CN115631166A - Product quality detection system and method - Google Patents
Product quality detection system and method Download PDFInfo
- Publication number
- CN115631166A CN115631166A CN202211309983.8A CN202211309983A CN115631166A CN 115631166 A CN115631166 A CN 115631166A CN 202211309983 A CN202211309983 A CN 202211309983A CN 115631166 A CN115631166 A CN 115631166A
- Authority
- CN
- China
- Prior art keywords
- quality detection
- detection model
- management service
- product
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 349
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000002372 labelling Methods 0.000 claims abstract description 33
- 230000007547 defect Effects 0.000 claims description 48
- 238000011156 evaluation Methods 0.000 claims description 45
- 238000007689 inspection Methods 0.000 claims description 44
- 238000012549 training Methods 0.000 claims description 36
- 238000012795 verification Methods 0.000 claims description 24
- 238000012360 testing method Methods 0.000 claims description 23
- 238000010276 construction Methods 0.000 claims description 19
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000011176 pooling Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000013434 data augmentation Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012372 quality testing Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
- General Factory Administration (AREA)
Abstract
The invention discloses a product quality detection system and a method. The system comprises: the method comprises the following steps that a detection terminal collects a product sample image and sends the product sample image to a detection model management service; the detection model management service establishes tasks according to the quality detection model, determines image annotation tasks, and distributes a first preset number of product sample images to the user terminal to perform product sample image annotation; the user terminal feeds back the labeling result of the product sample image to the detection model management service; the detection model management service also determines a quality detection model according to the product sample image and the labeling result, and sends the quality detection model to the detection terminal; the detection terminal also acquires an image of the product to be detected, and determines a quality detection result of the image of the product to be detected based on the quality detection model. The technical scheme solves the problems of high cost, low efficiency and poor accuracy of product quality detection, can reduce the detection cost and ensure the accuracy of quality detection while improving the product quality detection efficiency.
Description
Technical Field
The invention relates to the technical field of computer vision, in particular to a product quality detection system and method.
Background
In industrial manufacturing, product quality testing is an extremely important component, requiring observation, measurement, and testing of one or more quality characteristics of a product. And comparing the detection result with a specified quality requirement to identify whether the product quality is qualified.
In the existing product quality detection process, most detection processes are realized manually, a few detection contents with low detection difficulty and single detection dimension can be finished by using a sensor, and finally, manual spot check and recheck are carried out.
However, the quality inspection of the products manufactured industrially consumes a large amount of human resources, increasing the production cost. The accuracy of product quality detection is judged by human subjectivity, risks such as misjudgment and missed judgment exist, and the accuracy of a visual detection result is difficult to guarantee. Meanwhile, the efficiency of manual detection is low, and the requirement for efficient production is difficult to meet.
Disclosure of Invention
The invention provides a product quality detection system and a product quality detection method, which aim to solve the problems of high product quality detection cost, low efficiency and poor accuracy, can improve the product quality detection efficiency, reduce the detection cost and ensure the accuracy of quality detection.
According to an aspect of the present invention, there is provided a product quality detection system including a detection model management service, at least one detection terminal, and at least one user terminal;
the detection terminal is used for acquiring a product sample image and sending the product sample image to a detection model management service;
the detection model management service is used for constructing tasks according to a preset quality detection model, determining image annotation tasks, and distributing a first preset number of product sample images to the user terminal according to the image annotation tasks so as to label the product sample images;
the user terminal is used for feeding back the labeling result of the product sample image to the detection model management service;
the detection model management service is also used for determining a quality detection model according to the product sample image and the labeling result and sending the quality detection model to the detection terminal;
the detection terminal is further used for obtaining the image of the product to be detected and determining the quality detection result of the image of the product to be detected based on the quality detection model.
According to another aspect of the present invention, there is provided a product quality inspection method performed by a product quality inspection system including an inspection model management service, at least one inspection terminal, and at least one user terminal; the method comprises the following steps:
acquiring a product sample image through a detection terminal, and sending the product sample image to a detection model management service;
constructing a task according to a preset quality detection model through a detection model management service, determining an image annotation task, and distributing a first preset number of product sample images to a user terminal according to the image annotation task so as to label the product sample images;
feeding back the labeling result of the product sample image to the detection model management service through the user terminal;
determining a quality detection model according to the product sample image and the labeling result through a detection model management service, and sending the quality detection model to a detection terminal;
and acquiring an image of the product to be detected through the detection terminal, and determining a quality detection result of the image of the product to be detected based on the quality detection model.
According to the technical scheme of the embodiment of the invention, a product sample image is collected through a detection terminal and is sent to a detection model management service; constructing a task according to a preset quality detection model through a detection model management service, determining an image annotation task, and distributing a first preset number of product sample images to a user terminal according to the image annotation task so as to label the product sample images; feeding back the labeling result of the product sample image to the detection model management service through the user terminal; determining a quality detection model according to the product sample image and the labeling result through a detection model management service, and sending the quality detection model to a detection terminal; and acquiring an image of the product to be detected through the detection terminal, and determining a quality detection result of the image of the product to be detected based on the quality detection model. The scheme solves the problems of high product quality detection cost, low efficiency and poor accuracy, can reduce the detection cost and ensure the accuracy of quality detection while improving the product quality detection efficiency.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic structural diagram of a product quality detection system according to an embodiment of the present invention;
fig. 2 is a flowchart of a product quality detection method according to a second embodiment of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, apparatus, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. According to the technical scheme, the data acquisition, storage, use, processing and the like meet relevant regulations of national laws and regulations.
Example one
Fig. 1 is a schematic structural diagram of a product quality detection system according to an embodiment of the present invention, which is a product quality detection system applicable to a product quality detection scenario, as shown in fig. 1, the product quality detection system includes a detection model management service 110, at least one detection terminal 120, and at least one user terminal 130; the method comprises the following steps:
the detection terminal 120 is configured to collect a product sample image and send the product sample image to the detection model management service 110;
the detection model management service 110 is configured to construct a task according to a preset quality detection model, determine an image annotation task, and allocate a first preset number of product sample images to the user terminal 130 according to the image annotation task, so as to perform product sample image annotation;
the user terminal 130 is configured to feed back the annotation result of the product sample image to the detection model management service 110;
the detection model management service 110 is further configured to determine a quality detection model according to the product sample image and the labeling result, and send the quality detection model to the detection terminal 120;
the detection terminal 120 is further configured to obtain an image of a product to be detected, and determine a quality detection result of the image of the product to be detected based on the quality detection model.
The inspection terminal 120 may be an imaging device, such as an industrial high definition camera, an infrared thermal imager, and the like, for inspecting products at various production stages. The detection terminal 120 may collect the product image from a single shooting angle only, or may collect multi-angle images of the product to synthesize a three-dimensional image of the product. The inspection terminal 120 can acquire a product image during the operation of the production line, and perform defect, color and other aspects inspection on the product image based on the inspection model, so as to realize product quality inspection. The detection terminal 120 may send the acquired product image as a product sample image to the detection model management service 110 in a period of a certain time, so that the detection model management service 110 trains a quality detection model according to the product sample image.
Detection model management service 110 can be a server, cluster of servers, or the like, for implementing detection model management. After receiving the product sample image, the inspection model management service 110 may perform data augmentation on the product sample image to augment the product sample image dataset. The data augmentation modes can include mirror inversion, rotation, brightness transformation, contrast transformation, blurring, saturation change, and the like of the image. The purpose of data augmentation is to enrich the characteristics of the product sample image dataset so that the quality detection model can better generalize image characteristics.
The inspection model management service 110 may preset a quality inspection model construction task according to the quality inspection needs of the product. The quality detection model construction task can comprise information such as input image types, required sample quantity, detection precision, defect type quantity, distribution positions of defects in products, defect sizes and the like. The inspection model management service 110 may generate an image annotation task according to information such as the number of required samples, the number of types of defects, the distribution positions of defects in a product, and the sizes of defects. For example, the detection model management service 110 may randomly select a product image matching the required number of samples as a product sample image in the database of the detection model management service 110 according to the required number of samples. According to the information such as the number of image annotation personnel, the service capability and the like, the product sample image is distributed to each user terminal 130, so that the product sample image is annotated according to the information such as the number of defect types, the distribution positions of the defects in the product, the sizes of the defects and the like. The first preset number may be equal to the required number of samples, for example 3000, or may be a number increased by a preset ratio or difference to be higher than the required number of samples, for example 3000 × 120% = 3600.
The user can label the product sample image through a labeling tool such as a labeling box, a brush and the like provided by the user terminal 130. The user terminal 130 may generate an annotation result of the product sample image according to the annotation operation of the user, and send the annotation result to the detection model management service 110 for training the quality detection model. The labeling result may include information such as a defect position and a defect type of the product sample image.
The detection model management service 110 may use the product sample image as an input of the quality detection model, use the labeling result as an output of the quality detection model, and train the quality detection model to obtain the quality detection model satisfying the requirements of detection accuracy, defect detection type, and the like. After the training is completed, the detection model management service 110 may send the quality detection model to the detection terminal 120, so that the detection terminal 120 performs quality detection on the subsequently acquired product image to obtain a quality detection result of the product image to be detected. Similar to the labeling result, the quality inspection result may include information such as defect type, defect number, and defect position of the product to be inspected.
In this embodiment, optionally, the detection model management service 110 is further configured to perform sample cleaning on a first preset number of product sample images, and distribute a second preset number of cleaned product sample images to the user terminal 130, so as to perform product sample image annotation.
The detection model management service 110 may perform data cleaning on a first preset number of product sample images, and remove duplicate product sample images, so as to avoid that the same pictures are learned during deep learning, which wastes computational resources and storage resources.
In a possible solution, the user terminal 130 is further configured to respond to a quality detection network setting operation of a user, and obtain a construction parameter and a training setting parameter of the quality detection network;
the user terminal 130 is further configured to send the construction parameters and the training setting parameters of the quality detection network to the detection model management service 110 in response to the training submission operation of the user;
the detection model management service 110 is further configured to determine a quality detection network according to the construction parameters, train the quality detection network according to the training setting parameters and the product sample images and the labeling results, and output a quality detection model.
In this embodiment, the user terminal 130 may provide a building tool of the quality detection network for the user, for example, the building tool may include a position setting tool of a hierarchy such as a convolutional layer, a pooling layer, and a full connection layer. The user can customize the quality detection network by setting up a tool. The user terminal 130 may automatically acquire the construction parameters and training setting parameters input or selected by the user when detecting the quality inspection network setting operation. The building parameters may include a hierarchy setting parameter, a hierarchy connection parameter, and the like, and the hierarchy setting parameter may include parameters such as a convolution kernel size of the convolutional layer, a convolution step size, a pooling type of the pooling layer, a pooling step size, and a neuron number of the full connection layer. The hierarchy connection parameters may include input-output relationships between the hierarchies, such as the next hierarchy of convolutional layer 1 to connect pooling layer 1. The training setting parameters may include iteration number setting parameters, learning rate setting parameters, batch setting parameters, activation function setting parameters, optimizer setting parameters, and the like. The user terminal 130 may send the obtained construction parameters and training parameters to the detection model management service 110, so as to obtain a quality detection model through training.
Specifically, the inspection model management service 110 may construct a quality inspection network according to information such as a hierarchical setting parameter and a hierarchical connection parameter. The product sample image is used as the input of the quality detection network, the labeling result is used as the output of the quality detection network, and the detection model management service 110 can train the quality detection network according to the training setting parameters. After the preset number of iterations is reached, the detection model management service 110 may output the quality detection model obtained in the last iteration, or may output the best quality detection model in the last iteration according to the evaluation indexes such as the verification accuracy and the loss rate, for example, the quality detection model with the highest verification accuracy.
According to the scheme, the quality detection network can be customized by a user so as to realize flexible model construction and training, and a quality detection model fitting an actual application scene can be obtained.
On the basis of the scheme, the training setting parameters comprise iteration times and verification result evaluation indexes; the marking result comprises a defect marking frame and a defect type matched with the defect marking frame;
the detection model management service 110 is further configured to output a verification result evaluation index of each iteration according to the quality detection network, the product sample image, the defect labeling frame, and the defect type matched with the defect labeling frame; wherein the verification result evaluation index comprises at least one of accuracy, precision, intersection ratio, recall rate and loss rate;
the detection model management service 110 is further configured to determine a quality detection model according to the verification result evaluation index of each iteration.
The inspection model management service 110 may divide the product sample image into two parts, one part for training and the other part for verification. When the quality inspection network is iteratively trained, the inspection model management service 110 may perform quality inspection on the product sample images in the inspection set according to the quality inspection model obtained each time. And calculating a verification result evaluation index according to the comparison result of each iteration by comparing the quality detection result output by the quality detection model with the corresponding labeling result. According to the comparison result of the defect detection frame and the defect marking frame output by the quality detection model, the detection management service can calculate the intersection ratio. According to the comparison result of the defect identification type of the defect detection frame and the defect type matched with the defect marking frame, the detection management service can calculate the evaluation indexes of the verification results such as accuracy, precision, recall rate, loss rate and the like. After the preset number of iterations is reached, the detection model management service 110 may use the quality detection model obtained from the last iteration as the quality detection model deployed in the detection terminal 120. The detection model management service 110 may also determine a target quality detection model in the past iteration according to the verification result evaluation index, and use the target quality detection model as a quality detection model deployed in the detection terminal 120.
Optionally, the detection model management service 110 is further configured to generate a model evaluation chart according to the verification result evaluation index of each iteration, and send the model evaluation chart to the user terminal 130.
Detection model management service 110 may generate a model evaluation chart, such as a graph, based on the validation result evaluation indicators for each iteration. The detection model management service 110 may send the model evaluation chart to the user terminal 130, so that the user can visually check the model training condition, and further obtain a quality detection model with a better detection effect by modifying a quality detection network, a product sample image data set, and the like.
In a specific example, the detection model management service 110 may draw a loss rate change curve with the number of iterations as an abscissa and the loss rate as an ordinate, so as to visually check whether the quality detection model meets the loss rate requirement. Similarly, the inspection model management service 110 may draw an accuracy rate change curve with the number of iterations as abscissa and the accuracy rate as ordinate to visually check whether the quality inspection model satisfies the accuracy rate requirement. The inspection model management service 110 may draw a confusion matrix according to the recall rate and accuracy to visually check the identification accuracy of each type of defect.
According to the scheme, the verification result evaluation indexes can be visualized, so that a user can know the detection effect of the quality detection model in time, and then the quality detection model is corrected according to the detection effect.
In a preferred embodiment, the detection terminal 120 is further configured to send a product quality detection result to the detection model management service 110 and the user terminal 130;
the user terminal 130 is further configured to respond to a re-judgment result determination operation of the user and send a re-judgment result to the detection model management service 110;
the detection model management service 110 is further configured to determine a test evaluation result of the quality detection model according to a second preset number of product quality detection results and a re-judgment result matched with the product quality detection results;
the inspection model management service 110 is further configured to generate quality inspection model update information according to the test evaluation result, and send the quality inspection model update information to the user terminal 130 and the inspection terminal 120, so as to retrain the quality inspection model.
After the quality of the product image to be detected is detected, the detection terminal 120 may send the product quality detection result to be reviewed, such as the product quality detection result in the commissioning phase, the product quality detection result in the periodic spot check, or the product quality detection result in the specified defect type, to the detection model management service 110 and the user terminal 130. The user can re-judge the product quality detection result through the user terminal 130. In response to the re-judgment result determination result of the user, the user terminal 130 may transmit the re-judgment result to the detection model management service 110 for model evaluation.
The inspection model management service 110 may compare the product quality inspection result with the re-judgment result, and determine a test evaluation result of the quality inspection model according to the comparison result. Similar to the evaluation index of the verification result, the test evaluation result can represent the evaluation result through indexes such as accuracy, precision, intersection ratio, recall rate, loss rate and the like obtained in the test stage. And if the test evaluation result does not meet the preset model operation condition, determining the update information of the quality detection model according to the test evaluation result. The quality detection model update information may include update information such as a network structure, training settings, and data samples. The inspection model management service 110 may determine that an update target is required based on the test evaluation result.
If the data sample needs to be updated, the testing model managing service 110 may transmit the quality testing model update information to the testing terminal 120, and the testing terminal 120 may re-collect the product sample image or provide a product sample image that satisfies the preset requirements. Meanwhile, the detection model management service 110 may also send the quality detection model update information to the user terminal 130, so that the user terminal 130 updates the construction parameters and the training setting parameters according to the new product sample image data set.
If the network structure needs to be updated, the test model management service 110 may send the quality test model update information to the user terminal 130, so that the user may modify the construction parameters and the training setting parameters through the user terminal 130. Meanwhile, the inspection model management service 110 may also send the quality inspection model update information to the inspection terminal 120 to reasonably arrange the inspection tasks in the retraining process of the quality inspection model.
The scheme can update the quality detection model according to the re-judgment result, and is beneficial to adapting to wider quality detection application scenes, thereby realizing self-adaptive quality detection.
According to the technical scheme, a product sample image is collected through a detection terminal, and the product sample image is sent to a detection model management service; constructing a task according to a preset quality detection model through a detection model management service, determining an image annotation task, and distributing a first preset number of product sample images to a user terminal according to the image annotation task so as to label the product sample images; feeding back the labeling result of the product sample image to the detection model management service through the user terminal; determining a quality detection model according to the product sample image and the labeling result through a detection model management service, and sending the quality detection model to a detection terminal; and acquiring an image of the product to be detected through the detection terminal, and determining a quality detection result of the image of the product to be detected based on the quality detection model. The scheme solves the problems of high product quality detection cost, low efficiency and poor accuracy, can reduce the detection cost and ensure the accuracy of quality detection while improving the product quality detection efficiency.
Example two
Fig. 2 is a flowchart of a product quality detection method according to a second embodiment of the present invention, and as shown in fig. 2, the method includes:
s210, collecting a product sample image through a detection terminal, and sending the product sample image to a detection model management service.
In the scheme, the method is executed by a product quality detection system, and the product quality detection system comprises a detection model management service, at least one detection terminal and at least one user terminal.
S220, constructing a task according to a preset quality detection model through a detection model management service, determining an image annotation task, and distributing a first preset number of product sample images to a user terminal according to the image annotation task so as to label the product sample images.
And S230, feeding back the labeling result of the product sample image to the detection model management service through the user terminal.
S240, determining a quality detection model according to the product sample image and the labeling result through the detection model management service, and sending the quality detection model to the detection terminal.
And S250, acquiring the image of the product to be detected through the detection terminal, and determining the quality detection result of the image of the product to be detected based on the quality detection model.
In this scenario, optionally, the method further includes:
responding to the quality detection network setting operation of a user through a user terminal, and acquiring construction parameters and training setting parameters of the quality detection network;
responding to the training submission operation of the user through the user terminal, and sending the construction parameters and the training setting parameters of the quality detection network to the detection model management service;
and determining a quality detection network according to the construction parameters through the detection model management service, training the quality detection network according to the training setting parameters according to the product sample images and the labeling results, and outputting a quality detection model.
On the basis of the above scheme, optionally, the training setting parameters include iteration times and verification result evaluation indexes; the marking result comprises a defect marking frame and a defect type matched with the defect marking frame; the method further comprises the following steps:
outputting verification result evaluation indexes of each iteration through a detection model management service according to a quality detection network, a product sample image, a defect marking frame and a defect type matched with the defect marking frame; wherein the verification result evaluation index comprises at least one of accuracy, precision, intersection ratio, recall rate and loss rate;
and evaluating the indexes according to the verification result of each iteration through the detection model management service, and determining the quality detection model.
In a preferred embodiment, after the image of the product to be detected is obtained by the detection terminal and the quality detection result of the image of the product to be detected is determined based on the quality detection model, the method further includes:
sending a product quality detection result to a detection model management service and a user terminal through the detection terminal;
determining operation by responding to a re-judgment result of a user through a user terminal, and sending the re-judgment result to a detection model management service;
determining a test evaluation result of the quality detection model according to a second preset number of product quality detection results and a re-judgment result matched with the product quality detection results through the detection model management service;
and generating quality detection model updating information according to the test evaluation result through the detection model management service, and sending the quality detection model updating information to the user terminal and the detection terminal so as to retrain the quality detection model.
The product quality detection method provided by the embodiment of the invention can be executed by the product quality detection system provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution system.
It should be understood that various forms of the flows shown above, reordering, adding or deleting steps, may be used. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solution of the present invention can be achieved.
The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A product quality detection system is characterized by comprising a detection model management service, at least one detection terminal and at least one user terminal;
the detection terminal is used for acquiring a product sample image and sending the product sample image to a detection model management service;
the detection model management service is used for constructing tasks according to a preset quality detection model, determining image annotation tasks, and distributing a first preset number of product sample images to the user terminal according to the image annotation tasks so as to label the product sample images;
the user terminal is used for feeding back the labeling result of the product sample image to the detection model management service;
the detection model management service is also used for determining a quality detection model according to the product sample image and the labeling result and sending the quality detection model to the detection terminal;
the detection terminal is further used for obtaining the image of the product to be detected and determining the quality detection result of the image of the product to be detected based on the quality detection model.
2. The system according to claim 1, wherein the inspection model management service is further configured to perform sample cleaning on a first preset number of product sample images, and distribute a second preset number of cleaned product sample images to the user terminal for product sample image annotation.
3. The system according to claim 1, wherein the user terminal is further configured to obtain a construction parameter and a training setting parameter of the quality detection network in response to a quality detection network setting operation of a user;
the user terminal is also used for responding to the training submission operation of the user and sending the construction parameters and the training setting parameters of the quality detection network to the detection model management service;
and the detection model management service is also used for determining a quality detection network according to the construction parameters, training the quality detection network according to the training setting parameters and the labeling results according to the product sample images and the labeling results, and outputting a quality detection model.
4. The system of claim 3, wherein the training setting parameters include a number of iterations and a validation result evaluation index; the marking result comprises a defect marking frame and a defect type matched with the defect marking frame;
the detection model management service is also used for outputting verification result evaluation indexes of each iteration according to the quality detection network, the product sample image, the defect marking frame and the defect type matched with the defect marking frame; wherein the verification result evaluation index comprises at least one of accuracy, precision, intersection ratio, recall rate and loss rate;
and the detection model management service is also used for determining a quality detection model according to the verification result evaluation index of each iteration.
5. The system of claim 1, wherein the inspection model management service is further configured to generate a model evaluation chart according to the verification result evaluation index of each iteration, and send the model evaluation chart to the user terminal.
6. The system of claim 1, wherein the test terminal is further configured to send the product quality test result to the test model management service and the user terminal;
the user terminal is also used for responding to the re-judgment result determination operation of the user and sending the re-judgment result to the detection model management service;
the detection model management service is also used for determining a test evaluation result of the quality detection model according to a second preset number of product quality detection results and a re-judgment result matched with the product quality detection results;
and the detection model management service is also used for generating quality detection model updating information according to the test evaluation result and sending the quality detection model updating information to the user terminal and the detection terminal so as to retrain the quality detection model.
7. A product quality detection method is characterized in that the method is executed by a product quality detection system, and the product quality detection system comprises a detection model management service, at least one detection terminal and at least one user terminal; the method comprises the following steps:
acquiring a product sample image through a detection terminal, and sending the product sample image to a detection model management service;
constructing a task according to a preset quality detection model through a detection model management service, determining an image annotation task, and distributing a first preset number of product sample images to a user terminal according to the image annotation task so as to label the product sample images;
feeding back the labeling result of the product sample image to the detection model management service through the user terminal;
determining a quality detection model according to the product sample image and the labeling result through a detection model management service, and sending the quality detection model to a detection terminal;
and acquiring an image of the product to be detected through the detection terminal, and determining a quality detection result of the image of the product to be detected based on the quality detection model.
8. The method of claim 7, further comprising:
responding to the quality detection network setting operation of a user through a user terminal, and acquiring construction parameters and training setting parameters of the quality detection network;
responding to the training submission operation of the user through the user terminal, and sending the construction parameters and the training setting parameters of the quality detection network to the detection model management service;
and determining a quality detection network according to the construction parameters through the detection model management service, training the quality detection network according to the training setting parameters according to the product sample images and the labeling results, and outputting a quality detection model.
9. The method of claim 8, wherein the training setting parameters include a number of iterations and a validation result evaluation index; the marking result comprises a defect marking frame and a defect type matched with the defect marking frame; the method further comprises the following steps:
outputting a verification result evaluation index of each iteration through a detection model management service according to a quality detection network, a product sample image, a defect labeling frame and a defect type matched with the defect labeling frame; wherein the verification result evaluation index comprises at least one of accuracy, precision, intersection ratio, recall rate and loss rate;
and determining a quality detection model according to the verification result evaluation index of each iteration through a detection model management service.
10. The method according to claim 7, wherein after the product image to be detected is obtained by the detection terminal and the quality detection result of the product image to be detected is determined based on the quality detection model, the method further comprises:
sending a product quality detection result to a detection model management service and a user terminal through the detection terminal;
determining operation by responding to a re-judgment result of a user through a user terminal, and sending the re-judgment result to a detection model management service;
determining a test evaluation result of the quality detection model according to a second preset number of product quality detection results and a re-judgment result matched with the product quality detection results through the detection model management service;
and generating quality detection model updating information according to the test evaluation result through the detection model management service, and sending the quality detection model updating information to the user terminal and the detection terminal so as to retrain the quality detection model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211309983.8A CN115631166A (en) | 2022-10-25 | 2022-10-25 | Product quality detection system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211309983.8A CN115631166A (en) | 2022-10-25 | 2022-10-25 | Product quality detection system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115631166A true CN115631166A (en) | 2023-01-20 |
Family
ID=84907498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211309983.8A Pending CN115631166A (en) | 2022-10-25 | 2022-10-25 | Product quality detection system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115631166A (en) |
-
2022
- 2022-10-25 CN CN202211309983.8A patent/CN115631166A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105051781B (en) | NI Vision Builder for Automated Inspection and the method for the long-range quality inspection for product | |
| CN112966772A (en) | Multi-person online image semi-automatic labeling method and system | |
| KR102116264B1 (en) | Main image recommendation method and apparatus, and system | |
| CN109060817B (en) | Artificial intelligence reinspection system and method thereof | |
| CN107004123A (en) | Iterative defect filters out technique | |
| CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
| CN114862832A (en) | Method, device and equipment for optimizing defect detection model and storage medium | |
| TW202139131A (en) | Adaptive learning for image classification | |
| CN114266944B (en) | Rapid model training result checking system | |
| CN109815855B (en) | Electronic equipment automatic test method and system based on machine learning | |
| DE112019005951T5 (en) | Centralized analysis of multiple visual inspection devices | |
| WO2017016886A1 (en) | System and method for providing a recipe | |
| CN115495608A (en) | Defect detection method and system based on model | |
| KR20200096992A (en) | Semiconductor measurement and defect classification using electron microscope | |
| CN103913150B (en) | Intelligent electric energy meter electronic devices and components consistency detecting method | |
| US10134122B2 (en) | System and method of distributed processing for machine-vision analysis | |
| US20230325413A1 (en) | Error Factor Estimation Device and Error Factor Estimation Method | |
| CN111368823B (en) | Pointer type instrument reading identification method and device | |
| CN115631166A (en) | Product quality detection system and method | |
| CN111212434A (en) | WIFI module quality prediction method, device, equipment and storage medium | |
| CN113962558A (en) | Industrial internet platform evaluation method and system based on production data management | |
| US20220074874A1 (en) | Computer-implemented method for analysing measurement data from a measurement of an object | |
| CN114021744A (en) | Method and device for determining residual service life of equipment and electronic equipment | |
| CN112906805A (en) | Image training sample screening and task model training method and device and electronic equipment | |
| CN111666439A (en) | Working method for rapidly extracting and dividing medical image big data aiming at cloud environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |