WO2022017197A1

WO2022017197A1 - Intelligent product quality inspection method and apparatus

Info

Publication number: WO2022017197A1
Application number: PCT/CN2021/105434
Authority: WO
Inventors: 刘鹤辉; 王黎明; 李国志; 刘西洋
Original assignee: 南京认知物联网研究院有限公司
Priority date: 2020-07-22
Filing date: 2021-07-09
Publication date: 2022-01-27
Also published as: ZA202107467B; CN111830039A; CN111830039B

Abstract

The present invention relates to the field of industrial intelligence, and more particularly, to an intelligent product quality inspection method and apparatus. The method comprises the following steps: step S1: performing image data collection on a product; step S2: inspecting the product according to collected image data to generate an inspection result; and step S3: carrying out analysis according to the inspection result and processing the product, wherein storage time limits are set for the image data and the inspection result. The apparatus comprises: an image collection module, a storage module, an inspection module, a processing module and an intelligent management module, wherein the image collection module performs image data collection on a product; the inspection module inspects the product according to image data; the processing module reads a product inspection result from the storage module, carries out analysis according to the inspection result, and performs subsequent processing on the corresponding product; and the intelligent management module manages data in the storage module. According to the present invention, the stability and efficiency of product quality inspection can be improved.

Description

An intelligent product quality detection method and device

technical field

The invention relates to the field of industrial intelligence, and more particularly, to an intelligent product quality detection method and device.

Background technique

Visual inspection refers to the use of machines instead of human eyes to make measurements and judgments, and to convert the captured target into image signals through machine vision products (ie, image capture devices, divided into CMOS and CCD), and transmit them to a dedicated image processing system. The pixel distribution, brightness, color and other information are converted into digital signals; the image system performs various operations on these signals to extract the characteristics of the target, and then controls the on-site equipment actions according to the results of the discrimination. Is a valuable mechanism for production, assembly or packaging. Visual inspection is invaluable in its ability to detect defects and prevent defective products from being shipped to consumers. The feature of machine vision inspection is to improve the flexibility and automation of production. In some dangerous working environments that are not suitable for manual work or where artificial vision is difficult to meet the requirements, machine vision is often used to replace artificial vision; at the same time, in the process of mass industrial production, using artificial vision to check product quality is inefficient and not accurate. , The use of machine vision detection methods can greatly improve production efficiency and production automation. Moreover, machine vision is easy to realize information integration, which is the basic technology to realize computer integrated manufacturing. Visual inspection involves taking an image of an object, detecting it and converting it into data for the system to process and analyze, ensuring compliance with its manufacturer's quality standards. Objects that do not meet quality standards are tracked and rejected.

In the field of computer vision, the workflow of product inspection is generally as follows: the camera program takes a picture and saves the photo to a certain location on the disk, the quality inspection model reads the photo from the disk to start the quality inspection, and sends the inspection result to the result processing program after completion. These programs are generally run in series, which leads to the following defects in the existing visual inspection: 1. The dependencies between programs are too high and the stability is poor. The quality detection model relies on the camera program to output photos. Once the disk read and write errors, the whole solution will not work; the result processing program depends on the output of the quality detection model. When there are multiple result processing programs, a problem in one of the links will lead to an error in the overall solution. It needs to be re-tested and adjusted, and the test results cannot be given in time. 2. The photo read and write time is long and the efficiency is low. When faced with large photos, the disk read and write time is long. When a photo is larger than 4M, even if SSD is used, it takes nearly 1s to write or read a single photo to the disk. The efficiency is too low for industrial testing. Therefore, there is an urgent need for an intelligent product quality detection method and device that can improve stability and efficiency.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention provides an intelligent product quality detection method and device, which can improve the stability and efficiency of product quality detection.

The technical scheme adopted by the present invention is:

An intelligent product quality detection method, comprising the following steps:

Step S1: collecting image data for the product;

Step S2: Detecting the product according to the collected image data to generate a detection result;

Step S3: analyze and process the product according to the detection result;

The image data and detection results set a storage time limit.

Specifically, firstly, the image acquisition program is started, and the image acquisition program sends an instruction to make the photographing device capture and generate an image of the product to be detected. After the image is captured and generated, a storage device is used to store the photographed image. Then the inspection program reads the image data from the storage device, detects the product according to the image data, generates inspection results for the inspection program after product inspection, and uses the storage device to store the generated inspection results. Finally, the product processing program reads the product detection result from the storage device, and performs corresponding processing on the product according to the product detection result. Because the inspection program relies on the images captured by the image acquisition program, if the reading and writing time of the image data is too long or an error occurs during the reading and writing process, the inspection program will not be able to output the inspection results normally, and the product inspection will not be carried out. In addition, the product processing procedure is also dependent on the inspection procedure, and the process of generating or reading the inspection results is too slow, which will cause the procedure to be blocked, and the inspection of the product cannot be carried out. Considering the high dependence between various procedures, as long as one of the inspection steps is wrong, the entire product inspection program cannot be carried out. This solution starts from improving the stability of the entire product inspection, and sets a storage time limit for image data and inspection results. Once the image data or inspection results exceed the set time limit during the process of reading and writing, they will be automatically cleared and the next product will be processed. detection. Setting the storage time limit eliminates the problem that the overall image data reading and writing speed is too slow, and secondly ensures that the overall detection result generation and interpretation can be completed quickly, which solves the program blockage from two aspects and improves the overall product quality inspection scheme. stability.

Further, the setting process of the described storage time limit is:

Obtain the image detection time of the product and the image storage time interval;

Calculate the image retention time according to the image detection time and the image storage time interval;

The calculation formula of the image retention time is:

Among them, i is the current image serial number _{of the currently detected product, C k} is the detection time of the k-th image of the currently detected product, and F is the time interval for the image to be stored;

The storage time t is calculated by data fitting according to the image retention time of a plurality of products, and the storage time limit is set as t+d according to the storage time t, and the d is a fixed constant.

Specifically, first obtain the image detection time of the product and the image storage time interval, and then calculate the image retention time of m products through the calculation formula of image retention time, and each product has n images. Collect the image retention time T _ij (i∈(1,m), j∈(1,n)) of each image, obtain the continuous function of the storage time t from the _{collected image retention time T ij, and according to the storage time t} The storage time limit is set as t+d, d is a fixed constant, when i=1, T ₁ =0, 1≤k≤i-1.

Further, the step S1 includes:

Step S1.1: create a first queue;

Step S1.2: image capturing of the product and generating image data, and storing the generated image data in the first queue;

Step S1.3: If there is a product without image capture and image data generation, proceed to step S1.2, otherwise end step S1;

The step S2 includes:

Step S2.1: create a second queue;

Step S2.2: According to the image data stored in the first queue, the corresponding products are detected and the detection results are generated, and the generated detection results are stored in the second queue;

Step S2.3: if the product corresponding to the image data in the first queue has not been detected and the detection result has not been generated, continue to perform step S2.2, otherwise perform step S2.4;

Step S2.4: if step S1 has not ended, wait for the first queue to store the newly generated image data, then continue to execute step S2.1; if step S1 has ended, end step S2;

The step S3 includes:

Step S3.1: analyze according to the detection results in the second cohort;

Step S3.2: processing according to the analysis result;

The steps S1, S2, and S3 are executed in parallel without interfering with each other; if the image data in the first queue remains in the first queue for longer than the storage time limit, the image data exceeding the storage time limit is automatically cleared; The time when the detection results in the second queue are stored in the second queue exceeds the storage time limit, the detection results that exceed the storage time limit are automatically cleared.

Specifically, queues are created according to product information. The queues are of different types. Different types of queues store different data. Problems in one queue will not affect the other queue. At the same time, this facilitates the isolation of the two queues and reduces the The stability and performance problems of image data, especially image data with large pixels, are caused when transferring between programs. The step S1, step S2, and step S3 are performed in parallel, without interfering with each other. Specifically, the image data is generated after the current product is photographed, and then stored in the first queue. If there is a next product, the steps of photographing, generating and storing are repeated. That is, step S1 is repeated until all products are photographed; the detection step is performed simultaneously with the photographing step. When there are product images that have not been read in the first queue, the product images in the first queue are read, and the product images are paired according to the product images. The products are tested, and the test results are stored in the second queue, and the test steps are repeated until all products are tested, that is, step S2 is repeated. Because step S1 and step S2 are performed at the same time, step S2 does not need to wait for all product images to be photographed in step S1 before starting the detection, which saves time. In the same way, the processing step is carried out at the same time as the shooting step and the detection step. When there is an unread product test result in the second queue, the test result in the second queue is read, and the corresponding product is carried out according to the test result. The process is performed until all product detection results are analyzed, that is, step S3 is repeated. Because step S3 does not need to wait for all products to be photographed in step S1, and does not need to wait for step S2 to detect all products before processing, time is saved and the performance of the entire solution is improved. The image data and detection results in the queue have a storage time limit, and the image data and detection results that lead to timeout will be automatically cleared to maintain the stability of the model and avoid system paralysis caused by reading and writing high-pixel images and analyzing complex models.

Further, the step S1.1 is specifically to create a plurality of first queues according to the number of products; the step S2.1 is specifically to create a plurality of second queues according to the number of detection models for detecting the products.

Specifically, multiple products can correspondingly create multiple first and second queues, and each queue can perform steps at the same time, enabling multiple products to perform quality inspection at the same time, effectively utilizing system resources, and improving efficiency. The isolation also ensures the stability of the scheme.

Further, the image data collected in the step S1 is stored in the form of a binary stream, and the step S2 acquires the image data in the form of a binary stream to detect the product.

Specifically, in order to deal with the problem of long time for high-pixel image reading and format conversion, a binary stream-based access interface is defined. The photos captured in step S1 are stored in the cache by calling the storage interface in the form of binary streams, and step S2 is directly obtained. Binary stream, saving the time of image transcoding.

An intelligent product quality detection device, comprising: an image acquisition module, a storage module, a detection module, a processing module and an intelligent management module;

The image collection module collects image data for the product, and the collected image data is stored in the storage module;

The detection module detects the product according to the image data, and the detection result is stored in the storage module, and the image data is the image data stored in the storage module by the image acquisition module;

The processing module reads the product detection result from the storage module, analyzes according to the detection result, and performs subsequent processing on the corresponding product;

The intelligent management module manages the data in the storage module.

Further, the intelligent management module includes: a format detection unit and a storage time limit unit; the format detection unit is used to detect the format of the data in the storage module; the storage time limit unit is used to set the storage time limit of the data in the storage module.

Further, the storage module implements the storage function with a cache.

Specifically, the cache improves the speed of reading and writing images and reading and writing detection results.

Further, the storage module includes: a first storage unit and a second storage unit; the first storage unit is used for storing product images; the second storage unit is used for storing product detection results.

Further, the storage module stores the image data in a binary stream.

Compared with the prior art, the beneficial effects of the present invention are:

(1) The storage module realizes the storage function with a cache, which improves the speed of reading and writing images and reading and writing detection results.

(2) After the product image is taken, the binary stream is directly written into the cache, and then the binary stream is directly read from the cache for detection, which saves the reading and writing time of the image and improves the overall solution performance.

(3) The mutual independence of queues and steps improves the stability of the scheme on the one hand and the performance of the scheme on the other hand.

Description of drawings

Fig. 1 is the flow chart of the present invention;

FIG. 2 is a module relationship diagram of the present invention.

detailed description

The accompanying drawings of the present invention are only used for exemplary illustration, and should not be construed as limiting the present invention. In order to better illustrate the following embodiments, some parts of the drawings may be omitted, enlarged or reduced, which do not represent the size of the actual product; for those skilled in the art, some well-known structures and their descriptions in the drawings may be omitted. understandable.

Example

Fig. 1 is the flow chart of the present invention, as shown in the figure, a kind of intelligent product quality detection method of the present embodiment, comprises the following steps:

Step S1: collecting image data for the product;

Step S3: analyze and process the product according to the detection result;

The image data and detection results set a storage time limit.

Further, the setting process of the described storage time limit is:

The calculation formula of the image retention time is:

According to the image retention time of a plurality of products, the storage time t is calculated by data fitting, and the storage time limit is set to t+d according to the storage time t, and the d is a fixed constant.

Further, the step S1 includes:

Step S1.1: create a first queue;

The step S2 includes:

Step S2.1: create a second queue;

The step S3 includes:

Step S3.1: analyze according to the detection results in the second cohort;

Step S3.2: processing according to the analysis result;

Specifically, queues are created according to product information. The queues are of different types. Different types of queues store different data. Problems in one queue will not affect the other queue. The stability and performance problems of image data, especially image data with large pixels, are caused when transferring between programs. The step S1, step S2, and step S3 are performed in parallel, without interfering with each other. Specifically, the image data is generated after the current product is photographed, and then stored in the first queue. If there is a next product, the steps of photographing, generating and storing are repeated. That is, step S1 is repeated until all products are photographed; the detection step is performed simultaneously with the photographing step. When there are product images that have not been read in the first queue, the product images in the first queue are read, and the product images are paired according to the product images. The products are tested, and the test results are stored in the second queue, and the test steps are repeated until all products are tested, that is, step S2 is repeated. Because step S1 and step S2 are performed at the same time, step S2 does not need to wait for all product images to be photographed in step S1 before starting the detection, which saves time. In the same way, the processing step is carried out at the same time as the shooting step and the detection step. When there is an unread product test result in the second queue, the test result in the second queue is read, and the corresponding product is carried out according to the test result. The process is performed until all product detection results are analyzed, that is, step S3 is repeated. Because step S3 does not need to wait for all products to be photographed in step S1, and does not need to wait for step S2 to detect all products before processing, time is saved and the performance of the entire solution is improved. The image data and detection results in the queue have a storage time limit, and the image data and detection results that lead to timeout will be automatically cleared to maintain the stability of the model and avoid system paralysis caused by reading and writing high-pixel images and analyzing complex models.

An intelligent product quality detection device, FIG. 2 is a module relationship diagram of the present invention, as shown in the figure, including: an image acquisition module, a storage module, a detection module, a processing module and an intelligent management module;

The intelligent management module manages the data in the storage module.

Further, the storage module implements the storage function with a cache.

Further, the storage module stores the image data in a binary stream.

Obviously, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the technical solutions of the present invention, and are not intended to limit the specific embodiments of the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principle of the claims of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

An intelligent product quality detection method is characterized in that, comprises the following steps:

Step S1: collecting image data for the product;

Step S2: Detecting the product according to the collected image data to generate a detection result;

Step S3: analyze and process the product according to the detection result;

The image data and detection results set a storage time limit.
A kind of intelligent product quality detection method according to claim 1, is characterized in that, the setting process of described storage time limit is:

Obtain the image detection time of the product and the image storage time interval;

Calculate the image retention time according to the image detection time and the image storage time interval;

The calculation formula of the image retention time is:

Among them, i is the current image serial number of the currently detected product, C k is the detection time of the k-th image of the currently detected product, and F is the time interval for the image to be stored;

The storage time t is calculated by data fitting according to the image retention time of a plurality of products, and the storage time limit is set as t+d according to the storage time t, and the d is a constant.
The intelligent product quality detection method according to claim 1, wherein the step S1 comprises: step S1.1: creating a first queue;

Step S1.2: image capturing of the product and generating image data, and storing the generated image data in the first queue;

Step S1.3: If there is a product without image capture and image data generation, proceed to step S1.2, otherwise end step S1;

The step S2 includes:

Step S2.1: create a second queue;

Step S2.2: According to the image data stored in the first queue, the corresponding products are detected and the detection results are generated, and the generated detection results are stored in the second queue;

Step S2.3: if the product corresponding to the image data in the first queue has not been detected and the detection result has not been generated, continue to perform step S2.2, otherwise perform step S2.4;

Step S2.4: if step S1 has not ended, wait for the first queue to store the newly generated image data, then continue to execute step S2.1; if step S1 has ended, end step S2;

The step S3 includes:

Step S3.1: analyze according to the detection results in the second cohort;

Step S3.2: processing according to the analysis result;

The steps S1, S2, and S3 are executed in parallel without interfering with each other; if the image data in the first queue remains in the first queue for longer than the storage time limit, the image data exceeding the storage time limit is automatically cleared; The time when the detection results in the second queue are stored in the second queue exceeds the storage time limit, the detection results that exceed the storage time limit are automatically cleared.
An intelligent product quality detection method according to claim 3, characterized in that, the step S1.1 is specifically to create a plurality of first queues according to the number of products; The number of detection models to perform detection creates multiple second queues.
The intelligent product quality detection method according to claim 1, wherein the image data is stored in the form of binary stream.
An intelligent product quality detection device is characterized in that it comprises: an image acquisition module, a storage module, a detection module, a processing module and an intelligent management module;

The image collection module collects image data for the product, and the collected image data is stored in the storage module;

The detection module detects the product according to the image data, and the detection result is stored in the storage module, and the image data is the image data stored in the storage module by the image acquisition module;

The processing module reads the product detection result from the storage module, analyzes according to the detection result, and performs subsequent processing on the corresponding product;

The intelligent management module manages the data in the storage module.
The intelligent product quality detection device according to claim 6, wherein the intelligent management module comprises: a format detection unit and a storage time limit unit; the format detection unit is used to detect the format of the data in the storage module ; The storage time limit unit is used to set the storage time limit of the data in the storage module.
The intelligent product quality detection device according to claim 6, wherein the storage module realizes the storage function by means of a cache.
The intelligent product quality detection device according to claim 7, wherein the storage module comprises: a first storage unit and a second storage unit; the first storage unit is used for storing product images; The second storage unit is used to store the product inspection results.
The intelligent product quality detection device according to claim 6, wherein the storage module stores the image data in a binary stream.