product image.

3. The method according to claim 2, wherein the deep neural network module is a SE-ResNet neural network.

4 . The method according to claim 1 , wherein, before inputting the product image into a preset defect detection model and acquiring the category of each pixel in the product image, the method further comprises: 5 .

Acquiring training data, the training data includes: product images larger than a preset number and corresponding real defect information;

For each product image in the training data, determine the real category of each pixel in the product image according to the real defect information corresponding to the product image; input the product image into a defect detection model to obtain the product image The category of each pixel in ;

The coefficients of the defect detection model are adjusted according to the category and real category of each pixel in each product image in the training data.

5 . The method according to claim 1 , wherein processing the product corresponding to the product image according to the defect information comprises: 5 .

According to the defect information, a processing operation for the product corresponding to the product image is determined, and the processing operation includes any one or more of the following operations: controlling the robotic arm to process and prompt the product corresponding to the product image personnel to process;

The product images, corresponding defect information, and processing operations are stored in a production database.

6 . The method according to claim 5 , wherein after processing the product corresponding to the product image according to the defect information, the method further comprises: 6 .

Judging whether the number of product images in the production database satisfies a preset condition;

If the number of product images in the production database satisfies the preset condition, prompt the personnel to manually review the defect information corresponding to the product images in the production database, and obtain the product images that have passed the review and the corresponding defect information;

According to the approved product images and the corresponding defect information, update the training data to obtain the updated training data;

The defect detection model is trained according to the updated training data.

7. A quality detection device, characterized in that, comprising:

The acquisition module is used to acquire the image of the product to be detected;

an input module, configured to input the product image into a preset defect detection model, and obtain the category of each pixel in the product image;

A determination module, configured to determine defect information corresponding to the product image according to the category of each pixel; the defect information includes: whether there is a defect, defect category, and defect location;

A processing module, configured to process the product corresponding to the product image according to the defect information.

8. The device according to claim 7, wherein the defect detection model comprises: a deep neural network module and an upsampling processing module;

The processing process of the defect detection model to the product image is:

9. The apparatus according to claim 8, wherein the deep neural network module is a SE-ResNet neural network.

10. The apparatus according to claim 7, further comprising: a first training module;

The acquisition module is further configured to acquire training data, where the training data includes: product images larger than a preset number and corresponding real defect information;

The acquisition module is further configured to, for each product image in the training data, determine the real category of each pixel in the product image according to the real defect information corresponding to the product image; input the product image into the defect a detection model to obtain the category of each pixel in the product image;

The first training module is configured to adjust the coefficient of the defect detection model according to the category and the real category of each pixel in each product image in the training data.

11. The device according to claim 7, wherein the processing module is specifically configured to:

12. The device according to claim 11, further comprising: a judgment module, a prompt module, an update module and a second training module;

The judging module is used for judging whether the number of product images in the production database satisfies a preset condition;

The prompting module is used to prompt the personnel to manually review the defect information corresponding to the product images in the production database when the number of product images in the production database meets the preset conditions, and obtain the product images that have passed the review and the corresponding product images. defect information;

The updating module is used to update the training data according to the approved product images and the corresponding defect information to obtain the updated training data;

The second training module is used for training the defect detection model according to the updated training data.

13. A quality detection device, characterized in that, comprising:

A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the quality detection method according to any one of claims 1-6 when the processor executes the program.

14. A non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the program is executed by a processor, the quality detection method according to any one of claims 1-6 is implemented.

15. A computer program product, which, when executed by an instruction processor in the computer program product, implements the quality detection method according to any one of claims 1-6.